AI is reshaping SEO faster than most marketers can keep up—and the rules you’ve relied on for years may already be outdated. In this episode, Patrick Stox reveals what’s really happening behind the scenes and what you need to do now to stay relevant.
https://page2pod.com - AI is transforming SEO faster than ever—and not everyone is ready for what’s coming next. In this episode, Patrick Stox from Ahrefs breaks down the real impact of AI on search, content, and the future of SEO jobs.
From zero-click searches to programmatic content and edge SEO, this conversation dives deep into what’s actually working right now—and what might disappear entirely. If you’re in SEO, content, or digital marketing, this is a must-listen episode to stay ahead of the curve.
• 🤖 Why AI is creating fear, burnout, and uncertainty across the SEO industry
• 📉 The rise of zero-click search and what it means for website traffic
• 💡 Why conversions matter more than traffic in the AI search era
• 🧠 How SOPs and automation are reshaping SEO workflows
• ⚙️ What edge SEO is and how it enables faster implementation
• 📊 The shift from top-of-funnel content to product-led SEO strategies
• 🧩 How Ahrefs is using AI skills to scale content updates
• 🚀 The real opportunity behind programmatic SEO (and its risks)
• 🧪 Why original insights and expert input are now critical for ranking
• 📣 The growing importance of content distribution beyond your website
• 🔗 How brand mentions and third-party content impact AI visibility
• ⚠️ Why SEO jobs—especially junior roles—may shrink in the future
• 🌍 Why hreflang and international SEO are being ignored by LLMs (for now)
This episode is packed with actionable insights on how to adapt your SEO strategy in an AI-driven world.
If you want to stay competitive in SEO as AI reshapes the landscape, this episode gives you the clarity and direction you need.
👉 Subscribe for more deep dives into SEO, AI, and digital marketing trends that actually matter.
💬 Comment below: Do you think AI will replace SEO jobs—or make them more valuable?
🔗 Tools and Resources Mentioned:
• Follow Patrick Stox on LinkedIn → https://www.linkedin.com/in/patrickstox/
• Google Uses ~40 Canonicalization Signals → https://ahrefs.com/blog/canonicalization/
• The Most Common hreflang Issues Across 374,756 Domains → https://speakerdeck.com/patrickstox/hreflang-study-and-interesting-issues-brighton-seo-2023-patrick-stox
• Ahrefs official site → https://ahrefs.com/
• Will Critchlow episode discussing Edge SEO → https://youtu.be/syE6fe4lIr4?si=M\_6VaSQXydFritu3
Jon Clark (00:00)
Welcome to the Page 2 Podcast where we uncover the strategies, systems and tactical decisions that move brands beyond page two and into real visibility across search and answer engines. For most of the history of SEO, technical work meant long development queues, slow deployments and a constant fight to get even the smallest fixes implemented. But AI and a new wave of automation tools is starting to change that equation. Our guest today is Patrick Stox, product advisor, technical SEO and brand ambassador at Ahrefs
Patrick spends his day working on everything from JavaScript, SEO and product features to large scale data studies, which means he is seeing firsthand how the role of technical SEO is evolving. What we're really unpacking in this episode is how automation is collapsing the gap between ideas and execution. Patrick explains how edge technologies can modify websites before they're even served to users, how AI workflows can automate repetitive SEO tasks, and why technical teams are moving towards systems that can deploy fixes almost instantly.
But there's a deeper shift happening underneath all of this because while automation makes SEO faster and more powerful, it also raises a new question. If one person can now do the work of an entire team, what does the future of SEO jobs actually look like? Patrick has some surprisingly candid thoughts on that, including why experience and creativity may become the most valuable skills in the industry. If you've learned something new today, take a second to subscribe to the Page 2 Podcast. Leave us a rating or review and let us know what resonated. We'd love to hear your thoughts.
All right, here we go.
Jon Clark (01:33)
Welcome to episode 112 of the Page 2 Podcast. I'm your host, Jon Clark, and I'm joined as always by my partner at Moving Traffic Media, Joe DeVita. Today we're joined by someone I've had the pleasure of hearing speak numerous times and occasionally been able to speak face-to-face in the exhibit halls, but never had the chance to really sit down and pick his brain like we're going to do today. So personally, this one's pretty exciting for me.
Joe (01:42)
Hi.
