The ultimate guide to SEO | Ethan Smith (Graphite)

[00:00] I think people under-resource SEO a lot of times and over-resource ads. So if you're Zillow, you're going to spend... [00:05] tens of millions of dollars on ads, or if you're eBay, you're going to spend tens of millions of dollars on ads. [00:10] Why would you not have a really great SEO team? The amount of traffic you get is probably equal to that. So if you're going to spend $100 billion on ads, why would you spend... [00:20] $50,000 on SEO that doesn't make sense. [00:32] I interview world-class product leaders and growth experts to learn from their hard-won experiences building and scaling today's most successful companies. [00:40] Today my guest is Ethan Smith. [00:42] Ethan is possibly the smartest person on SEO you will find. Ethan has worked with companies like Masterclass, Thumbtack, Robinhood, Medium, and Honey to develop and execute their SEO strategies. One of my goals with this podcast is to give you tactical and actionable advice for how to build and grow your own product. And SEO is one of the most powerful and least understood growth levers. This episode contains more advice, tools, tactics, and guidance on how to win at SEO [01:12] I have ever come across, and I suspect it will blow your mind. We talk about when to focus on SEO, whether you ever should, talk about all the things you need to get right to win at SEO, when to hire, who to hire, the most useful tools, and so much more. I will now stop talking and get right into it. And with that, I bring you Ethan Smith.

[01:33] This episode is brought to you by Coda. Coda is an all-in-one doc that combines the best of documents, spreadsheets, and apps in one place. I actually use Coda every single day. It's my home base for organizing my newsletter writing. It's where I plan my content calendar, capture my research, and write the first drafts of each and every post. It's also where I curate my private knowledge repository for paid newsletter subscribers. And it's also how I manage the workflow for this very podcast. [02:02] from being a tool that makes teams more productive to one that also helps bring the best practices across the tech industry to life with an incredibly rich collection of templates and guides in the Coda doc gallery, including resources from many guests on this podcast, including Shreyas, Gokul, and Shashir, the CEO of Coda. Some of the best teams out there, like Pinterest, Spotify, Square, and Uber, use Coda to run effectively and have published their templates for anyone to use. [02:32] you're ping-ponging between lots of documents and spreadsheets, make your life better and start using Coda. You can take advantage of a special limited-time offer just for startups. Head over to coda.io slash Lenny to sign up and get a thousand dollar credit on your first statement. That's coda.io slash Lenny to sign up and get a thousand dollars in credit on your account. [03:02] offering powerful self-serve product analytics. If you listen to this podcast, you know that it's really hard to build great product without making compromises. And when it comes to using data, a lot of teams think that they only have two choices, make quick decisions based on gut feelings or make data-driven decisions at a snail's pace. But that's a false choice. You shouldn't have to compromise on speed to get product answers that you can trust. With Mixpanel, there are no trade-offs. Get deep insights at the speed of thought

[03:32] you grow. Mixpanel builds powerful and intuitive product analytics that everyone can trust, use, and afford. Explore plans for teams of every size and see what Mixpanel can do for you at mixpanel.com. And while you're at it, they're hiring. Check out mixpanel.com to learn more. [03:50] Ethan, welcome to the podcast. Thank you for having me. It's my pleasure. So you're the CEO of Graphite, which is one of the top growth agencies in the world. And you guys are especially good at SEO, which to a lot of people feels like this huge dark art. And so I'm really excited to dig into a lot of the nitty gritty of SEO and the tactics and the strategies around winning at SEO. And I suspect this episode is going to end up being very rich and very tactical. [04:20] But before we get into that, [04:22] Could you spend just like a minute talking about how you got into growth, how you got into SEO, and then just a bit about Graphite and what you all do over there? [04:31] Yeah, sure. So, [04:33] My career started in user research and user experience design, so very different from SEO. And I would do user testing and ethnography and things like that. [04:41] and UX workflows. [04:44] So I was working at this commerce company in 2007 and I made what I thought were pretty nice designs and nobody was using them. So I wanted to figure out how to get people to actually come. We were buying ads, but they're really expensive. And so... [04:59] By necessity, I basically, in my spare time, decided to also learn SEO on top of

[05:04] doing user tests and things like that. [05:07] And so I got into it, number one, by we hired an SEO consultant. And so he sort of, and we hired several SEO consultants, but, and several of them were not good, but this one was great. It was Leo Heriano from shopping.com, eBay. So he kind of gave some playbooks and then we took that and did a bunch of testing and most of the tests failed. Some of them succeeded. And so we sort of iterated over about three years and [05:31] Came up with what at the time was the programmatic SEO search page type of a strategy. And then since then, I went into growth and product broadly in SEO as part of what I did. [05:43] And then [05:45] Four years ago, Whirlpool bought the company that I was at before, and then I decided to specialize in SEO. But most of my career actually was in product and growth broadly, and SEO was just a part of that, which I think is interesting because to do SEO well, I think it needs to fit into a broader product context and a lot of what's effective in SEO is actually making core product changes. So I think that that's actually a... [06:03] a perfect profile, even though it was somewhat unintentional. [06:06] I haven't heard of someone's company being bought by Whirlpool yet. You're the first. Can you talk a bit about Graphite, which is the company and the agency that you run now? [06:15] Sure. So when I left my last company, [06:19] I was doing consulting as a, as a solo practitioner and I, and I had been doing it on the side since 2014, starting with Dumbtack. But then I, [06:27] decided to do it full time. I actually didn't intend to do that because I always thought of [06:32] consulting and agencies as sort of distasteful and not something that I wanted to get into. And it was somewhat by accident where Masterclass reached out to me, Honey reached out to me, Ticketmaster reached out to me. So those are the first projects that I worked on.

[06:47] And I hadn't been planning to have that be my full time role. I was planning to sort of explore and maybe build a product. [06:54] Those projects went really well. And then I got more inbound interest from companies when Masterclass did well and Honey did well. So I decided to start an agency. [07:05] We build out SEO and content strategies. So I thought broadly about all the different things that I could do and where the biggest need was, where the most underserved market was. [07:16] And SEO is by far more underserved than SEO. [07:19] Anything else that I was good at. [07:21] And so, [07:22] Graphite focuses on building those FCO and content strategies. And a lot of it's inspired by the early work we did with Masterclass around editorial SEO. So... [07:30] That's graphite. And I think one unique thing about us is that I found that getting things done is actually more important sometimes in the strategy. So a lot of what we've tried to do differently is to. [07:42] is for us to build as much as we can on those projects. I think the distinguishing characteristic is that we can get things done [07:49] and that we can build stuff. So that's a bit of background. Awesome. Just off the bat, what are some of the biggest myths about SEO? [07:57] Even number one biggest myth are technical audits. So we have people reaching out saying we want a tech audit. [08:05] And, [08:06] Essentially what that is, is [08:08] Give me a list of bugs to fix. [08:10] which is different from help me grow. And so a tech audit is basically, it's this, [08:16] strange thing in SEO where you basically

[08:18] Get Screaming Frog, which is this tool. You buy it for, I think it's $150. You crawl the site. [08:24] And then you give a report and you say, here are bugs. [08:27] which is not that useful. What the company actually wants is they want to grow. [08:31] And so we'll try to reframe that. And number one, technical SEO, I think is misunderstood. You want a strategy around your technical components and you want a general SEO strategy to help you grow. [08:43] So I think that that's the biggest myth. And then I think... [08:47] Similar to that is that [08:48] people over index on technical seo and not enough on editorial seo some companies can benefit from programmatic seo which is related to second class seo technical seo is mostly [08:59] internal link architecture and a few other things. [09:02] and bugs but that's not how you grow the way that you grow is with programmatic and with editorial seo and i think editorial seo and by that i mean articles and and content and [09:11] manually written content is underappreciated. So I think those are the two biggest myths. [09:16] Okay, I have a billion questions I want to ask, and you're touching on a bunch of these things already. Before we get into that, I just want to kind of frame our conversation and start at the top a little bit and come back to some of these things because this is really important foundational stuff. There's kind of three things I want to focus on today. One is when to SEO as a company, two is how to SEO, and three is just to cover a few important concepts that folks need to understand if they want to be really good at SEO. Does that sound good as a kind of a base? [09:46] Cool. And a quick shout out to Brandon Lee, who's the co-founder of this company called Power, which I'm an investor in who is all about SEO and asked them what to ask you. And he gave me a bunch of great questions. So those are going to be inter spliced into our chat.

[09:58] Awesome. Plus one for Power. I'm also an investment. Okay. Awesome. Oh, man. Okay, great. Go Power. Okay. So to start, I want to help people understand when to focus on SEO, when it makes sense as a growth channel and timing around when to invest in SEO. So question for you, what are just attributes of a product or a company that tell you that SEO could be a growth driver and or a massive growth driver? Because I imagine SEO isn't useful for everybody. How do you think about that? [10:28] There's two big things. The first thing is, is the addressable market large? So is the addressable market for SEO large? [10:35] And for most categories, it is. The second is, do you have authority? Do you have existing traction? [10:41] If you start from zero and you have no traction, so we talk with Seed. [10:46] seed stage companies and series A companies [10:49] Typically, they don't have a lot of authority and it's too soon. [10:52] And Google doesn't want you to just be an SEO site and they want you to be a credible person. [10:57] Domain? [10:58] before they rank you. And if you're starting from zero, they don't have enough signals that you are. So, [11:04] The way that we assess that is... [11:06] I will go into the first signal I'll look at is what's your traffic? What's your current traffic? And your non-SEO traffic is actually an authority signal. [11:14] So I'll go into SimilarWeb, I'll put the domain in, I'll look at the total traffic, and I'll want to see at least... [11:20] I would say 1000 visits a day roughly for non-SEO at least. [11:25] you have five visits that that's very little and then the second thing i'll look at is the number of referring domains so i'll go into ahrefs or scm rush and look at the total number of referring domains i'll try to have at least 1 000 roughly referring domains you could grow

[11:40] with less than that and less than a thousand visits, but it's much harder. [11:44] and the more you have, the easier it is to grow, and the faster you can grow, the more you can grow. [11:49] And then in terms of the addressable market, that's a little bit [11:51] more complicated, but [11:53] Most markets are pretty large, actually, in SEO. [11:57] but we want it to be large. So WithPower is actually an interesting example. So they're a clinical trial lead gen site [12:05] They want to acquire people to take clinical trials. [12:08] How many people are typing online, I want to take a clinical trial? Not very many. [12:13] But the number of people that could be taking clinical trials is very large. [12:17] And so if we target the persona, like what's the key demographic of who might be [12:21] a candidate for taking clinical trial. It's very, very large. And so, [12:25] We can then create content that targets that like gig economy or college students or [12:30] People like that. [12:31] And so if you think about targeting the persona, most sites have a very large addressable market. But the way that I would think about the addressable market is that which is what's what product am I offering? What are the use cases for that product? What is the persona? [12:45] What's the size of that? [12:47] And then I would assess that typically by looking at [12:51] external benchmarks. So, [12:54] If I'm a shopping site, I can and I haven't started yet. I can look at other shopping sites. I can see how much traffic they have. So again, I can go into similar web. [13:01] If I'm wish.com and I want to see what my traffic potential is, I can put in Walmart and Wayfair. [13:08] put it in similar web, look at their total traffic. You can get similar webs free for all of this.

