RS308: Programmatic SEO with Stuart Brameld

[00:00:05] Speaker A: Hello. Welcome back to rogue startups. I'm your host, Craig Hewitt. Today we're talking all about growth, growth experiments, growth data, programmatic SEO, so using code and automating, a lot of content creation and SEO with Stuart from growth method. Stuart and I have known each other for about five years, ever since the first, first and only big snow tiny comp Europe. So kind of ski trip with me and three other founders in Chamonix, France. Stuart is one of those. We've kind of stayed in touch ever since then. I think it's kind of the power of this like world that we're living in these days. You can meet somebody once, stay in touch and then have them back on the podcast like this to talk about something really, really interesting. And Stuart's doing some really cool things with programmatic SEO and running growth experiments. [00:00:49] Speaker B: And I'm excited to share that with you today. [00:00:51] Speaker A: Let's dive into this conversation with Stuart from growth method. Okay, so Stuart, I've known you for, I was thinking about it five or six years. You went to the one and only kind of ski retreat we did when I lived in France. That was pretty awesome. I think about that weekend a lot, but we haven't caught up in a while until recently and love to dig into kind of what's going on with, with you and growth method. For folks who aren't familiar with growth method, you want to give a quick update on who you all are and what you do. [00:01:26] Speaker B: Yeah. So the website says, I think project management for growth marketing teams, something like that. But I find it's quite useful to describe maybe what it isn't rather than is first. So Asana and Trello and ClickUp and all the project management tools, everyone knows a lot of the focus is around activity rather than impact. So there's loads of like tasks and subtasks and delegation and Gantt charts and list views and calendar views and like all this stuff that isn't focused around impact. And so we're trying to basically do the opposite. So we connect to Google Analytics or Salesforce or HubSpot and then like everything you do in the platform is tracked against a growth goal and a metric. So it's basically trying to make it very hard to do work that doesn't have impact. That's still a work in progress. We're got a good number of customers on board, but yeah, definitely not finished yet. [00:02:26] Speaker A: Got it. So I know from chatting that you try to focus a particular type of customer for this because I would guess smaller companies like ours, maybe not a great fit. Big old huge companies like Asana maybe not a great fit, right. Because they're just probably too big to fit into an opinionated tool and a workflow like you kind of espouse. Is that fair? [00:02:51] Speaker B: Yeah. Customer size is like really mixed. So I've got a couple of customers with like a team of five people, like the whole company, and then, like one at 50 and then one at like 7000. So that doesn't seem to be fit in terms of size. I think it's more culture and willingness to adopt more of an experimentation type approach to marketing, like growth and experimentation. I find as we'll come on to, I think a lot of the, most of people reach out through SEO because they've been, like, googling and trying to find this stuff. So the first call I'm typically hearing something like, I came across you guys having looked for some way to manage experiments better, something like that. And I found I had a few calls with more traditional b, two b marketing teams, and trying to change the culture of a team that's already not operating in that way. It's not fun and often not possible. So I'm trying not to do that, at least not right now. [00:03:50] Speaker A: Yeah, gotcha. Gotcha. Okay, so kind of mindset more than company size, experimental, you know, companies who embrace really, like an experiment first kind of mindset with. When it comes to growth. [00:04:03] Speaker B: Yeah. Or a willingness to, willingness to adopt it. Like, if you're a big enterprise company that runs a marketing campaign every quarter with little data and like the hippo says, this is gonna be the subject of this campaign and we're gonna spend 100 on ads and we don't really know what's gonna work. That's not where I'm at. [00:04:24] Speaker A: Yeah, gotcha. Okay. What is like, like growth experiments? What does that mean to you or, like, to your customers? [00:04:35] Speaker B: So I can describe it from the sort of software point of view. So when you create, you can think of an experiment, like a task, I guess, and I've been through, should have been a task or a sprint or an experiment. Like, it sort of doesn't matter what it's called, but various things are sort of set in stone. You talked about opinionated. Like, software is quite opinionated. So when you, I guess the first thing is an experiment is 28 days long. So in growth methods, when you get to 28 days, we automatically move it into analyzing, so it's not in progress anymore. So you basically can't go longer than 28 days without recording some learning. And then there's various other things like, you have to write a hypothesis. The hypothesis has to be aligned to a goal which is preset in advance and tied to