Jon Clark (01:56)
He's the current product advisor, technical SEO and brand ambassador at Ahrefs Joe and I are very excited to dig into everything that that entails. Patrick Stox, welcome to the show.
Patrick Stox (02:05)
Hey, thanks for having me guys.
Joe (02:07)
We are trying to figure out what you do every day. It seems like you're so busy. You're a prolific writer, you're always at these know you are a moderator for a huge Reddit community, you have these meetup organizations, you organize all these meetups. Can you just talk to us about what a normal day or week looks like for you?
Patrick Stox (02:30)
Yeah, it's sort of varied. I actually do the technical SEO for ahrefs.com. We run on a lot of JavaScript, different sections of the site, different frameworks, different setups. Pretty much all the programmatic projects that we have, which not just Ahrefs, but other websites.
What you said, that's writing, speaking, data studies. And then I actually do work on the a lot of product ideation, new features, filters, presets, all that kind of stuff. Sometimes prototyping, more of that now. And lately, it's just pretty much all skills.
Joe (03:03)
Are you working like a 60 hour work week? How are you doing it all? How do you balance the work with Ahrefs and everything else?
Patrick Stox (03:11)
I mean, everything else really isn't that much. It's always having help. There's other moderators for TechSEO, our local meetups, there's four of Conference, there's four of us also. It takes a community to run events. There's everything from door greeter to setting up chairs, AV. It's more than one person should be doing typically.
Joe (03:31)
Okay.
Jon Clark (03:32)
It really does take a village with things like that. I mean, you guys are incredibly nimble team too, so I can imagine everyone just sort of pitches in pitches in where they need to. I guess we can dive right into the AI side of things. I thought it was interesting just in doing some of the research that back in your days with IBM, you were sort of already starting to explore, I guess, AI in its infancy. And now that you also sort of manage all these communities that Joe mentioned.
I'd love to hear maybe how you're seeing that conversation sort of just quickly evolve. Like, what's the biggest maybe shifting conversation that you've been hearing and being involved across these communities? I feel like there's definitely sort of the peaks and valleys with the more that we learn about this and new sort of buzzwords coming out. But is there anything sort of you can put your finger on in terms of something that specifically is changing more recently?
Patrick Stox (04:23)
Yeah, fear, I think. It used to be we were automating things like I wrote a script for redirect automation like 2015, 2016, something like that. It was the first one out And it was to help people like to make their jobs easier. Now, with skills and agents and everything, I think everyone's like, am I going to have a job in a year or two, five years?
Jon Clark (04:25)
Haha
Mm-hmm.
Patrick Stox (04:49)
So I think, yeah, fear, uncertainty, doubt, it's all that right now. I think there's a lot of, you know, a lot of change. It's causing a lot of burnout for people. So I'm hearing more on like the mental health side than has typically been brought up in the past. You know, tech SEOs in general got to work on cool stuff. Most of them were fine, I think, from that side. And now it's like, there are more tools, more systems, more things to do,
and all that doubt about like, am I automating my job away?
Jon Clark (05:16)
Right, right. I mean, I guess maybe with the initial release of all these tools, it was sort of fascination. you sort of had the fear snowballed, then it got sort of exciting because you were able to build all these things on the fly. And now you can really start to see how what you're building can replace your job. Is that sort of the fear path that you're sort of experiencing?
Patrick Stox (05:36)
Yeah, I mean, for like devs, content writers, I think for a while they were like, it's not good enough. And it really wasn't. I would say at least from the dev side, like there was a massive change around December last year. And now like I can like one-shot stuff in apps, like it's auto correcting issues with all the planning features and everything. Like that stuff is scary now.
It's so much better writing than it was. And now with skills, like you can break the writing down into different processes. It's something we do. And we built like, I don't know, 15 ish skills just for the Ahrefs blog to like go in and do updates, changes, different workflows, basically, because it's all like the skills for people that, I don't know, are basically like SOPs. Um, you know, you're, you're creating a process document and you're telling it what to go do. So if you got something repetitive that you do like that,
Jon Clark (06:20)
Mm-hmm.
Patrick Stox (06:27)
It can pretty much be automated now.
Jon Clark (06:29)
I mean, we've always taught the benefits of SOPs, both in terms of scaling and onboarding, but also just for process and efficiency in just the agency ecosystem in general. And I feel like the agencies who have a deep library of those types of documents are a little bit ahead of the game just to begin with, because they can literally take that SOP and upload it into a skills memory or use it to sort of define the
prompts that go into that So yeah, I think you're exactly We were just sort of demoing something that we've been playing around with where it's basically a set of 12 to 14 specific skills around updating content. And I think I built it in like six hours and stopped like perfect, but like, you know, even six months ago, I would have never been able to do that myself. There's really incredible the speed at which this is going.