[13:12] And then I can get a sense of how big I can get. The one other thing I'll mention is that there are product competitors and audience competitors. So for something like I'm at a WeWork right now. So. [13:27] For WeWork, I could look at the traffic for other... [13:30] you know, rental office companies. [13:33] Or I could look at companies that are ranking for the kinds of things that I would want to rank for. And they may not be direct front of competitors. So like fintech is interesting. So for Robinhood. [13:42] Robinhood could look at other [13:44] sites that allow you to sell stocks in crypto, or they could look at Investopedia. Investopedia doesn't allow you to buy stocks, but they have a bunch of traffic. [13:51] It's not a product competitor, but it's an audience competitor. And so these competitors basically can tell you what the size of that market is. So that's how I think about it. [13:59] the addressable market so ideally it's large it typically is and then the more authority i have the more i can compete [14:05] If I'm starting from zero, it's probably too early. But once I have traction, SEO can then multiply that traction. [14:10] Okay, wow. There's so much there. I'm going to try to summarize because I think that was just like a million nuggets in one answer. So the way to think about this, if I'm just kind of reflecting back what you just said, if SEO is worth investing in, you also kind of answered when it makes sense, which is interesting. One is just, do you actually have, you have like a thousand or more visits a day just organically through search is a sign that maybe it's a good time before that. Is that?

[14:40] Okay, got it. And other than search meaning, like what are examples that when you say that? Usually direct and paid, but direct, paid, email, social, anything else. And why is it that it's not from search? [14:53] So Google wants to see that you're not just an SEO site. I see. And if you're not getting any traffic from anywhere else and it's only from SEO, then you're an SEO person. [15:03] Farm. [15:04] essentially got it so the more traffic you have that's not from seo [15:08] that means that you're credible the more referring domains you have that means you're more credible more authority the more shares you have more authority so the more of these signals the more you're a known brand in google's eyes and the more authority you have and the better you can rank so it's the non-seo traffic [15:22] piece that I especially want to look at. Got it. And then the other piece is that there's a thousand domains linking to you. It's something you'll look for, right? Yeah. And it depends. But yeah, those are my rough [15:34] back of the envelope benchmarks. And if you're not hitting these benchmarks, is the idea, consider investing in this so that you can eventually start doubling down on SEO, or is it just let it happen and kind of see if it even happens on its own? You could do either. Usually, people don't specifically focus on [15:51] links or non-SEO traffic for SEO. They just do it because they want to grow their company. So usually that's what happens and it happens for free. But you could be intentional about it if you want to. Awesome. Okay. So that touches on when to start focusing on SEO. And I'm going to ask you if there's anything else you should think about about the when. But it feels like the other piece is even more important is should you even consider SEO and the total adjustable market is a big part of that. And you've mentioned the site SimilarWeb. It's just similarweb.com, right?

[16:21] you how much organic traffic you're getting from search, how much your competitors are getting. Is that roughly a good way to describe that? Yeah, it has total traffic and then it has traffic by channel. [16:31] and it has countries and other things but i'll look at total traffic and then by channel and then total times percent non-seo equals [16:37] total total traffic from non-seo and you may have mentioned this but again what are heuristics looking at that data that's like okay this market is big enough for seo well it depends on you know how much traffic you want but um the heuristic i mean the heuristic is to look at other sites that are your product competitor your audience competitor and you can get their totals and then you can [16:58] say whether or not those totals are meaningful, but I would do totals times what I think my conversion rate is equals sales. [17:05] compare that with other channels so how much could i get paid how much could i get messy how much can i get from social [17:10] And then that's how you would decide which channel and whether or not SEO is the channel to focus on. Got it. So the time is like already is happening. They're going to your competitors. This is what you can eat away if you invest in SEO. Yes. Awesome. [17:22] Okay, before I get into other attributes of when it makes sense to invest, is there something about the actual product? [17:29] The way it works, say, generates content automatically like Lastdoor or Zillow or Zapier. Is there something else that tells you this is going to be really useful? This is going to be a good product to use for SEO or the opposite one's like probably not going to work out for SEO? Sort of. So when we're looking at all of our different channels, SEO, paid, social, other, we want to do impact versus scope.

[17:54] And we talked about impact scope is another one. [17:57] So if you have no content at all and no UGC, then the scope is higher to... [18:02] Compete. [18:03] in the addressable market. If you have a ton of UGC like Glassdoor, the scope is lower. So the more UGC you have, the more like for Zillow, the more addresses you have for a shopping site, the more SKUs, more products you have, the scope is lower. It's easier to compete for those. So definitely consider that. But if you have no UGC and no SKUs, you could still compete through editorial LCO and through other ways. [18:25] so it's not that that would tell you you cannot if you don't have those things but those things [18:30] make it easier. Therefore, the scope is lower. Therefore, impact versus scope ratio makes more sense. This is a good time, I think, to introduce these buckets of types of SEO. You mentioned programmatically editorial. Can you just talk about what these buckets are and what they what they mean? [18:44] There's programmatic SEO, there's editorial SEO, and then there's technical SEO. So programmatic SEO are pages that are automatically generated from a database of information. So for [18:55] eBay. [18:56] Product pages are generated automatically from information in their database or for Zillow. [19:01] home pages, address pages are generated based on information in their database. [19:05] For a site with UGC like Glassdoor, that is programmatic generated as well. So that's programmatic SEO. It's usually category pages and item pages. Item page meaning product page. [19:16] page about a home page about a video. So it's like an individual item and then there's category pages, which are grids or lists of items. So those are the two main page sites for programmatic SEO.

[19:27] Then there's editorial SEO. So editorial SEO is where a person actually sat down and wrote a piece of content, like a writer wrote something like an article or a guide or a listicle. [19:38] And, uh, [19:39] Editorial SEO is more and more actually driving more traffic than programmatic SEO. Programmatic SEO used to be almost all SEO products. [19:48] when I started and editorial SEO has overtaken that. Wow. So that's editorial SEO. And then there's technical SEO. So technical SEO is just the infrastructure. So like internal link architecture, tags, redirects, [20:00] page speed.com. [20:01] that applies for everything. [20:03] But those are the three main types. [20:05] Got it. Okay. So these are basically three areas you can invest in also as a company that's investing in SEO. Which of these is the best [20:15] most lucrative you may have mentioned editorial maybe is on the rise if you could choose because you can't really make ugc and so i guess if you could choose is there one that's the best route if you could choose it depends but the short answer is editorial seo is usually the best but it depends on your category [20:32] Given your addressable market, you want to see the page type. [20:35] that maps to the queries that you want to target. [20:38] And sometimes that's an article and sometimes it's not an article. So for Zillow, that would not be an article. It should mostly be programmatically generated pages. Or for an eBay, it's mostly going to be programmatically generated pages for most commerce companies. But for most other companies, it's editorial SEO. [20:54] So editorial SEO is always an opportunity for every single company if there's an addressable market.

[20:59] And programmatic SEO is sometimes an opportunity. [21:02] The nice thing about programmatic SEO is [21:04] is that you don't have to spend money for each page that you create. [21:08] So Zillow doesn't spend money for every single home page that they create. eBay doesn't spend money for every product page that exists. [21:15] Editorial SEO, every page has a cost. [21:18] So that's kind of what I think about. What are some examples of really good editorial SEO ish companies, sites? Yeah. What comes to mind? I'll give you a couple that I work with and then I'll give a couple that I don't work with. So Masterclass is one of the projects that I'm most proud of and they do a great job. [21:37] and build a lot of content about their classes and stuff that's related to their classes. [21:41] Another example is BetterUp, which is a coaching platform so people can get a virtual coach to improve in their career and build a lot of content around it. [21:49] how to give difficult feedback, how to create a five-year plan, things like that. [21:53] So those are two projects that we worked on that I'm really proud of. A couple other examples that I did not work on. So NerdWallet is one of the top editorial SEO companies, and they have guides around credit cards and financial information, and their stuff is top-notch. HubSpot does a really good job, so they have a lot of. [22:11] content around productivity, [22:13] And I've done really well. The other is DotDash. DotDash is a conglomerate of many different publishers like Allrecipes and Investopedia, and they consistently over and over again publish. [22:24] perform really well they're probably the most successful editorial seo [22:28] company. So those are a few examples. Awesome. I want to transition to how to be good at SEO and how these operations run and how to do this. Well, before I do that, coming back to a question I wanted to finish asking is when does it make sense to invest in SEO? Is there anything else to think about for a startup that's like, not yet, not yet, that tells you that's time other than these benchmarks you shared of a thousand non-search visits a day and a thousand links?