something like Ga four or salesforce or HubSpot like we talked about. You have to include data in that hypothesis and then you start running it, basically whatever you're going to do. So rather than maybe a simple example is instead of a company that spends lots of time and effort publishing blog posts but never going back to them and looking at them again, we'd be like, come up with a hypothesis of why you're going to write on this topic and what traffic do you think you're going to gain from it, or conversions, whatever your metric is, outline like your reasons behind that. Run it for 28 days, see if you've got some positive impact, like you're heading in the right direction. It might not be. If your goal is conversions in 28 days, you're probably not going to get conversions or you might not. Depends on the size of the company, but you should get some positive signal you're moving the right direction and that should give you the confidence to either like move forward and try an exoration or scrap it. So you're just sort of trying to increase that kind of velocity and learning and sharing those learnings across the team as much as possible. [00:06:21] Speaker A: Gotcha. Gotcha. Okay, so 28 days. To me, that's kind of like a good, like short to middle term view. Like you said, you might not be 100%, like have an answer, but you'll have like some, some data that you're probably trending the right direction. A couple questions around that, like are we talking like page views or conversion events or time on page? Or like what's a, what's an endpoint metric that people try to like align that goal with? [00:06:49] Speaker B: Yeah, mostly conversion. Some conversion metrics usually like leads, mqls, trials, demos, meetings, books. [00:07:00] Speaker A: And how many experiments do you suggest people run at a time? [00:07:05] Speaker B: It's a really good question. So we've got like a health metric and there's a lot of back and forth around what this metric should be, but it's basically experiments completed in a month and we give you an average. And most of the new teams I'm basically aiming for between one or two to keep it really simple to start with, and then it moves up from there. But you can imagine with content, if you were to adopt this approach, that you get to a point where you've got a baseline number of pieces of content and then you're almost forced by the software to come back and review them every 28 days. So if you're trying to get to position one for a piece of content and you're currently like, on page five, it might take you three or four iterations to get to position one, but you're continuing to moving towards that goal and learning along the way. And then every time you complete it, you share the learnings with the team, talk through them, etcetera, etcetera. So it's sort of encouraging people to be moving in the right direction and to adopt this, like, more iterative mindset and not like, I published it, so it's done right. Got it. Actively try to prevent people from not doing something once they completed the first experiment. So you do that 28 days, you maybe analyze the metrics, you find you're on page three and you're trying to get page one. Like, when you hit complete, it pops up with a thing and says, do you want to share this with stakeholders? And you can copy a shareable link that's valid for 40 days. And then the primary button is, I can't remember quite what the language is, but something like, iterate on this or create a new version. I think it's create a new version. You basically encourage to, like, do the next thing, because in most situations, it's quite unlikely that you're done. It's perfect. There's nothing that could potentially be improved with the thing that you've just shipped, whether that's ads or SEO or a b testing, constantly thinking about what could I do to make it better. [00:08:58] Speaker A: Okay, so the counter argument to this kind of testing based growth mindset that I've heard is most of us, and I'll say kind of like small to medium sized businesses are businesses that don't have a ton of page views or users or impressions or conversion events, don't have enough data to be testing things as, like, a growth mechanism. You know, it's more like just a, like, on off tests or. Or just don't test it all and go with your intuition. Like, what, what's your position on that for? Like, so I'll just say, castos, we have about 100,000 unique website visitors a month. We get hundreds of trial starts a month, and we convert, you know, 70% of those. Like, do we get enough data to run these kinds of marketing and growth experiments? [00:09:46] Speaker B: Yeah, you've got a lot more data than a lot of my customers. Yeah. I think a b testing and statistical significance has given experimentation, like, a little bit of a bad rap or, like, I think there's most business decisions are made without a lot of supporting data. Like, you're not, you don't. If you think about like the number of decisions that you're making on a daily basis, most of them aren't made with statistical significance. So, so like, to kind of put quite that amount of pressure on a marketing team I think is not particularly helpful. You can still, like, you can still track metrics and feel confident that you're moving in the