Maybe we can talk a little bit about the data that you're seeing. There's maybe this fear from the employee side of things, but how is that translating into the data? You guys are releasing incredible studies, data-rich studies around all the data that you're collecting. How is that translating into what you're seeing in terms of traffic? Like, is that fear translating down to the website level? I think we all know the answer, but we'd love to hear
maybe some of the data points that you guys are looking at.
Patrick Stox (07:44)
Yeah, I mean, I think with zero click rise and everything, everyone's traffic is going down. But that's not necessarily a bad thing. Really, companies should always just be looking at conversions, revenue, that kind of thing anyway. In general, the conversion rates are higher from AI search. So even though less traffic, you might still be getting nearly the same amount of conversions.
Hopefully at least I wish I had better data on that. I think for the most part, everyone is somewhat okay there, but I think it will translate a little less. Like, to give an example, our blog traffic. Ahref's blog traffic is way down. And yeah, we probably shouldn't care about like top of funnel informational queries, but we do because our entire blog strategy was take people from, you know, problem aware, like what is this information?
I got this thing I need to fix to solution aware. Like everything how to solve this with Ahrefs, here's the process. So for us, that's not great. A lot of businesses, you know, rely on ads and stuff for their revenue, not necessarily just conversions and revenue like a SaaS would. They are, how do I put that nicely? No, they're, they're kind of screwed. I won't put that nicely.
I think they're going to have trade-offs of, I just block these systems totally? We saw, what was it, the recipe sites where these amalgamations, nonsense recipes aggregated. Now they're trying to make it nicer and be like, OK, this is actually the recipe from this website,
Jon Clark (09:08)
Mm-hmm.
Patrick Stox (09:17)
but still they're showing the recipe from that site and then like, click over there. But those recipe sites, like they're making money through affiliate you know, by this tool I'm showing and ads and now, now I don't know what they're going to do. I think eventually they have to make that trade off of you know, do I leave this public or do I take this to a private community walled garden?
just saw it with, I don't know if y'all saw this, rtings.com. They do a lot ratings and reviews for electronic devices. It's R-T-I-N-G-S, skip the A. And yeah, they basically just took it where now walled garden approach. And I think we're gonna see more of that.
Jon Clark (09:51)
Ha
I the sites like the New York Times and maybe others who've been investing in that subscription model for years leading up to this are probably in a good spot from that perspective. It's a tough decision that these sites are going to have to make to stay solvent, really.
Patrick Stox (10:12)
Yeah, I wanna, I don't know that that's even gonna work though. Cause like New York times, content syndication goes out there. There are scrapers. A lot of these AI systems site sources that would like never be cited in traditional search. They don't have the reputation. They're kind of spammy. They're still being cited in those systems, at least for now. So even if they choose to go that path, someone who stole their content may actually show instead and be giving away
all the data.
Jon Clark (10:42)
Yeah, I mean, that's been an always problem for syndication, right? Like, how do we manage for what we want the canonical version to be versus some of the sites that maybe have more strong domain authority and sort of taking that top spot or taking that authority away from the site? I know Yahoo syndication was a big problem for a while, where they would basically outrank anyone who syndicated content to them. So yeah, it's...
old problems are new again, right?
Joe (11:08)
Can we maybe
just talk a little bit more about maybe the changing strategy you have for the blog? It seems like you've always had this two-pronged approach where you focused maybe more than you should have on top of funnel content and less than you should have on bottom-of-the-funnel solution content. I imagine that strategy has to shift a little bit for you now.
I mean, you guys are a solution company. Do you just try to go a little bit heavy? Do you abandon the top of funnel for a while to fill out your bottom of funnel content? Is that the approach for 2026?