[22:53] Yes, well, we want to decide of everything that the company could work on, what is the relative priority of SEO versus marketing? [23:00] everything else paid social and things like that so we want to compare it with these other things and we want to look at the cost [23:06] And so depending on the cost and the opportunity and how that compares with other channels, [23:10] That's how I would. [23:11] how I would make the decision about when and whether to do SEO. Got it. So yeah, very specific to what other opportunities they have. Like if paid is working really well, maybe not worth investing in SEO just yet. That makes sense. Yes. Or Airbnb. Brand is working really well. Let's dive into brand. Yeah. Okay. So shifting a bit to how to do this well, just broadly, what are the buckets of [23:41] as a company? What are the investment areas that they have to do well at? So it depends. And here's the workflow that I would go through. So first, we want to define the addressable market. [23:51] and [23:51] Given the addressable market and what are the different topics and keywords that I want to acquire on? [23:57] And then what is the page type that I need for that? So for Zillow, the addressable market is mostly address pages. For Masterclass, it's mostly articles. [24:06] And so we look at benchmarks, we see what keywords they're ranking on. [24:10] Then we take those keywords. Then we figure out the page type. [24:14] Then given the page type, we want to figure out the product requirements. So what needs to be on the address page on Zillow? Then we want to figure out the content strategy. What are the different subtopics and themes of content that needs to be on each page?

[24:27] And that's specific to the different page types. [24:29] And then last is the infrastructure. So how do we help Google find all of these different pages via links? [24:36] And so that's the workflow. [24:38] that i go through it depends on your addressable market what [24:42] what that strategy looks like, but those are the components for a high-impact SEO strategy. To unpack a couple of these, so there's a page type. You said that if it's an article or an address for Zillow, how do you do that? How do you figure out what kind of pages that you should create? So the short answer is you take a few keywords you want to rank for, you put them in Google, and you look for themes of the page types that appear. So for Masterclass, you type in how to [25:10] And you see that all the results are articles with a recipe. [25:15] That's a very clear correlation. [25:17] that if I want to rank, I need a page like this. So a masterclass is an interesting example. [25:22] They have classes and chapters and each chapter is like they have how to put in a chapter for for Gordon Ramsay. [25:30] And what we could do is we could have a chapter page and a chapter page is a transcript of [25:35] of what Gordon Ramsay said. [25:37] And that actually won't rank. The page type is wrong. Google is not ranking chapter pages or transcribed text. [25:44] the ranking articles. [25:45] And so that's the short answer. The longer answer is if you want to rank for how to poach an egg, you would want to know, well, there's a thousand different keywords that this page is actually going to be targeting. [25:55] So type in all thousand of those different keywords and then look for the patterns of the

[25:59] page types for all thousand keywords. Now that's impossible to do manually. [26:03] but [26:04] If you were to be very rigorous, you would essentially do that. But you're basically looking for what are correlations of URLs or page types that are [26:11] that are ranking for these terms. Usually there's one or two page types. So sometimes there's two different ones and it's like a... [26:19] A fractured intent? [26:21] And so an example would be best cameras. So you'll have a category page and a listicle page typically. [26:28] So you could have... [26:29] one or the other. [26:30] but you will need one of those if you have a product page you're not going to rank so you're basically looking for correlations of which page types are ranking for the terms that you want to rank for and that's how you come up with your page type and similarly you're going to have [26:41] if you want to know what should go on the page you again look for what are the correlations of the components [26:47] of the pages that are ranking. [26:48] that are ranking for the terms that I want to rank for. For example, do I need a map or do I need an address or do I need a phone number? A way to answer that is to see what are the patterns of the pages that are ranking for the terms you want to rank for. So it sounds like essentially you're this detective trying to understand and reverse engineer what Google wants out of this general topic. Is that the way to think about it? [27:08] Yes. And the only thing that I would say is, [27:11] I wouldn't have the goal be copy everyone else and just do what we think the algorithm wants. I would take inspiration. [27:19] and make a great page. So, [27:22] If it looks like maps are useful and phone numbers are useful, you should include that. But I wouldn't just [27:28] optimized purely to get a visit and then stop. I would take as inspiration. Users in general seem to want to be able to contact the business.

[27:35] And I want to be able to [27:36] find the business, which is why there's a map. So what are some other ways that I could help users do that? That's how I would... [27:43] how I would come up with requirements. Got it. So look at what Google is finding valuable. See what makes sense, what doesn't. Think about what else would make it even more valuable. Like the way to think about it isn't, [27:55] what's Google's secret crawling algorithm? It's like, what is going to be useful to somebody that is looking for this thing? And Google's probably constantly trying to make it do that, even though it's not perfect. And I imagine the more you're connected to, what is this? How is it going to be useful to a human? The less likely an algorithm change is going to screw you in the future. [28:14] That and the better your engagement. [28:16] So if it looks like users really want to be able to find a business and you do that better than anyone else, your engagement will be higher and your entire engagement will cause you to rank higher and it'll cause you to rank higher right away. [28:26] So it's not only a way to be robust against algorithm changes. It's a way to rank higher right now. [28:33] And engagement is click through on the page that Google sends you or something else? It's click through. So it's. [28:40] If I typed in best camera, [28:42] What's the rate of clicking on my position versus somebody else? [28:47] So that's signal one. Related to that is, did you go back? [28:51] So you clicked on my page, then you went back and you clicked on somebody else. That's a very bad signal. [28:56] So that's engagement on google.com. [28:58] The second part of engagement, which is probably even more important, is did you stay on the page? [29:03] And did you stay on the page for a long period of time?

[29:06] So Google's tracking everyone through Chrome and Android. [29:10] and they can actually see what happened after they left Google. [29:12] And ultimately, Google wants you to end your session or find the answer, find what you wanted. [29:17] on that page so the more you click on on our url and then go to the page and then stay there and spend time and get your answer that's a key ranking signal [29:28] Wow, I didn't know any of that. That's awesome. What does the operation look like to generate these pages once you've kind of figured out I need, say, address pages with a map and a category? What does it look like to actually generate? And do you need like millions of pages, thousands of pages? What's like the order magnitude? And then how do you actually go about generating these? It depends, but it's usually not millions unless you're eBay or Zillow, but usually it's less than a million. It depends on the page type. So for a product page, [29:57] Typically, you would index all of your product pages and however many product pages is the number of URLs. [30:02] But you wouldn't want to index things that are thin or empty or we have a product and it's out of stock and there's no description and there's no reviews. You might not want to rank all of those, but usually for product pages or item pages. [30:13] you would index all of them that are of some threshold of quality. [30:18] For category pages, you want to... [30:21] find all of the categories that go after keywords where the category page is the right page site. So what are all the queries where category page is ranking? [30:30] And then you don't want a page for every singles keyword. [30:33] So when I started in SEO, [30:35] you would have a category page for every single keyword. And so...

[30:39] I was at a shopping site and we had [redacted address] that you could possibly search for a product, [30:45] had its own page and that worked really well. [30:48] The pages were of dubious utility because of that, but that worked well. That no longer works well. [30:55] And Google does not want you to do that. So you'll see in there, [30:58] Guidance. [30:59] that um [31:00] You should not have many minor variations of every single way that you could search. You should have themes or topics. [31:06] And so a topic for best cameras, best cameras is going to rank for. [31:10] the best cameras, 10 best cameras, cameras, best, best digital cameras, all these things are going to map to a single camera category page. And then there's a bunch of other category pages. So you basically take, [31:21] the 20 million commerce keywords and then you group the keywords at the topic level and then you launch just those [31:28] So that's how you do a category page. And then for an article or listicle, you would basically build out an article or listicle page type, and then you would have writers... [31:36] write articles and then publish them through Contentful or Webflow or some sort of CMS. [31:40] Oh, cool. I want to talk about that, just like the actual platform. But this might be a good time to introduce this idea of topics you mentioned a few times. I was going to ask you about that towards the end. But just this idea of, I think, topics versus keywords is, I think, the way you recommend people think about what to focus on. Can you talk about that? [31:57] When I started an SEO, again, [31:59] It was keyword based. Every keyword had one page. [32:03] and now many keywords have one page. [32:06] So if you're a commerce site and you have 10 million pages, you're going to have a

[32:10] It has to be automatically generated. There's no way for you to afford to write [32:14] 10 million pages. [32:16] So you have to automatically generate it. [32:17] And that led to a bunch of low quality content. So, [32:21] Since 2008 with Google Pen, there's been a bunch of other algorithm changes. [32:25] Google wants to again. [32:27] find a page that covers many different keywords and the algorithm has improved. So for best cameras, they know that [32:34] Best digital cameras is the same as best cameras is the same as something else, some other variation of that. And so the algorithm has improved and it has a semantic understanding. [32:44] So the algorithm is targeting topics, whether that's our intention or not, you are targeting topics. [32:49] And so there's typically about [32:52] Any given page is typically going to rank or be able to rank for about 200 to 2000 different keywords. [32:59] And um... [33:01] This is important for a few different reasons. So the first reason is that [33:04] the search volume for what we're targeting is not the search volume for one keyword so for best cameras [33:10] The search volume is not just the search volume for best cameras. The search volume for 2,000 different variations of best cameras. [33:17] And so if you think the [33:19] The search volume for best cameras is a thousand. [33:22] The topic of best cameras is probably 100,000. [33:25] It's probably 10x at least the search volume of the keyword. [33:30] And so when we're making prioritization decisions or, [33:32] Should we go after this? [33:34] We want another right search volume. If we have the wrong search volume, then [33:38] then we're going to make bad prioritization decisions on a very large scale.