right direction without needing that. And I think the sort of b two c e commerce world where you've got a really short transaction cycle, like you're buying a pair of shoes, the transaction happens straight away. There's millions of people looking at the site and we're like, well, if I add this kind of free delivery option below the buy now button and test it with our millions of users and we get an uplift, then, like, that's how a b testing should work. And there's just so much more to sort of experimentation, I think. I think a huge amount of it is the learning piece. And if you're going back to my original example, if you're writing a piece of content, I keep coming back to SEO just because I know it's something we're going to talk about a little bit. But like, if you write a piece of content, publish it and you're done and you never look at anything afterwards, you're learning nothing. Whereas if you go through this approach and start to understand, like, how long does it take us to rank? Which page do we normally rank on? How long does it take us to move up the things up, the rankings, like what sort of thing? If we start doing loads of internal linking, what impact does that have? And you share that with the team, it just goes out via emails. As you're going this way, you have so much more insight and understanding, which develops into the likelihood of your next experiment working just increases over time. [00:11:44] Speaker A: Yeah, I know. I remember Rob walling talking about a marketing change log. We have it with development. We know we have source code and version control and everything, but we often don't have this for marketing where like, oh, I updated the homepage copy here, or we added this blog post here to go back and correlate the data you see in Google Analytics or in chart mogul or in the bank account at the end of the day. And kind of what I'm hearing is this is a much more kind of beefed up version of that for active marketing teams who want to be able to kind of track and attribute things they're doing. To like, actual outcomes. And the bottom line. [00:12:21] Speaker B: Yeah, it's really interesting. At one point, I think the positioning was around like GitHub for marketers, which I thought was quite interesting. I quite like that idea. I don't think it's particularly in the design phase. So you move through a couple of stages within growth method and there's indesign or planning running, analyzing, complete that planning bit like we don't require, we don't require anything other than a small piece of text to describe what you're doing. The most useful function of just having some idea of what you've done or are doing is firstly, the team get a notification as soon as something moves in progress. So like, if you're one of a team of ten, everyone knows you've just shipped something without having to like find it. And also, as you said, if it doesn't work and you come back and you decide to unwind what you've done, you've got a clear idea of what you've done. But that's like points or a sentence or something like that. [00:13:18] Speaker A: Yeah, yeah. I mean, cause for us it's just someone pops something in slack. Hey, I published this video or this blog post is live, or you get your own email. It's like keeping track of marketing efforts is challenging some. [00:13:32] Speaker B: Yeah, yeah. [00:13:33] Speaker A: Cool. Okay, so I want to switch gears sort of to talk about programmatic SEO. So kind of. Well, why don't you define how you view what programmatic SEO is and I'll see if I understand it. Right. [00:13:51] Speaker B: Yeah, I thought about this before we chatted, actually. I think the sort of true industry definition is probably more like what the likes of transferwise and Zapier do with, like, they have products that lend themselves very well to automatically creating SEO content from the product itself. Right. So transfer wise have thousands of pages for every variation of every currency exchange in the world. And I think every bank rate for every currency exchange in the world. And zapier obviously do the same thing for their zaps. So you can do like connecting HubSpot to Google sheets, connecting HubSpot to slack, connecting HubSpot to everything. Right? And again, thousands and thousands of pages. So their products lend themselves really well to that. And you can automate, they can automate a huge amount of content. That's definitely like the traditional and probably best way of doing it if there's an angle within your product that allows you to do it. What we've started doing at growth methods more is like, is there a way to adopt some of that, like programmatic approach without pulling it solely from a product. So probably the easiest way to describe it is I have an airtable database with a ton of content in it. And a lot of that content powers what's on the WordPress site. So there's tables for like growth courses, growth consultants, growth experts, growth agencies, growth hacking agencies, all of that kind of stuff. And that powers a lot of content in a sort of programmatic type way. [00:15:40] Speaker A: Okay, cool. So we'll use a fictional example and I want to walk through how we're thinking. We're embarking on a couple of programmatic SEO initiatives, some powered with AI, most all