Patrick Stox (11:41)
I think that was the approach for like 2025 even. Yeah, a lot of the informational content we used to write, you're not seeing as much of that on the blog anymore. You're seeing a lot more data studies, tests, that kind of thing. now outside the blog, yeah, our product marketing team and all, like we're, you know, comparison pages and,
you know, more, more on the product pages than we've ever done before. Like we, we didn't even have pages for a lot of the stuff, a lot of the solutions we probably should do like industry pages, use cases, kind of things. there's a lot of strategy shift, like you said, and it's because yeah, if everyone's just getting the information they need just by asking, you know, well, even in Google between AI overviews, AI mode, if, someone even goes to Gemini or
ChatGPT, Claude, wherever you're searching, like the answers are just pretty much there now and rarely people need to dig in further. Some stuff they do though, cause, like, some things are more complicated than you're like, you know, three paragraph thing is going to answer.
Jon Clark (12:43)
I did want to dig into maybe a relatively newer feature of Ahrefs, sort of around like the edge SEO concept. I think it's maybe a newer concept for a lot of people. We had Will Critchlow on the show a couple episodes ago and he really sort of preached the benefits of how they've set up their A-B testing ⁓ capability where it sort of sits on the edge.
You're able to roll these things out without necessarily a ton of technical, really like bandwidth, because you can all sort of do it on the edge, if you will. Can you maybe to start, can you sort of explain what that is, like the concept behind it, how it's implemented, and maybe some of the benefits there?
Patrick Stox (13:23)
Yeah, the easiest explanation is that it changes your website before it's ever served to bots or users. And that can be like a middleware system, which like sits between your server and CDN, or it can be at the CDN level. So you've got things like Cloudflare workers, or workers, everyone has workers now. But Cloudflare was like the first ones that made it really available to people. I think Akamai might've been the first that actually had these.
And you can just do anything. You can add content, remove content, fix your canonical tags. You can implement redirects. I think redirects were one of the first use cases that were done with this. And that's powerful. it's just whatever you wanna fix just it. You know, it's funny because I love this. This is like my dream project. And,
we started building this out a few years back and now we've got tons of stuff like you can use AI to, you know, automate, automatically write 800 meta descriptions and implement them with the push of a button. So titles, meta descriptions, canonical tags, meta robots. I think we just launched alt text generation and implementation. All these things are like,
basically done for you. And now I'm like second guessing, like, is that even the right way to go in this modern age? Cause now it's like, well, what if I you know, let's, let's take a WordPress instance. For example, I could literally use their API and go make those changes on the page level. I, if it's a template thing, I could have an agent go in and actually change it. And now it's done in the actual system rather than, you know, through a worker on the edge, but
the edge workers just kind of blend your systems together. Like there's a concept of serverless in-dev technology and like basically, I mean, it's someone's server somewhere, but it's a CDN typically, like the whole website's hosted on the CDN in that case. And it just like blends the tech together, like, yeah, your server and the CDN, it doesn't really matter where the thing is. Like it matters what
Jon Clark (15:01)
Mm-hmm.
Patrick Stox (15:17)
the final output of that is. And that's why that technology is great, but at the same time, I think most people just want it changed in one system now.
Jon Clark (15:25)
Yeah, which I guess makes sense. I think where the real opportunities are is if it's a really, it's like a homegrown CMS or, you know, the SEO capabilities of the CMS are limited. I think in one, one of the, one of the podcasts you referenced an example from, from IBM where it would take like six hours to change. or sorry, maybe it six weeks to change like a canonical tag or something, which is just absurd.
And for the cost to implement it on the Ahrefs side, I think that trade-off is probably easy to push through from a finance perspective. I think one thing I was curious about is if you have sort of the, or I guess maybe some questions that we get from clients when we suggest it, is can you take us through like your decision tree of implementing on the JavaScript snippet level or the Cloudflare worker level? Is it a limitation on
what sort of workers you have available to you? or is there a specific decision based on the type of website you have? Does that make sense?
Patrick Stox (16:18)
I think the decision really is just like what's available to you. Like we only support Cloudflare workers right now. If you're on Cloudflare, great. If you're on Akamai, you're out of luck on the HR system. And in which case you could use the JavaScript snippets, so not fully out of luck. But the problem is that certain things are not supported that way. ⁓ So if you, with a worker, rewrite,
Jon Clark (16:24)
Mm-hmm.
Patrick Stox (16:40)
⁓ say, a noindex to an index, MetaRobots. That's gonna change before it's served to the bots or users, like I mentioned. That means it actually will be seen as indexed. Now, if you try to do that with JavaScript, they actually see that noindex first. So now they've seen a noindex and an index. And guess what? They're gonna take the most restrictive version. So they're actually still gonna noindex that page. So you can't actually make that work.
going to be, yeah, there's a few things that just will not work with the JavaScript snippet. And like we, I think in the tool, like we don't even allow you to do those things.