[33:42] quarter magnitude. [33:44] The second is if our topic is going to rank for a thousand keywords, we don't want to write a thousand articles. And if we did, we wasted a bunch of money. So, you know, for best cameras, if I had. [33:53] Best cameras, best digital cameras, best rated digital cameras. [33:57] We wasted money if we wrote an article or made a page. [34:00] for all of us. [34:02] and the order of magnitude can be substantial of wasted money. [34:05] And then the last is that the page that is actually ranking for for that. [34:10] needs to fulfill the intent of all thousand keywords. [34:13] And if there's a theme that you're not talking about and you didn't know about it, [34:17] then you're not fulfilling the intent and you're not going to rank as well. [34:20] So for all of these reasons, it's much, much better to think in terms of topics rather than keywords. What's an example of a topic, maybe from an example of a company you worked at versus a keyword? What's like the how high level do these topics get? Sure, I'll give two examples. So an example with Masterclass is Butter Lettuce. [34:38] And the keyword, I think it ranks for about 400 different keywords. [34:43] And there are themes like health benefits and other kinds of lettuce that [34:48] and recipes. So if we only define what it was, what butter lettuce is, and we did not talk about health benefits, [34:56] then we have this gap. [34:57] and we're not going to perform as well. And part of my masterclass, [35:00] Early on was able to outrank Food Network, who had way more authority in food. [35:06] and food. [35:06] is because other companies have these gaps. They were not looking at all [35:11] 400 keywords and they didn't talk about health benefits and therefore they didn't rank as well because a google's content score said

[35:17] you're not comprehensive and being users who were searching for butter lettuce health benefits. [35:21] didn't get the answer, and then left and went somewhere else. [35:24] One other example that I'll give. [35:27] is with BetterUp. We have this article five-year plan. And if you look at that article, I think it ranks for about a thousand different keywords. [35:34] And there are themes like examples and templates, personal versus business five-year plan and five-year versus 10-year versus one-year plan. [35:43] And again, [35:45] Because we have all these subtopic themes, our content is comprehensive, content scores higher. [35:51] and users who are searching for five year plan template get a template and now we have a better engagement score. [35:58] So those are two examples. And one thing that I'll say is the way that we can know how we should cluster keywords, like should we have one page for template versus examples? Should we have two separate pages? [36:09] You can basically put in five-year plan template, five-year plan examples, and you can look to see how much overlap. [36:15] of the results are there. So if the results are completely different for those two keywords, you should probably have two different pages. But if they're all the same, Google's saying, [36:23] these keywords are essentially a single concept or a topic so that's how you can [36:27] think about you know which keywords should be clustered under one topic or another [36:32] That makes sense. I was going to ask you with the butter lettuce example, is that how you, how did you figure out that there was a big opportunity there? What was like the [36:40] The light bulb. So what we do is we'll look for competitors. So we'll look at a food network, all recipes, other sites, and then we'll basically get all of their pages and all of their keywords using a tool like Ahrefs or SEMrush.

[36:54] So if we're starting from zero and we have no information about food, we just go to other sites who have. [36:59] who have already done this. [37:01] and get all their information and data. [37:03] and figure out all the different pages that they have, all the different keywords that cluster to each page. [37:09] the search volume or the traffic for each of these, and then we can prioritize them based on search volumes. [37:15] So we have... [37:17] 10,000 food topics, sorted by search volume. [37:20] or traffic volume. Then we want to look at how competitive we can be, which is topical authority. Then we want to look at conversion intent. How does food versus comics versus photography convert? And then we have a prioritized list of [37:35] topics by expected conversions. So that's ultimately how we think about prioritization. Awesome. I'm just going to keep bouncing around based on where this is going. What are tools that you recommend and use most to help you in this? You mentioned a few. Screaming frog sounds awesome. What else? What else do you recommend? [37:54] Four tools. So tool one is Google Search Console. So Google Search Console is free. Anyone with a website can get an account. There's a bunch of really useful information you can get. [38:05] real traffic information you can do surf tracking and you can do surf tracking for free so every single keyword you rank for you'll have that in there so it's very useful that's my number one tool [38:15] My second tool is ClearScope. [38:17] So there's a bunch of different tools for content analysis, content scoring. [38:22] ClearScope is the best one.

[38:24] And ClearScope will tell you what I talked about subtopics and comprehensive content. So that tool helps you. [38:33] understands whether or not you have a gap in your content. [38:36] That's the second tool. [38:38] The third tool would be a keyword research tool. There's many of them. [38:43] scm rush and ahrefs i use the most ahrefs is my favorite but sm rush is good there's a bunch of other ones [38:48] They're all pretty similar. So pick one of those. [38:51] that's like 100 to 200 a month i think google search console is free clear scope is based on volume and then the last [38:57] tool or tools is that we have a bunch of internal tools uh graphite that that are not accessible but [39:04] We used to power... [39:06] some of the stuff that I talked about where there's not already a tool. So, [39:09] topics like i mentioned we have internal tools for that we have internal tools for things like what subtopics you need topical authority there's aren't tools for that so we basically built some in-house [39:19] We use that for when we're building strategies. [39:21] What about Screaming Frog? You mentioned that. Screaming Frog is good. It's a bit [39:26] niche. You don't have to have it. The application is Screaming Frog. [39:31] is one crawling. So I want to get all the URLs from some other site. [39:35] So that's a good use case. The second is you can audit stuff on your site so you can crawl your whole site and find error codes and redirect loops and things like that. [39:45] And then another key use case for Screaming Frog is your internal link coverage. So, [39:50] Do your pages on your site, do they have links pointing to them within your site or do they not?

[39:54] So that's what Screaming Frog does. [39:56] Screening Frog will be number five. [39:58] I would just say you can build an SEO strategy without Screaming Frog, but it's certainly useful. It's got a place in my heart now, Screaming Frog. I use it every day. Okay, okay. So if you're a real pro, Screaming Frog, if you're not using it, you're not real pro. That's my conclusion. One of the classic issues with SEO that people often worry about is how long it takes to see anything, to see if it's working, to see results. So it's always this like, "Oh, we'll get to it because it's going to take us six months or a year to even see any impact." [40:27] Is that true? Does it sometimes... [40:29] show you results sooner. How do you think about timelines on SEO? It depends, but usually I'll say six to nine months. [40:36] but it's a function of [40:38] how much authority you have and how much existing SEO traction you have. The more authority you have and the more traction, the faster it is. So, [40:45] If a company like Netflix makes a change, you'll see it the next day. [40:50] And if a brand new company... [40:52] Strikes from scratch, it'll take. [40:54] Probably a year. So you start on SEO today, takes maybe three months to build the page types, do all the analysis, come up with a strategy, build everything, launch it. [41:02] write 20 articles. So, [41:04] three months to actually [41:06] launch stuff and then three to six months to see it. [41:10] start to rank and to grow. [41:12] So what's happening, uh, [41:14] The earlier we are in less authority we have, Google is testing our pages. [41:18] on keywords that are less popular to see if our engagement is good [41:23] And then with better engagement, we'll rank more and more and it'll grow exponentially. And so, again, back to the previous conversation, Google wants to see, well, when we rank for best cameras, what's the click through rate on Google? And did you stay on our page?

[41:35] or did you go somewhere else the higher that score is the better engagement is the more we can now rank for more competitive uh [41:42] keywords within that topic, then we're ranking, you know, we have 20 articles and we're ranking for those things. Now we can rank for other stuff that's related to those 20. [41:50] And so it sort of compounds and grows that way. So the rough timeline is around... [41:55] three to six months post of launch. Having said that, the larger the site is and the more authority it is, it could take [42:02] days so for you know for for a large site like a ticket master and ebay it's probably one to four weeks is the rough timeline [42:09] once you're really optimized. [42:10] So as a founder, you're probably listening to this and be like, "Damn, that's such a long-term investment. That's not going to show impact. And I have all these other things I got to work on. Should I be doing SEO?" Is there things that along the way, say before six months that are pointing to you're heading in the right direction, keep going, or just like, "Okay, this isn't going to work." [42:28] just stop now. Is there any kind of early signals, leading indicators? Yes. I wouldn't say that there are signals that you should, well, there may be signals that you should stop now, but we want to look for leading indicators. [42:39] And a leading indicator is, are we ranking at least number 20 for anything? So usually what we'll see is we'll launch something and maybe in the first week we're ranking number 15 for something and we get one click or no clicks. But we go from number 15 to number 12 to number eight. Now we're at position eight. We have some impressions for this keyword. We don't have that many clicks because most people don't click on number eight. [43:02] But we're starting to see signs of progress. Then we go from number eight to number five. Now we get more clicks.

[43:07] Now we know if we can rank number five for these interesting head terms. [43:11] We can rank. So let's say we have 10 articles. [43:14] We're ranking number five for keywords for these 10 articles. We can rank number five for a thousand articles. And so you can see these leading indicators and signs of opportunity early on, even with a small number of pages. And I would wait until I see that to then scale it. It's common to have to incorrectly conclude failure. I listened to the Yuri Grammarly episode and he talked about false negatives. Like we tried this thing. It didn't work. Therefore, it won't work. [43:43] It's common for the first try or for several tries in SEO to just not be implemented correctly. [43:49] And so... [43:50] If you're not ranking for anything at all, [43:53] either the strategy is wrong or there's no opportunity. It's probably that the strategy is wrong or that you don't have enough authority. [43:59] But that's how I think about it. [44:29] at competitive rates. Startups choose Lemon.io because they offer only hand-picked developers with three or more years of experience in strong, proven portfolios. Only 1% of candidates who apply get in, so you can be sure that they offer you only high-quality talent. And if something ever goes wrong, Lemon.io offers you a swift replacement so that you're kind of hiring with a warranty. To learn more, just go to Lemon.io slash Lenny and find your perfect developer or tech team

[44:59] And if you start the process now, you can claim a special discount exclusively for Lenny's podcast listeners. 15% off your first four weeks of working with your new software developer. Grow faster with an extra pair of hands. Visit lemon.io slash Lenny. [45:15] Amen. [45:16] When you're starting to invest in SEO, say a founder is listening to this and they're like, "Yes, SEO, let's do it." How do you recommend they hire for this person/have somebody take it on? Do you recommend finding someone that's like an SEO god that has done this a lot and bringing them in-house or finding someone young and hungry that's going to learn it all or bringing an agency [45:38] As an early stage startup, what do you recommend people do skill-wise and people-wise to start on this? [45:43] it depends on the strategy and it depends on the stage of the company and the opportunity and budget so [45:49] In terms of the strategy, [45:51] programmatic SEO is more specialized. It's harder to find people who are great at programmatic SEO. [45:56] Editorial SEO isn't easy, but there's a much larger set of people who can do editorial SEO than to do programmatic SEO. Programmatic SEO just has many different nuances around indexation logic. I talked about the category pages and clustering keywords at the category level. That's a lot of work. It's hard to do. It's specialized. [46:17] how to deal with UGC. So if we're a UGC site like Pinterest, [46:22] It's very complicated and you want to understand which pages you want to index and you need to build scalable systems. So that's a specialized skill. And there aren't that many people.