kind of powered with AI. And I think this is really cool. So let's say we are a review site. I think this is a fair thing. We want to create a bunch of a versus b comparisons. So Ford, f 150 versus Chevy, whatever it is these days or these elgato lights we were just talking about, versus some other thing I'm imagining we build spreadsheet with sort of like rows for each of the options. Right. And then, or columns for each of the options and rows for like price and image and description and pros and cons and best for like use cases or something like that. Right. And then you take the spreadsheet and somehow get that into like a structured type page in WordPress for each of the permutations of a versus b and a versus c and a versus d. And you create hundreds of pages like that. [00:17:05] Speaker B: Yeah, exactly. Yeah. [00:17:07] Speaker A: Okay. [00:17:08] Speaker B: And that the versus bit, actually with I used OpenAI, but the ability for it to now browse using bing search, it does an incredibly good job in the testing that I've done creating a versus b in a table without you having to give it any information whatsoever. So if you could say create HTML comparison table of Trello versus Asana, and it will list out like ten core features and benefits and tell you what's got what in it without you having to power anything. So yes, exactly as you said. But you may need to do less work than you think, particularly if the products or services you're comparing are kind of pretty well known. Like it's likely to be in their model because it's a bigish company, then they'll do a pretty good job. [00:18:00] Speaker A: Yeah. Interesting. Okay, okay. [00:18:02] Speaker B: And I did it for exact taste and it came out like the content was. [00:18:06] Speaker A: Yeah. [00:18:07] Speaker B: I'd say like probably as good as I could have done on my own with like half an hour of time. [00:18:14] Speaker A: Yeah, yeah, yeah. Don't let me forget. I want to come back to, like, if that's the case, is this still worth doing? Because I think that's, that's just like a big question. It's like if the bar is so low or the barrier is so low. But, but before we do that, we haven't quite, we did what a lot of programming tutorials did is you do this and you do this and then you have a Ferrari over here. But we skipped like the 87 steps to get the Ferrari. How do you go from the spreadsheet to WordPress? [00:18:43] Speaker B: I use make.com, but you could use zapier or I think there's loads similar. Right? Most people I, most people seem to use make or zapier seem to be the big two. [00:18:53] Speaker A: Okay. [00:18:54] Speaker B: So. And yeah, it just pushes it straight into, there's literally a zap or scenario in makes terms for push to WordPress or create a post. [00:19:06] Speaker A: So you have like when a new row is added in this Google sheet, make a post in WordPress with these attributes. [00:19:13] Speaker B: Exactly. Yeah. [00:19:14] Speaker A: Wow. Okay. [00:19:15] Speaker B: And where it gets quite interesting is, so I have a database like an airtable base with all of this like content in this example, like comparison content in it and then a separate one for posts on the website. And then every time I push one, I record it in the posts table. And then you can start doing stuff like they're all tagged as well. So if there's a topic that I'm writing about, trying to generate some topical authority about something like prioritization frameworks works really well for us. And there's like, they're all tagged as prioritization framework. And then when you write a new one or update one, you can say like go and query that post database and tell me all the other ones. And you can then related content and interlinking. And then you could tell OpenAI when you're writing about this reference related posts that it's pulled from airtable. [00:20:12] Speaker A: Okay. [00:20:13] Speaker B: And it'll do like, you can tell it, update the last updated time, publish my little bio, stick this CTA in the middle of it. Generate a meta description for me from the content you've created. It gets, gets really interesting. [00:20:26] Speaker A: Yeah, I mean, this is in the weeds for folks who don't use WordPress, but all with the block editor. Are you pasting in HTML or just plain text or what? [00:20:38] Speaker B: So I do the formatting via make.com dot. So I'll literally wrap it in a p tag or an image tag or whatever it might be and then just push it straight. WordPress, I think it's like when you view it. It's native pre Gutenberg, but you can just hit convert to blocks and it will show you all those blocks if you need to edit anything. [00:20:56] Speaker A: Right. But on the front end it's just good. [00:20:58] Speaker B: Right. It's all formatted. [00:21:00] Speaker A: Yeah. Interesting. Okay, we're halfway through. We're to the point where we have about 90% of the spreadsheet and we're struggling with how do we go from the spreadsheet to get it into WordPress? We've looked at a plugin called WP Allimport, which lets you map spreadsheet or CSV fields to a post. So you could do like post title image, like featured image, all that kind of stuff. But I think the place we're struggling a little bit is like, it's not exactly the same as what you're talking about, I don't