Jon Clark (17:13)
Got it, got it. Another thing I was curious about is if you, I don't know, let's say for budgeting reasons or whatever,
they have to remove it or maybe they change CMSs completely and build a lot of that functionality within the CMS so it's no longer needed, whatever the argument might be. What happens when that feature or that worker is removed or that snippet is removed? Basically all those implementations go away as well, correct? Yeah.
Patrick Stox (17:36)
Yeah, that's one
of the risks. And in Ahrefs, if you canceled Ahrefs, you still have the worker, but you can no longer edit it. it would still, the changes that were there would still be there. And then if you start changing other things or try to change something, it may not actually let you without you going into Cloudflare.
Jon Clark (17:45)
I see.
Got it, got it.
Joe (17:56)
How do
you, do you have to, so for like a big regulated company like a finance, someone in finance or something, how do you address the concerns of the, like the development team with all this process in place that this solution is more than just quicker, it's safe too?
Patrick Stox (18:12)
Is anything safe? I mean, ⁓ would I even have a job if everyone did everything perfectly? No, I think the big issue has been like getting it for years was getting it through infrastructure teams DevOps. Anyway, I think will Critchlow you mentioned sells his as an AB testing system that that's a great
Jon Clark (18:14)
Hahaha.
Mm-hmm.
Patrick Stox (18:32)
positioning or at least it was for the time because yeah, you could just sell it as like I'm testing this and while the test is is there it's technically fixed, you know, so that that was great. ⁓ it's a good way, good, good work around to get that done. But now like teams, they want it changed in the main system, but maybe didn't have the resources now again, you know, you could have an agent do it. You could have a worker do it. I think
you know, the way I would position it now, if I was trying to sell this internally in a company is automated fixes, AI fixes AI, everything. Every exec, every C-suite person wants more AI, more, faster deployment, faster everything. We got to move fast. That's, that's the story I keep hearing. And that is the way to sell this. You want this stuff fixed, like whether it's a worker, whether it's an agent, boom, we
do it in 100 the amount of time it took before. Push a button, done.
Jon Clark (19:27)
Yeah, speed plus AI is a great way to position that. Last question about Edge. I was curious around how you're thinking about potentially using it content, right? you guys have a great sort of content auditing tool it of highlights chunks opportunities within the content itself. Is there, I don't know, maybe you can't say, but is there a future version of this where you could go in and
I don't know, sort of marry some of those two concepts together, right? So you could sort of reorder content, things like using an edge worker versus having to go, I don't know, through a much more cumbersome content fixing process.
Patrick Stox (20:07)
and again, whether it's on the edge or whether it's a direct write to, you know, an API in one of the systems either way is fine. And yeah, it's something that's like on our roadmap has been for a while where like we, you know, if you, it should be easy. We pull out the content you have, you change it, we put it back.
And again, whether that means edge or API or whatever, I don't care personally, like as long as it gets there. And I don't think people really care. Like as long as they can make the changes and it goes back and the changes are there, then I think people will be happy.
Jon Clark (20:43)
Got it, got it. I was curious to get your opinion on, speaking of content, on like programmatic content. I think with the original launch of AI, right, there was just a smattering of all sorts of programmatic projects. I know Ahrefs has, I guess, dabbled in that a little bit. How are you thinking about programmatic content today
knowing that we're seeing some algorithm updates rolled out that may be penalizing certain types of AI generated content or focused content maybe? You still see a play there from an SEO and growth perspective.
Patrick Stox (21:14)
Yeah, that is what I am most excited about for this year, in fact. We did programmatic projects that people didn't even realize were programmatic. And they're doing great, phenomenal. Literally tens of millions of visitors every month that they are. For instance, I'll give this away for the first time. I don't think I've actually told people this, but
all of the AI writing tools that we launched as free tools for they were basically just ChatGPT wrappers. We even asked the AI to like tell us how to use the tool and blah, blah, blah. Like here's what it's for, here's how to use it. I think that was like 70 or a hundred tools and those are, I'd have to look, I think like close to 20 million visits a month themselves.