[46:33] If you can hire somebody who's great at that, [46:35] I, you know, it's, [46:37] I'm biased towards in-house, even though I run a consulting agency, I'm still biased in-house. So if you can get that person, that's great. [46:44] There just aren't that many of them. [46:45] But on that editorial SEO side, yeah, getting someone young and hungry could absolutely work. [46:52] On the opportunity and budget side, so if you're Zillow and you have SEOs worth a billion dollars to you, [46:59] just spend the money i think people under resource seo a lot of times and over resource ads so if you're zillow you're going to spend [47:06] tens of millions of dollars on ads or if you're ebay you're gonna spend tens of millions of dollars on ads [47:11] Why would you not have a really great SEO team? The amount of traffic you get is probably equal to that. So if you're going to spend $100 billion on ads, why would you spend... [47:21] $50,000 on SEO. That doesn't make sense. [47:24] So I think it's based on the opportunity and how complex the strategy is that you want to focus on. Awesome. That is really interesting. The two routes that you talked about, programmatic versus editorial, what kind of person would you look for for the programmatic? Is it like a growth PM? Is it an engineer? Something else? [47:43] It would be a growth person with technical ability. And I would look for somebody who's done it before. I would not look for someone who, [47:52] who has done it before at a well-known brand name company, necessarily. What I would do if I were looking for someone is I would try to find sites that have had really good SEO growths.

[48:04] Thank you. [48:05] where the percentage of SEO is really high, almost too high, like 80% of the traffic is SEO. [48:11] company I haven't really heard of. [48:13] because that's harder to do. It's easier to do programmatic SEO at a well-known brand name company. [48:19] It's harder to do at a company that... [48:22] that isn't well-known, [48:23] Yes. [48:24] And, you know, [48:25] The people who are brand name companies, everyone's competing to hire them. And the people at companies that are not brand name, [48:31] are less competitive. So, [48:34] They've had a harder route to go through. [48:37] And, you know, there are fewer people recruiting them. So if I were to look for a programmatic SEO person, [48:44] I would basically scrape the web, find all the sites that have grown through SEO. Percentage is really high. They're outperforming. They don't have that much authority. So they've made a ton out of the small opportunity that they had. That's how I would find a programmatic SEO person. It reminds me of GoCool. It gave this great advice that when you're looking for a specific function, instead of trying to just scrape for everyone that does this skill, look for the company that is really good at that specific thing. And then just try to hire those people. [49:14] place to hire SEO people. Yes. What is it that this person does day to day? So with programmatic SEO, it's not like sitting there and writing. It's basically building code and architecting taxonomies and updating sites, basically working with engineers to just create thousands of pages, millions of pages automatically, right? Yeah. [49:33] Yes, and the SEO person doesn't need to write code necessarily. I'll tell you what I did when I was in-house doing programmatic SEO. So most of my time was spent, A, looking at other companies that are performing well and trying to understand SEO.

[49:48] what's working. So that involves finding the other sites, [49:52] basically auditing what they're doing, going into Ahrefs or SCMRush. [49:57] getting all their URLs and keywords, which pages are performing the best, what are the patterns of the pages that are performing the best, [50:04] you know it seems like they tend to have these particular characteristics [50:08] So a lot of my time is spent just looking at other people's sites. [50:11] even more so than my own site. [50:13] Try to figure out what's working and reverse engineer what's working. [50:17] Then given that, [50:19] I'm trying to recreate that. So, [50:22] for a food network how do i recreate a recipe page like what are all those different components i'm coming up with product requirements [50:29] and then working with engineers to build them. I'm spending probably a lot of time on programmatic, [50:34] on the category page side going through lots of keywords and then i'm probably usually manually clustering the keywords and coming up with those browse categories that i mentioned so a lot of my time is just figuring out which pages are actually being created and coming up with that list [50:46] or working with engineers. [50:49] to do that. And then we're launching them. And then I'm spending a bunch of time, A, doing data analysis to see what's working and what's not working and why. And then last is tests, but coming up with ideas of what to test. [51:02] What I'll say is that [51:03] I spend more time and get more value. [51:06] On. [51:07] assessing other people's sites and assessing our own site than I do on testing. [51:12] Testing validates. [51:14] a hypothesis but the hypotheses come from [51:17] analyzing other people's sites and analyzing my own site.

[51:21] So that's where the valuable qualified hypotheses of what is probably working comes from. It's not from brainstorming. [51:28] The test validates the hypothesis. [51:30] So that's what I'm spending most of my time on. Is there an example that comes to mind of what you just described where maybe you test it too much instead of just looking at what other sites are doing or learn something by just scanning sites recently? [51:44] Well, I try not to do that, but sometimes companies test too quickly. So let's say that you're starting from zero. [51:52] If you're starting from zero and nothing exists, [51:55] then. [51:56] You shouldn't be testing things. You should be launching things. You should be testing things once things have launched. [52:01] So, [52:02] So testing is something I would do at like month six or later. [52:06] The beginning should be around building stuff. And when you say testing, basically writing A-B tests, right? It could be an A-B test or it could be a sequential test. And it depends on the tests, but you could make a site-wide change and then see the entire page type. [52:20] you know, went up or down and a sequential test, meaning, [52:23] change the entire page type. [52:25] at the same time that's still data. [52:27] It's not as clean. You're not controlling for every variable, but it's still useful. [52:31] But yeah, you can do a split test. [52:33] The one thing that I'll say is some things are hard to test. [52:36] For example, [52:38] There are domain-wide... [52:39] things like Google Panda or Google has a bunch of domain-wide [52:43] scores, and you can't split test that. [52:46] You could only split test that if you had 100 sites and you change 50 of them and you don't change those. But you can't split test everything. There are also things that are hard to decouple. So internal link architecture, a lot of times it's a very interconnected graph and to truly separate things.

[53:01] parts of the site can be hard. [53:04] So usually split testing. [53:06] is the best route, but [53:08] You could also do a sequential test, and some things just can't be tested. [53:11] Are there any tools that you find useful for testing SEO or is it just the same old A/B test? We usually build them in-house. You could use tools. What I'll say is the main effort is bucketing. So what's in the control versus the test bucket. [53:27] Thank you. [53:28] Or you have five different buckets. [53:30] But bucketing actually is not hard, especially if you have thousands of pages. [53:34] So I don't think the tools do it. [53:36] an especially great job at bucketing beyond what a human could do. [53:40] So for Zillow, just [53:42] take a random sample of whatever percent you want. And that's a perfectly fine bug, but you don't need a tool for that. [53:46] You might want to like compare the traffic composition of both buckets to make sure that they're roughly equivalent. But that's basically bucketing. And then in terms of the implementation of bucketing, it's actually not that hard either to build in-house. There are testing tools. They're perfectly fine. I just typically will build it in-house. [54:03] or rather do it in-house. Okay. Coming back to the two types of people that you want to look for, we talked about the programmatic person. On the editorial side, what I think of there is I saw this story that Datadog hired an engineer and their job was to write blog posts and that was really effective for them early on to build their SEO cred. Is that the right approach? How do you think about hiring someone to run your editorial SEO strategy? That's a part of the right approach.

[54:33] writing really useful content and information, which is [54:36] Probably the most important part. [54:38] The other parts are... [54:40] What should I be writing about? What are the topics that I should actually be writing about? A lot of topics maybe nobody's searching for. Maybe they're interesting, but are not searched for. [54:49] So selecting which topics [54:52] we want to go after is important. And it goes back to what we were talking about earlier. What are the most searched for topics? So what's the search volume by topic? [55:00] What is our topical authority? Where can we actually compete within that set of topics? Where is their conversion intent? [55:07] All of that multiplied out equals expected conversions. [55:10] And then we prioritize that way. Then we want to. So then we have our list of prioritized topics. Then we want to make sure that the article is comprehensive and we cover all of the subtopics. [55:20] That is not intuitive typically to a writer. [55:23] So the engineer of Datadog that you mentioned, [55:26] probably just had a really good intuition about [55:28] What should be covered? [55:29] But a lot of people don't. So a lot of times you'll have a gap. And so again, you won't, you won't cover health benefits of butter lettuce on that article and you have a gap. So having the outline and the subtopics is really important. [55:40] and a writer's not going to, [55:42] A writer will guess that and they may be right, but they usually are not. So you start with your prioritized set of topics. Then you have your outlines, your structure of your article, and then you give that to a writer. [55:53] So that's an SEO person. [55:54] Then you give it to a writer who's ideally a domain expert, has expertise. Then they write it. Then it gets published. [56:01] And then you analyze it and then you feed that back into the strategy. So SEO manager does topics and outline. Writer does writing. SEO manager manages publishing or maybe content manager.

[56:12] manages publishing and then SEO manager looks at results and then that's kind of the workflow. [56:16] Amazing. This SEO person, so it's still a dedicated SEO person who's running an editorial SEO strategy and they have people that are doing the writing. They're not usually doing the writing. [56:28] And to your point, maybe the most important part is figure out what to write about and what it needs to touch on. Yes. And the only thing that I would say is that for editorial SEO, [56:35] I've seen multiple examples where the person doing that SEO manager stuff has no experience in SEO prior and will [56:42] show them some of our workflows and they'll pick it up [56:45] Like at Masterclass, that's what happened and the team didn't have prior SEO knowledge and [56:50] We shared our workflow and then they became great. And they were actually great at content more so editorial and editorial strategy. And then they learned SEO and then they're able to do that really well. So it doesn't have to be someone doesn't have to be a, you know, an SEO expert with deep expertise. Programmatic is where you'd want a lot of expertise over years. [57:09] an SEO, but on the editorial side, you can actually pick it up if you have the right workflow and the right tools. [57:14] What does that operation look like initially and at scale? How many articles do you have to pump out a day, a week initially and then later on? How do you structure this and what does that look like operationally? [57:26] We'll usually do 10 to 25 articles a month. [57:29] But it depends. Essentially, you have your addressable market. How many topics exist in your addressable market? [57:35] And how long do you want to take to go after them? And how many of them do you want to go after? If you have 10,000 topics, like Masterclass probably has 20,000 topics, let's say. So if you're doing 10 a month, I don't know how long that'll take. It'll take 200 years or something.