think. Because we're literally going to have for each post two different sets of content because we want to do comparisons. So it's not just like this is a review and these are the structured data about the review, but it's like this one and then this one. So I think we're struggling a little bit with how to do that. We may have to do like a custom page template or something a little more fancy with that. [00:21:59] Speaker B: And I think WVP import, I used to run WordPress agency, so I know a little bit. I think it's like, would be a one off or you, maybe you could schedule it, but I think it's more built for like, we're going to do an import for this period of time. Right. I think the advantage of the make Zapier type approach is it's more like always on in terms of your airtable database, of your features and comparison stuff. And if you find something you new tomorrow and one of your team goes in and updates it, it could be just polling it every day and updating the content or GPT five comes out and suddenly the content is written way better. You can just say like, rerun all our posts today. Yeah, I do the same. I put glossary content and if I add something to it, it just reruns every week. So then you know, like freshness as well, from a Google perspective is, I think it looks like it's continuing, getting updated, which I guess is what they want to see. [00:22:54] Speaker A: Yeah, no, it makes sense. Makes sense. Did you have any kind of fear about, and I'm sure you tested it, but what if this goes wrong? We're going to publish 100 blog posts that are all wrong and how are we going to undo that? We don't really know what we're doing. That's one of my fears. This project is 300 articles for us, probably. Yeah. And so, like, push and go on. That is. And there's not. I don't see an easy way to just do, like ten and then pick it up from eleven to 300. So how did you approach that? [00:23:29] Speaker B: I haven't done 300, which is probably. [00:23:31] Speaker A: The. [00:23:33] Speaker B: Time is probably the most I've gone to so far, so I can just keep an eye on it. I mean, I don't see why from a technology point of view, there's no issue. I think the only risk would be, like, Google goes, why have you just posted 300 articles when you normally do one a month or week or whatever it might be? Yeah, yeah. So maybe like 300 sounds like quite a lot to me. I'd be tempted to do like 20 or 50 at a time or something like that. [00:23:57] Speaker A: Yeah. [00:23:58] Speaker B: But again, if you use the kind of make zapier approach, you could just tell it to run one a day or one a ten a week. Yeah. Less of a big bulk upload. Yeah, yeah. [00:24:10] Speaker A: And I may be getting kind of ahead, making it more complicated than it needs to be, because I think some of the value of programmatic content isn't necessarily, like, is this the most beautifully designed, like, landing page ever? But, like, if we want to compare, like, the Ford f 150 versus the Chevy, whatever it is, if we just have, like, one, and like, if we just have the Ford and then we have the chevy below it and it's all kind of plain text and bullet points and a couple of images. I don't know that that's so much worse than a bunch of beautiful tables and all of this stuff. How much ux did you go into when you were creating these pieces? [00:24:51] Speaker B: Not a lot. And at the moment, they're pretty plain and vanilla. I would say the blocks that would be styled in Gutenberg are styled. When that gets pushed into WordPress, like a quote block, for example, looks decent and an image, like, looks decent. Obviously the text is formatted as it would be if you wrote it. So we don't really go outside. Like, they're not landing pages, they're posts, which is maybe part of the answer. Yeah, but it's very much the goal is sort of search traffic over anything else at this point. But where I quite like this programmatic approach is, I think there are a huge number of options for improving the design aesthetic quality of the content once that baseline is in place. So I've been looking at things like using Bannerbear to automatically create a really nice looking quote block in our brand and incorporate that into it. Then pretty soon I imagine the Dali image generation stuff will be good enough to say, create me a custom image based on this post and you put that in there and probably far off. That will be like creating a little video that describes this thing with my head in it. It's probably not that hard in the future as well. So I think if you've got the base content and this ability to connect these different services, I think there's tons that can be done with it. [00:26:14] Speaker A: Yeah, cool. Any kind of learnings that you want to pass along around, like how you're using chat, GPT and OpenAI for the content creation. [00:26:26] Speaker B: I use the I can't what it's called now, custom instructions, I think it's called something like that. So I use the API through make or zapier. But I have a set list, I can tell you what it says, actually. So I have a number of instructions, like five little bullet points that basically say you're a growth, marketing, agile marketing expert. You write for an audience of B. Two b marketers use UK English language only, right? With the