And we use that to then pitch like AI content helper and everything. We were pitching wordcount.com with it, which was another programmatic project that we have. And yeah, they are doing phenomenally. I would say if you're doing it for things that don't necessarily make sense for your business,
you're at risk if you're doing it in a way that's not really useful or in a way that is just, it is just the AI. So a lot of ours, like we did top websites. We had plans for a bunch of stuff that hopefully I will be doing this year. And like a lot of it is to showcase the Ahrefs data we have. Here's how many people search this. Here's like creative uses of the data. Here's like data about websites.
So all that is relevant and I'm not too worried about those. But if we start, you know, automating general informational content for our blog, maybe not a great use case. If I wanna do some in the glossary, I think that actually is a great use case, cause that's pretty general content. So long story short there, we actually did write a lot of our glossary content ourselves.
And then at a certain point, the volume was so low that we're like, we should just generate these pages. Still useful to have, but you're talking, you know, this is a few paragraphs. The thing that AI is really good at doing is summarizing and making things simple. So I think those were valuable. They worked. So it's just, there is an art to it where you can overdo it. You can have the things you need
Jon Clark (23:11)
got it, yeah.
Patrick Stox (23:29)
in there you can have things unrelated and it's going to be a mess. Like we actually have some like content generation tools coming where we are trying to guide people better. You know, we might have, just some ideas. I don't know what all we will end up doing, but like we could have a Slack bot that just goes and asks five people at the company, the people that work on the thing,
you know, a few questions, relevant questions and get their experience, expertise, their insights. Then it's not, we're just putting out the AI content. We're actually getting the content from experts that then we can mix in. And that could be even ⁓ an automated video call where the thing just asks or in any number of ways on automated phone call. If you got some old school people that just don't want to use a computer, there's a number of ways you can get the info out. And it's not that different than
what content writers should have been doing anyway. Like go interview the expert. Years ago, I used to have a tape recorder. I would take into engineers, working with like a radio tower company. And I'm like, I don't know anything about that. So like y'all tell me and I would actually record it at the time so that I could transcribe it and then like, you know, figure out what to put in the content, what was actually relevant. And so now with AI, like all that can be
scaled more than ever, you know, you could, you could, we've talked about that number of different things, but like gamifying it, like we have, you know, something for security in Slack and it asks people, but again, like if we just did five, 10 people or whatever, the relevant people at the company, given some kind of point system winner gets a hundred dollar gift card every week kind of thing. Like, I don't know, I don't know the right answer for that yet, but I do know that just putting out the content.
At scale is risky so many sites, you know the rank and tank. You see them go up, you see them go down. Usually at a two to three month cadence It will work for a little bit and then they go down and it can take down other parts of the website so that's that's part of the risk of putting out so much is like not just that content, but like all your content could potentially be pushed down.
But even some of the ones that tank, a lot of SEOs are like, oh, that's a huge loss or whatever. They went from like 100,000 up to like 6 million and then down to like 2.5 million. So they went from 100,000 to 2.5 million. That's a win. What are you talking about? They're doing great. That was a success, not a failure.
Jon Clark (25:46)
Right, yeah.
So man, so much in there. So the Slackbot example is an awesome use of AI. So you would get those responses back, and then you would then feed those responses into another AI
Patrick Stox (26:06)
Into the content generator, basically. Or if it's something that is a skill to improve the content, like, I'm not saying like, don't go ahead and launch all these pages, but don't lead them like that. Don't, you know, traditionally, I think a lot of folks put out content and then they never update it. Or they updated like two years from now, three years from now. With the programmatic, like, I treat them the exact same way I do a product.
Jon Clark (26:08)
Got it. And then that would sort of craft the content.
Mm-hmm.
Right.
Patrick Stox (26:30)
Cool, like let's go get some expert insights, so let's put that in, let's put Ahrefs data in here. Let's add this feature over here. Let's add this section here, like, keep improving it. The problem isn't that like it's generated by AI. The problem is it's like a base level content that it's, it's not quality and you could make that content better.
Joe (26:51)
I was hoping you were gonna get to that. It's like, you're just looking for better ways to lift up original ideas. It's like, you still have to come up with something that hasn't been said before. That's the hard part. Then you've got the packaging it and the promoting to distribute it is something that you can put a system around. Just coming up with that original idea is the hard part, the original thought.
Patrick Stox (26:58)
Mm-hmm.
Or just your own insights, your own experience. I I would say like even when like I'm writing a blog for Ahrefs, an informational blog or something, 80, 90 % of the info is just the same stuff that's already out there. It's general knowledge. That is what AI is good at. A lot of our content that SEO has created over the years basically just met that bar. We're writing about tires and glue and stuff that we don't know about.