[57:47] do you want to wait that long or do you want not want to go after all 10,000? You want to only want to go after 5,000. You have diminishing return over time. [57:54] But how many topics do you ultimately want to compete for and how fast do you want to take to get there? What I would usually do is I wouldn't write a thousand articles a month. [58:03] month one i would start small i would write [58:06] tend to [58:07] 30, assuming that I'm starting from zero. [58:10] 10 to 30 maybe. We'll usually do 25. [58:14] But, you know, do some volume around there. [58:17] And try to get those leading indicators that I mentioned before you... [58:21] go really deep on on seo once you have those leading indicators then you say okay over three months we wrote [58:28] 50 articles. Here's what the performance looks like. Looks like they're ranking number five. [58:32] You know, half the articles are performing, they're ranking number five for these different keywords. Okay, now we know that we can rank for a bunch of other stuff. [58:39] And then we just it's our it's an ROI question of what's the cost of writing each article? What's the return and how much do we want to spend? How fast versus ads versus other channels? So that's how I think about it. What I'll say is that other companies, once they get those leading indicators, they're [58:53] will write [58:54] 100 plus articles a month because, again, they don't want to wait 100 years to go after all the different topics. What does the MVP of that initial team look like? Is it an SEO person and a writer person? Is it just one person can do it all initially? What do you suggest there? [59:08] Probably someone doing strategy and someone writing. You could do both, but usually... [59:13] The SEO person does not want to write because... [59:16] People who love writing is a special group of people.

[59:20] And also they're, you know, domain experts as well. The SEO manager is probably not a domain expert in SEO. [59:26] what the article is about. So I would say [59:28] one seo person and one writer is the smallest groups that you could have that that would be able to do this effectively you could obviously have more when we do projects we usually have more [59:37] three to five people working simultaneously, but that's because we want to go faster and we don't want to wait too long. So if you want to be more cautious, minimum would be one SEO person and one writer. Awesome. Okay. There's a couple of things I wanted to come back to that I skipped. One is the page types. You mentioned there's types of pages you can write, category pages, address pages. Are there a set number of types of pages for people to put into their head? [1:00:01] There's a few pages that cover probably 90% of all use cases. So on the programmatic side, there's category pages, which is a greater list of items, like 20 products. [1:00:13] There's an item page, which is a single item, which could be a product and address. [1:00:17] a video so those are the key page types of programmatic underneath category there's [1:00:23] Brand, subcategory, a bunch of different ways you could slice categories, but it's essentially a grid of items. So that's a category page. [1:00:30] and a product page. Those are the two main page types. One other one I'll mention is like a question and answer on Quora or a forum, so UGC pages. [1:00:38] Thank you. [1:00:39] So those are the primary page types on programmatic. And then on the editorial side, it's essentially a page with content. And that can be an article, a listicle. A listicle is like a list of items packaged as an article. You could have a guide. You could have a product marketing page, like a landing page for a feature that you have. So those are the...

[1:00:57] I would actually break out product marketing pages and homepages as sort of a separate miscellaneous bucket, but like product marketing page, homepage. [1:01:06] Support. [1:01:07] Those are the main page types. There are a few, those are the main page types of like eight and 90%. Then there are special cases. So special case would be, [1:01:15] tool pages like a mortgage calculator or a body mass index. [1:01:19] calculator or Zapier has integration pages, Google Sheets plus Gmail. So that's a special page type for them. So those are the main page types. And coming back to how you decide which of these page types to do, you basically look at what is Google finding most valuable within the space that you're coming after, right? Yes. What's your whole address on market? [1:01:42] by page type which ones have the most opportunity which is essentially impact [1:01:46] what's the scope of all these different page types, and then you prioritize based on impact and scope. [1:01:51] Awesome. The butter lettuce example, say you want to win butter lettuce as a topic. Do you have one page that becomes like your butter lettuce page that you're driving all your traffic to? Or do you try to have like 100 variations of it to win the broader butter lettuce topic? [1:02:21] lap of the keywords that I have? Or is there is a completely different set of [1:02:25] of results. And if it's completely different, then I would separate them out.

[1:02:29] The other thing that I would do potentially is if let's say the butter lettuce page is ranking number six for health benefits and we talked about it. [1:02:38] But maybe if we and that's a huge search line, if we really talked about it, we might be able to compete even better with this spin out sub article. [1:02:46] And you can do that by, again, looking at the keywords that you're ranking for. And if there's some theme where you did cover it, but you're not ranking quite as well as you want to. [1:02:55] break out that section of health benefit and then have a separate article for that. [1:02:59] Awesome. Man, there's so much meat to this. My brain is full, but I have many more things I want to ask. So let me ask a few more things. One is just like a meta question. It feels like there's a lot of conflicting advice about SEO. Everyone's always kind of giving you different opinions about the way to win SEO and things are always changing. Do you have any advice for folks that are trying to learn how to get better at this? They'll listen to this. They'll learn a ton. [1:03:29] what to believe, what's real, who to listen to. [1:03:33] So I think it's based on doing data analysis of your site, of competing sites to generate hypotheses of what you think is working and then testing them and validating those hypotheses. [1:03:45] which is different from Google said something and therefore I will believe it completely or somebody told me something. [1:03:52] So, [1:03:54] Google is a source of hypotheses to test. [1:03:56] analyzing other sites, something that someone said, something that another SEO person said. These are all hypotheses, but ultimately you want to test them.

[1:04:04] Of the hypotheses, of the sources of where you would generate your hypotheses, the [1:04:09] I would focus on looking at other sites and looking at your own site. So find... [1:04:13] what seems to be working on other sites and do that by [1:04:17] Again, putting them into Ahrefs, putting them into SEMrush, getting their top pages, trying to glean what patterns you think are correlating. [1:04:25] with success. [1:04:27] And then the same on your own site. So which pages are performing well, which ones aren't like an article that's 2000 words doing or the word count seems to be a correlation. Longer is better and shorter is worse. [1:04:37] Maybe that is causal. [1:04:40] then go test it and then validate it. I think also I spend a lot of time. So when I look at other people's sites, [1:04:47] one thing that I'll do that I've had a lot of success at [1:04:51] You're Steve. [1:04:52] contact the person who runs SEO at that company. [1:04:55] and then set up a meeting and then ask them a bunch of questions about, I think that this is what's what happened. Is that correct? Or... [1:05:03] What have you tested? What's worked and what hasn't worked? I, um, I remember I was trying to, [1:05:08] come up with an international SEO strategy. And so I did an analysis where I looked at every company and which companies had repeated success outside of the United States. So eBay, TripAdvisor, [1:05:20] a bunch of other sites. And there were clear patterns that TripAdvisor consistently outperformed over and over again. So then I analyzed TripAdvisor and tried to come up with what I thought was causing that. [1:05:31] And then I topped my network and tried to find [1:05:33] the person who's running an SEO TripAdvisor,

[1:05:36] which was hard to do, but I eventually found [1:05:39] Luke Levesque, who is running all SEO, including that. [1:05:42] And I asked him a bunch of questions and he answered some of them. And so now I actually [1:05:46] I actually know the answers. And so I don't even need to test some of these things. Like somebody else knows. [1:05:51] already built this and tested this, they can just tell me. So, [1:05:54] Finding the SEO manager, the people at these companies and just asking them the questions is actually a great way to do that. [1:06:00] to get those answers. [1:06:03] Awesome. Luke's coming on this podcast, actually, in not too distant future. [1:06:07] Luke's great. I'm looking forward to it. Yeah, he's a big deal now. You mentioned that some companies get a little blocked on thinking about strategy and just like planning and don't actually get anything done. And I think you had some thoughts to share just like how do you get shit done without being blocked if you can't get all the things done you want to get done? [1:06:28] The dirty secret of SEO and SEO consulting is, [1:06:32] is that [1:06:33] Most of what? [1:06:34] is recommended, never gets launched. [1:06:36] So we we interview [1:06:39] people from other SEO agencies or recruiting SEO at other companies. And the common thing that will happen is I'll say, [1:06:46] What did you work on? What was your strategy? Describe your strategy. And then they'll describe it in detail. And I did all these different things. [1:06:52] and then I'll ask what happened. [1:06:53] And they'll say, oh, well, actually, almost none of this got launched because it couldn't get resources. So the common problem is resources. [1:07:01] And, you know, every company is resource constrained. I think [1:07:05] A key part of why SEO especially gets disproportionately blocked is

[1:07:09] is number one, it's not part of the product org. [1:07:12] And the changes are... [1:07:14] Oftentimes, products changes. So Eli Schwartz has this book about product-led SEO, which is how it should be done. And it usually is not done that way. It's usually the product org has their stuff. And then there's the SEO team over here. And they're saying, please do this thing. And the product team says, well, I have all this other stuff that I do. [1:07:29] need to do so I'm not going to do that or I'm going to do that in a year. [1:07:33] And so [1:07:34] Having the org design and having things prioritized through product, I think is really important. Also, having buy-in at the executive level is really important. Usually when we get locked, I'll just try to ask the executive. [1:07:48] There's this really big opportunity. May we please get this thing launched. I think the other part of why SEO doesn't get resourced as it should is that it's mysterious and it's hard to describe it and people will... [1:07:59] say, well, how do you know that this is going to work? And you actually don't know if it's going to work. And so the fact that there's this mystery around it versus something else where there's not mystery. [1:08:07] the non-mysterious thing will get prioritized over the mysterious thing. And then the last is that it's just a lot of work. The stuff that I've been describing to you, to your point about there's a lot of meat, there's a ton of work to do at scale, like at Zillow. [1:08:21] You need a large team to do a whole bunch of different things. So you need a lot of resources to actually do all this stuff. And so it's just high scope a lot of times to do it. So for all these reasons... [1:08:30] The getting things done part is actually... [1:08:33] more of a factor of success or not than the actual strategy a lot of times. [1:08:38] There's a bunch of topics I want to touch on before we wrap up. This is already, I think, going to be the longest episode we've done, but well worth it because...