comprehension of readability level of a senior school student. Use active voice where possible. Like those prompts, like, they may not be the best prompts in the world, but they make a huge difference to the quality of the content you get back. So I think, yeah, think about those and make sure that powers like everything that you use. [00:27:09] Speaker A: You know, like this is a bit of a detour, but I am of the mind that I, I think I would like to hire like an AI coach or something. And that sounds so weird to say, but I feel like there's so much opportunity with all this and I mean, it's just, yeah, it's like, it's probably a perfect place for a coach. Instead of me going into chat GPT and futzing around with all this shit, if I could pay someone $50 an hour and meet with them once a week to show me. Okay, if you want to create content and you want to do this thing, these are the eight things you need to include in your prompt. Like, I can just imagine if you do that every week, how that compounds. Okay, so, audience, if you are up for this $50.30 minutes once a week, let's do it. Send me a message cragastus.com because I think it's huge. I mean, the little bit that I use chassis, bt and AI, it's amazing. It does so good to that 90% level, you know, and, like, fuck, if I could. If I could pay a little bit of money to get that last five or 7%, like, 100%, I would do it. And then it's not just help me write blog posts, but, like, help me automate the rest of my business, analyze our financials, help me manage this person, help me do all this stuff, because I think that, like, I mean, there's. I think there's so much about what we do right now that AI will change. I'll say at least, you know? [00:28:44] Speaker B: Yeah. So I first sort of started exploring it after I tried multiple times hiring writers, and maybe I just had bad luck. It just did not work. I just found it so frustrating. Like, you put a job out, you get loads of responses, you have to trawl through them all. I do apply test kind of page tests to a few people and, like, whittle it down. And then, like, someone's like, I've got another job. Well, it's a waste of time. And just, like, the content comes back was like, it was good, but it wasn't amazing. And I think I was paying reasonable rates. And I got to this point of, like, and then the business model just seems. And then I think it was around the time that, like, chat GPT was released and, like, the writers are charging by, like, per word or by numbers of words, and I'm like, this is, like, nuts at this point, given, like, where the technology's at to be saying you're going to try to extend a word just seems so far away from where the world is going. And as you said, the more you get into the sort of AI tools, the more you realize how capable they are. And I got to a point where I'm like, I think if I power this thing, it maybe takes a little bit of my time in terms of coming up with this framework and the structure and the tools and everything, but the end result is not far off what I get from a copywriter. And so it makes more sense to my time to do that than it does to use copywriting. It wasn't even a money thing. It was just a kind of quality of content efficiency. [00:30:18] Speaker A: No, I mean, I've heard folks like Brian Castle say they've split tested human generated content and AI content, and the AI content wins across the board, like, blind taste test kind of things, but also, like, traffic and engagement and rankings and stuff. So I saw on Twitter this morning, Josh Howard, who was a guest on the podcast before, has a thing where he's having AI write the entire script for his YouTube videos and it's kind of like a blind channel and he's doing all of the growth and all the scripting just with AI and he's taking it directly from there and he's shooting the videos. So I think I'm using a little bit of chat GPT to write YouTube scripts for me, but it's definitely far off because it doesn't sound like me. So you can say, hey, write me a YouTube script on this topic, but then it sounds like chat shibti and not like me. Cause I have a certain tone and style and stuff and I think, I don't know, for me, I think especially, like, when it's you talking, it has to be even more on point than written content. [00:31:16] Speaker B: Yeah, absolutely. Yeah. The other bit I just wanted to touch on was I think you still need a smart, knowledgeable person to power it. I think this idea that you can have no inside expertise in an industry and just create a load of content doesn't work. I know you did your mini series around YouTube SEO. I think all of that stuff still applies. I think using AI, if you're trying to use it for SEO traffic, using AI without understanding like, the basics of SEO and your industry really well won't help you because you still need, like, the content structure. You need to give it the right questions, the right context. You need to understand the topics that you're going to rank for and how these different articles within those topics interlink. [00:32:07] Speaker A: So, yeah, 100%. Yeah, yeah, yeah, yeah. And I think the last bit that, that is a bit of a skeptical take for me on programmatic content in general, but also just like a lot of content that AI can do really well today is it's really