So a lot of a lot of content writers a lot of SEOs created that kind of content just basically said the same things other people that was that was the bar but now that has to be the floor, like, you can't just do that because these AI systems that is what they are good at, that extra, you know, 10 15 percent is your own knowledge, your own expertise, your own insights,
your own tests you've run, like talk about that kind of stuff. And again, if it's not you, find the expert, find that person who their entire job is testing glue and seeing how long it dry, takes to dry and like what kind of conditions in the room and that kind of stuff that you might not find on other sites. That is what is going to allow that kind of content to win.
Jon Clark (28:23)
Relatively easy to do too. It's just like, you it's not like you need some, you know, an insanely expensive tool to measure the temperature in the room. You know what I mean? You just sort of set the temperature, you have a thermometer and you just repeat the process a couple of times. Granted, that's maybe harder to automate from an AI perspective, because you actually have to watch glue dry in that example. But you could be done without, you know, a heavy investment.
Patrick Stox (28:45)
you
Jon Clark (28:50)
It just takes that extra effort. I think, you sort of, you, you've had these, I don't know, evolutions in SEO where things were super easy and then you needed that extra effort to be able to compete. And now AI is sort of coming back to the stage of like everything's super easy, but now we're finding that it is still going to need that extra effort. Even if it is another component of AI, you still sort of need that extra effort to be able to make,
you know, content or whatever it is, stand out compared to everything else. That's sort of the baseline. You know what I mean?
Patrick Stox (29:19)
you go back what couple years now, the, helpful content update for the affiliates killed so many affiliates, but it was a lot of affiliates that were killed off were SEOs who don't really know anything about the niche. And so they were creating, you know, the, the base-level content. A lot of affiliates are still out there and they're killing it. And, you know, in a lot of cases, these folks are, you know, sending off
food samples to laboratories to be tested or like running elaborate tests on whatever setting different conditions for robot vacuum, I don't know if you all watch vacuum wars, I love that channel. Like it's just random stuff and like the folks who are the hardcore ones like that still won. And it's because like they know the niche, they know the product, they're doing the experiments, they're doing the testing,
Jon Clark (29:57)
Yeah.
Patrick Stox (30:09)
they have really cool studies, stories to share. And yeah, they will always be around. But the low effort, affiliate sites are no longer really viable.
Jon Clark (30:19)
So affiliate the site, are well known for some cases, being a pretty spammy neighborhood. Are you seeing any of that in LLM-created content? I don't know if you want to share specific examples to feed the beast, if you will, but are you seeing those types of things content that you're analyzing?
Patrick Stox (30:39)
Yeah, even with our crawler, the number of domains being registered, new domains, the number of pages being created. The web has gone exponential. Like it was already pretty exponential and now it's like straight line up over the last couple of years. And that will probably get worse. Even ourselves, like we, to test a bunch of stuff, basically automated an entire like content creation workflow last year, set up a bunch of
test domains, I can't tell you much about that yet. We'll probably write that up. But just to kind of see like, are they gonna work? What's gonna happen? ⁓ How do we guide people to not crash if they crash, basically? So we've been thinking through this for a while. And yeah, we'll see.
Jon Clark (31:24)
Ha
Joe (31:24)
Can I shift gears a little bit? Once you've got this great piece of content
and you start to see some initial success traffic from your site, you'll need to distribute moving beyond the page that you own to help with that distribution. Are there a set of rules or a set of steps that you could offer? Just some advice to someone, like you've got this great piece of content, now you've got to go beyond your website to really see the benefit from that original thought. Are there a few things you would suggest to focus on first?
Patrick Stox (31:56)
Yeah, I mean distribution other channels like you've got you you created something written, create a video for it, YouTube works well for both. Well YouTube still the second largest search engine, also cited in Google quite often also cited in LLMs. So I would have that if you've got a social media following like share it out.
You've got all your typical like outreach kind of stuff to be active in communities, Reddit forums, Slack groups. There's so many SEO Slack groups. Like now, usually if I post a data study, I'll go to like five different Slack groups and share it there. Cause then I know like a lot of people will see it there and then go share it on social media and like just try to get that attraction that way. But in general, like I'm pretty low effort on that kind of
thing. Y'all probably know way more about that than I do.