[1:08:45] Rarely have I heard this much depth on such a, as you said, mysterious skill and growth opportunity. So I'm really excited about all the stuff that we're talking about. There's a few things that you already touched on in my notes, but there's a few that we haven't. One is... [1:09:00] topical authority, which you've talked a bit about, but I'm curious, is there anything more to share about just the importance of becoming an authority at a topic as the core of your SEO strategy versus a keyword focus and things like that? This is probably the most important, least understood. [1:09:18] part of SEO. [1:09:19] So page rank was a key thing with Google where Google would prioritize things based on page rank. And it's based on academic publications where the journal article that was cited the most is better than others. So they applied that to their search engine. And that was a key part of why their search engine is better. And so it started with links. So the more links you have from other sites pointing to you, the more important you are, the higher you'll rank. And so people would build links and focus on that. [1:09:49] time that's gotten a lot more complex so there's more inputs the links is one and i mentioned referring domains that's part of links there's other inputs so non-seo traffic like i mentioned how much traffic are you getting from other channels [1:10:03] is a signal and google is looking at chrome for that as well so google actually knows your traffic sources and andrew [1:10:10] So non-aseotropic is an authority signal. Shares and social activity is an authority signal. Branded search. So the number of times people type in your brand name into Google, how many times people type in masterclass is an authority signal.

[1:10:22] you [1:10:22] So authority is more complex. So there's more inputs. [1:10:26] The second thing is that it's topical. [1:10:29] So it's not your total domain authority. It's what you're known for. And you'll see in the most recent Google helpful content update, they talk about domain experts or top, they don't use the word topic, but they refer to people where there's a [1:10:43] specific narrow expertise versus we can rank for anything like a Wikipedia. And so Google is looking for which sites are disproportionately known that have disproportionate authority for certain themes. [1:10:54] versus others. And so those themes come from [1:10:58] What is the anchor text of your backlink? Does it say chicken recipe or does it say digital camera? What is for shares? What has been shared the most? What is the content of what got shared the most? [1:11:10] branded search. So [1:11:11] master class plus gordon ramsey is authority for gordon ramsey so these are the inputs and that's how google's determining [1:11:19] those keywords and then they'll give you authority [1:11:22] for the... [1:11:23] things that are semantically related or adjacent to that. So an example that I'll give is [1:11:28] I was working with this recipe site and we had a viral blog post about jello shots. 10 jello shots worth the hangover. [1:11:37] And we got 4 million likes and millions of pins. It was hugely successful. [1:11:43] And so that post did really well. I think it ranked around number six for jello shots. [1:11:48] But the entire domain then started to rank for vodka recipes and whiskey recipes and things that were semantically related to Jell-O shots. And those pages were not even linked to from the blog post. So the blog post did well, but the entire domain benefited from Jell-O.

[1:12:06] From Jell-O Shot, [1:12:08] topical authority and from things that are adjacent to jello shots which are other other types of alcohol and so it was stuff adjacent and it was entire domain rather than single page [1:12:18] So that was a really interesting example. [1:12:20] And we see that a lot. An example with Masterclass is early on, they had authority for the names of the instructors like Gordon Ramsay, but not for food in general. And so... [1:12:30] We knew that if we had authority for Gordon Ramsay, then we'll have authority for things that are adjacent to Gordon Ramsay. So Gordon Ramsay is known for beef, Wellington, salmon. [1:12:38] how to poach an egg, steak, [1:12:40] And so we focused on things adjacent to Gordon Ramsay. And we knew we could rank there versus something like a butter lettuce, where we did not have authority for that yet. [1:12:48] We would not rank for that early on. [1:12:50] So we started for stuff that was adjacent to where our topical authority was. [1:12:53] And then we branched out from there. And this is actually a great way where earlier companies can actually rank, even if their competitors have way more domain rank than they do. [1:13:03] because again, Google is saying, [1:13:06] For a fintech site, maybe you don't have as much authority as Investopedia, but you're really known for Iron Condor, Crypto, I forget what it was called, but you're known for that thing. So what's your authority for Gordon Ramsay? What's your authority for… [1:13:19] just Gordon Ramsay relative to your total authority. And if you're disproportionately known for Gordon Ramsay, then you have very high topical authority for that. You'll outrank [1:13:29] other sites that have more total domain authority because you're disproportionately known for that thing when people think about the idea of a topic you mentioned a bunch of examples butter lettuce gordon ramsay is is it usually just like a topic versus a keyword a topic is like a one word or two word description of like a broad thing versus like caesar caesar salad with sardines and croutons what's like a good mental model for thinking about here's a topic that i can win versus like here's just a keyword

[1:13:57] The topic would be, again, we want to find keywords where the results are the same over and over again. So those keywords are now a group, that group of keywords, 300 keywords where the same results are showing up every time. [1:14:11] That's now a cluster of keywords. And therefore, it's a topic. The label of that topic is essentially, you know, what do you think is the primary theme of that? It's usually the highest search volume. [1:14:21] keyword within that keyword cluster like a butter lettuce so that's the name of the topic and then the way that you think about adjacent topics you can kind of think of adjacent topics is for food what does an ingredient graph look like so so for jello shots [1:14:34] There's drinks, meat, vegetables, produce. Underneath drinks, there's [1:14:40] juice and coffee and tea and then there's alcoholic drinks underneath alcoholic drinks there's whiskey and vodka and this eventually lower down there's jello shots so jello shots what's adjacent to that or other types of alcohol then you have other types of alcohol you rank for that what's adjacent to that [1:14:55] Other kinds of drinks. What's adjacent to that? Something else. But that's kind of how you can think about how the topic lives in a [1:15:01] in an overall graph is an ingredient graph or same with [1:15:05] with commerce, you know, home goods versus cameras. There's a whole tree structure. So that's how you can think about adjacent topics. [1:15:11] I wonder if there's a big map of these somewhere with all the traffic to each. So there is and there isn't. There are expert-defined taxonomies, and Google has them, but they typically don't use them. And they typically... [1:15:25] will generate grass somewhat on the fly because things will change and so

[1:15:29] I think people over... [1:15:32] overestimate the amount of expert curated taxonomies that Google uses. And it's actually a lot of [1:15:37] algorithmic dynamics definitions of a graph that are generated constantly and again they change so that's typically how it works [1:15:46] So, yeah. [1:15:47] How Google would actually find these adjacencies, I think it's by doing large crawls of the web and looking for co-occurrences and then structuring and graph that way, looking at search refinements. [1:15:58] looking at co-occurring results with ant sulfur butter lettuce here are all the results what words co-occur on the results there [1:16:05] So Google's basically, I think, creating a graph out of that. Part of why I say that is because we have [1:16:10] Some data suggests that. And then also there's a [1:16:12] Good SEO, the best SEO podcast. I think it's called Search Engine on the Record or something like that. But it's with Gary Isles and John Mueller. And I actually give quite a bit of information. There's one episode where Gary Isles basically... [1:16:25] That's what I just said. [1:16:26] where Google crawls the whole web, looks for co-occurrences and search refinements and all of this sort of feeding into how relationships are understood. [1:16:35] Cool. We'll definitely link to that podcast. A few more questions I definitely want to get through and then we can wrap up. One is I know you're a big proponent of internal links and the power of internal links and how that's often a underappreciated lever for growing SEO. Can you talk about that? Yes. So I mentioned I worked at this e-commerce company in 2007. [1:16:58] and [1:16:59] In 2007, Google Search Console had...

[1:17:03] a report of all of your pages and the number of internal links. [1:17:07] and they have that now, but the data are not all available. So it's, it's not, [1:17:13] It's not clean data. But then it was actually real data that was comprehensive and complete. So we could literally download and we had millions of pages and millions of products. [1:17:23] And I could download the entire set of every URL and the number of internal links that Google found. [1:17:29] And we had a bunch of product pages that were not getting... [1:17:32] traffic. So I wanted to understand why. [1:17:34] So I got that, all the URLs, all the internal link counts. Then I got traffic. [1:17:39] And then I basically did a correlation. I found that pages with more links got more traffic and pages with fewer links got less traffic. [1:17:45] So again, to the earlier question of how do we generate our hypotheses, we analyze data, we look for patterns, we generate a hypothesis. I think that links cause problems. [1:17:55] Traffic. [1:17:56] test that hypothesis so we made a change where we added more links to these products that [1:18:00] had no links and by links i mean links from within the site so within the site we're linking to these products more [1:18:06] And then we saw a huge increase. I think it was 300% increase. [1:18:10] in traffic to these product pages. [1:18:12] And so... [1:18:13] Then we were able to draw a clear causal relationship. And that was in 2007. That still works. And that's still a thing. [1:18:21] So, [1:18:22] I would say probably 90 to 95% of sites have this problem right now. [1:18:27] And, um... [1:18:27] The reason why this is a thing is Google is crawling down. [1:18:31] the web and your site through links.

[1:18:33] And if you don't have links, then Google doesn't have paths to find all of your pages. And it's not enough to put it in the sitemap. [1:18:40] you need it that's one link you need many paths for google to find pages [1:18:45] And the more paths, the better up to a certain point. [1:18:48] usually at least 10 links Google needs to. [1:18:51] to find a page and crawl it regularly enough to give it enough credit. [1:18:56] And um... [1:18:58] 90% plus sites don't have this. [1:19:00] The reason why is because... [1:19:03] Internal link algorithms are typically either [1:19:05] Facet-based, recommendation-based, or popular-based. [1:19:08] So for facet-based, what I mean by that is for thumbtack, [1:19:12] plumbers in San Francisco [1:19:14] They'll link to other cities and other service types. [1:19:18] or related. [1:19:20] will be and basically you'll have silos so you'll so you'll have pages you know you'll have cities and maybe they're not popular enough they didn't get linked to you'll have other services they didn't get linked to and so you'll have these gaps [1:19:30] in the site. [1:19:32] where you didn't have enough links to some of the [1:19:34] some of the pages. [1:19:36] popular, you're always going to have the same 10 most popular posts on every single page and the other... [1:19:42] Posts that are not the most popular don't get linked to most recent. Same thing. You'll only link to the most recent ones and not to the ones that are not the most recent related similar. So related algorithms are based typically on people who viewed this viewed this other thing. [1:19:55] And if you have a post that didn't get viewed a lot, you don't have data for that. And then it doesn't show up in the related algorithms. So related algorithms tend to skew towards. [1:20:02] a subset of pages. And so typically what you'll see is if you crawl, so you can look at this by getting Screaming Frog. So this is a good use case for Screaming Frog. Scrape the whole site, you'll get an internal link count. It's called Unique in Links and Screaming Frog.