top of funnel, right? Like you're probably not going to get chat GPD to write you like a conversion cop focused landing page for you, castos versus captivate or whatever, right? Like, probably not going to do as well as I could do, or Dennis from our team could do with all the context that we have, but it sure as hell could write a great microphone review article or listicle of best true crime podcasts or something like that. So that's my only kind of hesitation is I like to write content as low in the funnel as we can, and at least so far, it's not doing a great job of that. [00:33:00] Speaker B: Yeah, I agree. I sort of, in my structure, things like, I use the AI, most of the AI content is sort of child's content. To a parent theme, if you like. I wrote around attribution recently, for example. So, like, AI generated tons. Like, not tons, like maybe ten or so articles around, like, what is time decay, attribution, multitouch, multi channel attribution, all the different attribution models, right? First touch, last touch, et cetera, et cetera. And then I write the, like, what is attribution parent article that links all of those together. And that sort of model is worked across all of that content. Really, it's like you can't get it to write the kind of more nuanced, top level article, but it gives you, in my experience, gives you some topical authority to rank for that one article. [00:33:58] Speaker A: Cool. So circling back to programmatic, what are things that, when you are rolling this out, or if you're talking to me and we're about to publish 300 articles, what are things that you feel like people should be, like, aware of or worried about as they're getting into this? [00:34:18] Speaker B: I'd say start small. Like, maybe not. [00:34:21] Speaker A: Come on. [00:34:28] Speaker B: Yeah, I'd like get into the weeds with maybe try it for a few, a few articles. But like, more kind of figuring out. You're going to give it a pro. You're going to give like, OpenAI a prompt and it's going to give you a response back. Obviously, if you give it the wrong prompt, you're going to get a response back. Or it might start the sentence in a wrong way. So you're going to say, like, if you have a heading that says, I don't know, what is the Chevy 150? If that was the Ford 150. [00:34:55] Speaker A: Ford. But it's okay. [00:34:57] Speaker B: Ford 150. You probably don't want it. You might want to start in a really specific way, like maybe you want the forward 150 at the start of the sentence, for example. And sometimes you get the wrong prompt. It might come back with, like, I don't know, we can't browse. Browsing isn't available right now or something like that. So you need to be careful that you don't get those kinds of outputs. So I'd say start with the really simple. I started with glossary content. What is content? Mostly text. And then go into the versus content, as you said, maybe put listicles in middle. Actually, I go glossary listicles versus. And then start experimenting with, like, creating tables, adding quote blocks, including video content, those sorts of things. Yeah, yeah. [00:35:41] Speaker A: Cool. Cool. [00:35:42] Speaker B: Interesting. [00:35:42] Speaker A: I'm going to give a shout out. We had Spencer Hawes on the show probably just a couple episodes ago. As this goes, live is a fantastic plugin for WordPress called Linkwhisper, which automatically in bulk, adds internal links between articles. And so I think it's a perfect match for something like this because you go publish those 300 articles and then just go and say, add links. I mean, you have to go into each one, but it picks out, okay. From all these other places. Here's where you could link from that article to this article and then vice versa. So you would probably have to manually go through each of those 300 articles and add the handful of internal links. But it sure as hell beats manually doing that and finding like you were talking about finding all these existing articles to link to. So I'm sure he's thinking about how to automatically add those links, but there's a little bit of review and suggestion you have to do still. [00:36:37] Speaker B: Yeah, yeah. I'll take a look at that. Sounds careful. [00:36:39] Speaker A: Yeah, yeah. Neat. [00:36:42] Speaker B: Cool. [00:36:42] Speaker A: Stuart, anything you want to leave folks with on kind of growth and experimentation? Programmatic SEO? [00:36:51] Speaker B: I don't think so. If anyone's got any questions, very happy to receive emails and messages and tweets and anything else? [00:37:00] Speaker A: Cool. [00:37:00] Speaker B: I don't think I can share. Yeah, we can catch up afterwards if I can share some of the scenario in the make world scenarios, if we can share links to them. I think some of them you can share, like, public links and then the show notes as well be quite interesting to see. [00:37:13] Speaker A: Beautiful. Cool. So we'll either include links or maybe just, like, screenshots in the show notes. That'd be great. So awesome. Stuart, thanks so much for coming on. I appreciate it. I'm going to go dive into some AI and programmatic SEO. This is fun. [00:37:28] Speaker B: It's Valentine's Day, so I'm definitely not going to be doing that. [00:37:32] Speaker A: Great. Enjoy. All right, thanks.