[1:20:15] and you'll see a power curve where [1:20:18] 5% of pages are getting 95% of links. It's very extreme. [1:20:23] usually then you can see that a bunch of pages on your site get zero links or one link we'll usually look for at least five links [1:20:30] Ideally 10. [1:20:32] to say that this page has enough, and frequently they won't. [1:20:37] And none of these algorithms that I just described are going to solve it. So facet-based, related, popular, recent, none of those are going to spread enough links everywhere. [1:20:47] And so then when you fix it and you spread links everywhere, you're going to be able to do that. [1:20:50] then traffic goes up and essentially what you want is you want a shortest path from crawl point [1:20:55] So what I mean by that is a crawl point is where Google enters your site. [1:20:59] Google enters your site at your homepage, but it also enters your site on any popular page. So if your mobile app page gets lots of links, so you have a press release, it gets lots of links. [1:21:10] All of these are places where Google is entering your site. [1:21:13] frequently and it has a lot of authority. [1:21:15] And what you don't want is you don't want every page to be very far away in your link graph. You don't want to be very far away from these crawl points. You ideally want the few hops or few levels of your page. [1:21:28] site [1:21:29] from those crawl points you don't want a hierarchical tree structure you actually want a very flat graph [1:21:34] So how do you actually build this? This is a lot of work. So back to getting things done. [1:21:40] This takes a lot of effort. [1:21:42] So, [1:21:43] We actually built our own at Thumbtack. I worked on their internal algorithm. And then we and then at the shopping site that I mentioned 2007, we did the same thing.

[1:21:52] We've rebuilt that a few different times, but regularly the scope was too high. So we basically built that in-house. So we have an API that can power this on any site. We have it live on a bunch of sites. We have it live on Upwork. And other sites seen really good lift from that. But consistently we'll see. [1:22:07] If you don't have good link spread, [1:22:09] fixing it has significant opportunity, typically 25% plus. [1:22:13] opportunity, even like 100% opportunity sometimes if your links are really under optimized. So this is probably anything that I can recommend to companies as the quick win, the thing to do at the start, it's this. [1:22:27] What a nugget to have towards the end of this long conversation. Folks that are still listening, good job continuing to listen. And you found a great opportunity. [1:22:35] So to understand how this actually ends up being implemented, is it a simple way to think about it? There's a bunch of links in the footer and you smartly decide what links to include in that footer. [1:22:45] Yes and no. So footer links are not equal to body links. So links in the nav and in the footer are navigational links and they count, but they don't count as much. So Google will look at your page and separate out navigation. [1:23:00] But body links count a lot. And I think that the way Google is doing that is they're basically looking at the same pieces of information over and over again, or is it different? And so we'll usually put it at the bottom. [1:23:11] of an article or a bottom of a category page. You could put them anywhere, really. [1:23:15] It could be in the body, it could be at the top, it could be at the [1:23:18] It could be at the bottom, but I wouldn't put it in the footer. I would put it above the footer. Got it. Just hearing you talk about how this space works and how detective oriented it is and scientific, but also creative. I imagine it's very fulfilling to work in this space, just like intellectually.

[1:23:36] Is that what you find? [1:23:38] It's fulfilling now, but it wasn't when I started. When I started, I tested a bunch of things that didn't work. And it's very demotivating to do a bunch of work and none of it works. I think that's part of why people... [1:23:51] either don't go into SEO or when they do, they go out of SEO. [1:23:55] is that they try a bunch of stuff and it doesn't work. [1:23:58] But if you can persevere through them, [1:24:01] and start getting things to do work and start getting wins. And 80% of your company's traffic comes from the work that you did. That's very, very fulfilling. I think the other fulfilling thing is, [1:24:13] There's a certain profile of a growth person or an SEO person who really likes to do things they're not supposed to do and to break things and to do things that are unconventional. [1:24:23] And I'm that way. And so for those reasons, it's very flippling. [1:24:28] A couple more questions I want to get through and then I'll let you go. One is, I've noticed this proliferation of startups that are using GPT-3 basically to generate pages and to win SEO through just auto-generated AI-driven pages. Do you think this is an effective strategy for a startup? Do you recommend it? Do you think it's long-term going to last? How do you think about AI and SEO? [1:24:52] I mostly don't like it and don't recommend it, but there are exceptions. [1:24:57] So... [1:24:58] a couple things on this. The first thing is the content should be good and it should be useful. And we've played around with GPT-3. So GPT-3 is basically generating sentences that are similar to other sentences that have been written by somebody else. So GPT-3 takes...

[1:25:12] common crawl and very large data sets of a large snapshot of the whole web. So tons of sentences and then it trains a model to generate new sentences that are similar to other sentences that have been written. So they're grammatically correct or spelled correctly. [1:25:25] Sometimes they sound like they are written by a person. Sometimes they're not. But a lot of times they are. A lot of times it's indistinguishable from a human. [1:25:33] The problem is that [1:25:34] You want the sentence to be [1:25:36] factual. You want there to be underlying wisdom in the sentence. So I'll give you an example. We were playing around GPT-3 to generate a description of a product, of a soap. [1:25:47] And GPT-3 said the soap was oxide-free, but it was not oxide-free. But GPT-3 doesn't know that. [1:25:53] And some other thing was oxide free on the web. And so just said it was oxide free. So it made this false claim. [1:25:59] But if you're a user, you would never know that. It was written well, it was spelled correctly, it was grammatically correct, and it was just factually inaccurate. [1:26:06] And so, [1:26:07] To have content be useful and to use AI in a useful way, there needs to be underlying wisdom that the AI is communicating. [1:26:16] If you have structured data, [1:26:18] then it can be useful. So let's say that you're writing an article about a [1:26:22] a basketball game. [1:26:24] If you have structured data about each thing that happened in the basketball game, somebody scored this point at this particular time, and then this happened, something else happened. [1:26:32] You generate sentences that are factual, they're based on actual information or knowledge. [1:26:37] And the sentences are therefore useful. [1:26:40] with power is actually an interesting example. So they're extracting

[1:26:44] structure from clinical trial information and generating sense based on that structured data. [1:26:50] So in that sense, it's useful. But most AI-generated content... [1:26:54] It's just sentences that are similar to other sentences that have been written. [1:26:58] And so in general, I recommend against GAI, an example would be, how should I save for my retirement? [1:27:04] If I have an article about that and it's just not truthful and I'm making decisions about my retirement, I'm going to be a little bit more. [1:27:11] Based on this article, the GPT Throat, that's... [1:27:14] That's really bad. Or if I'm, you know, if I have cancer and I'm researching cancer options and this AI content suggests some treatment that. [1:27:23] is not a good treatment. [1:27:25] That's a big problem. And so... [1:27:27] I think we want to be really careful about how we use AI for content generation. Where AI is really useful is actually... [1:27:35] all the other stuff that I talked about. So if you think about the workflow of creating content, you start with what should I be writing about? What are the subtopics? So like, [1:27:43] 300 different keywords. What are the sub themes? [1:27:46] What's that outline? How is it performing? [1:27:48] All of these things can be done really well with AI. [1:27:51] So AI can extract structure. They can help me decide which topic to write by going through these thousands of keywords and clustering them and looking for overlaps and things like that. [1:28:00] It could tell me what my topical authority is. [1:28:02] They could sort them. [1:28:04] It could go through the keywords and find these subtopic themes. It could do that, all of that really well. [1:28:09] and then a human can write a piece of content, [1:28:11] So given all that information, [1:28:13] Human can write a piece of content. They're a domain expert. They have wisdom.

[1:28:17] And they can write an article based on that structure. That's a great application of AI. What's not a great application of AI is writing an article, suggesting cancer treatments based on no underlying wisdom. [1:28:28] That sounds dangerous. Say someone is listening to this episode and they're like, okay, I'm excited. SEO is something we should be thinking about. What is the first... [1:28:41] And second thing they should do after finishing this episode to explore the opportunity. [1:28:45] First thing that they should do is assess the size of the addressable market. So is it large? Do that by finding, again, product competitors and audience competitors. How much traffic do they have? Is that meaningful? The second thing that they should do is they... [1:28:58] Can I compete? Do I have enough authority? So how much traffic do I have that's not from SEO? How many referring domains do I have? What is my total authority? [1:29:06] How much traction do I have? [1:29:08] Then they would want to ask, of all the things I could be working on, how does SEO compare with all of my alternative things that I could focus on? [1:29:16] And then what's the scope of that? [1:29:18] So do I want to spend my money on that or do I want to spend it on something else with that amount of money? Does that make sense? [1:29:23] That's what I would do. [1:29:24] Ethan, this episode is bonkers. I think there's millions of dollars worth of value in this last hour and a half that we've been here. I am so appreciative of you making the time. I'm going to go eat some butter lettuce for sure. Before we wrap up, where can folks find you online if they want to reach out or learn more about Graphite and you? And how can listeners be useful to you?

[1:29:54] And one thing I'll add is I talked about how I learned a lot from asking other people at companies, uh, [1:30:02] what worked and what didn't work. [1:30:04] And what you're doing with your podcast is essentially getting those people on a podcast and sharing it with everyone. So. [1:30:11] When I was starting in growth, I had no information or no guidance other than that consultant that I mentioned to point me in the right direction. So it was really painful. So. [1:30:19] The fact that there's an operator who's talking to other operators and practitioners and getting that information, giving access to everyone is really meaningful. So I just wanted to thank you for that. It's my pleasure, man. It's really fun for me, too. And I'm learning a ton. So thank you. Thanks for being here. Absolutely. Thanks for having me. [1:30:49] podcast. You can find all past episodes or learn more about the show at Lenny's podcast.com. See you in the next episode.