Jon Clark (32:44)
I was curious, like how, so you guys recently integrated, I think it was YouTube, Tik Tok, some of these other platforms, Reddit, I think as well, into the Brand Radar sort of ecosystem. Do you have any good examples of how folks are using that tool to improve out of the either citations or AI visibility? Have you seen any good case studies from that yet? I know those features are relatively new. Anything there?
Patrick Stox (33:09)
So Brand Radar was built a little differently than most. Y'all probably know that before we had even custom prompt tracking, like we had big databases of stuff so that people could go look. And a lot of it is just, are people talking about my brand? What are they saying? Are they saying good things? Are they saying bad things? Are they mentioning my competitors and not me?
So like you can go in there and see all that. I recorded a video the other day where I was actually going through Brand Radar for a bunch of different use cases, and like, some of the things I saw were, you know, like SEMrush owns other websites and they're putting out all this comparison content on like backlink, traffic think tank, exploding topics. And like, they were sort of controlling some of that narrative of like the best SEO tool kind of thing.
Jon Clark (33:48)
Mm-hmm.
Patrick Stox (33:52)
But yeah, if you want to check out my social, I put that on like Twitter and LinkedIn and like that was just basically me looking at Ahrefs in Brand Radar to say like, these are things that seem to be working, the things other people are doing, places we need to be, you know, websites that were had like reviews, comparisons, that we probably need a relationship with. We had an internal conversation, do we need to bring back, the affiliate program even.
Like we killed that off many years ago, but I'm like, you know what? These people are making money, recommending competing products. So like, they're being cited in, these systems. And like, if we really want to be in contention there, we probably have to come off money one way or another, whether that's they are affiliated for us or we pay them. I don't, otherwise we just accept that we are, you know, second place in a lot of those listings. And if those are the ones that are,
Jon Clark (34:18)
⁓
Patrick Stox (34:44)
constantly being cited, I want to be first. I won't, cause when it comes out the other end, when these AI search systems, they go and look and they're like, oh, Semrush first, Ahrefs second. I guess what they say, Semrush first, Ahrefs second. Cause like, you know, 80 % of the things are like Semrush first, Ahrefs second. And so 80 % chance that's the order it's going to come out. And yeah, it's a, you know, all the link builders are from SEO are in a good place cause
man, you're gonna need so much outreach in this new era. I feel like third party sites, influencers, communities, all way more important than in the past. And it's far beyond your website now. It's like what the web says about you. How does the internet talk about you?
Jon Clark (35:29)
Yeah, I thought previsibles acquisition of internet marketing ninjas was a pretty smart one. I mean, they basically tacked on that whole outreach capability, you know, that's been around for years. Like they definitely have a good process there.
Joe (35:41)
The outreach that someone focused on back linking has always done is still super important. The writing of original content, still super important. Understanding technical ways to optimize your website, still super important. Seems like the future of SEO, you seem a little bit scared about it, but I feel like it's, this is like a peak time for us where we should be like...
we're going to become more important than ever. I've read where you are fearful that the number of jobs will be constricted because of all the automation put in place. And I really like the point that you make about someone just getting started in SEO. So maybe we could get you to comment on that too. If you felt like SEO was a good place to start your career or not now.
Patrick Stox (36:26)
Yeah, it's much harder. It's hard in every industry. Even the junior devs, do you need junior devs anymore? Your product manager, your... folks like me, anyone can go build tools now. But then, yeah, it's difficult to say that will disappear because then who will actually clean up that code as a senior dev and stuff?
Who will do the SEO work? I do think, you know, with a lot of the automation, you go from teams of 10 to like teams of one, or, you know, one person who did one thing maybe now has like five different roles that they take in in a company. to me, it is a scary time, it is risky. I do think jobs will go away, I do think the junior end will dry up.
But I hope I'm
Jon Clark (37:15)
I think we do have a couple of questions around maybe career, but before we get there, I did want to ask one of the presentations I saw you deliver was around, hreflang tags, It's one of my favorite, I don't know, areas of SEO, because I think it is so complex, so few people understand it well, and when you can explain it well to a client and it sort of clicks and then they implement it and see the benefits of it,
It feels much more fulfilling than updating a title tag for But I think one of the original stats was something like, you know, 67 % of domains had an hreflang mapping issue or some sort of hreflang issue. And I was really curious around,
I guess, A, are you going to update that study anytime soon? And B, are you seeing that translate to LLM types of traffic? Because I feel like we haven't, you know, there was this period where there was a lot of conversation around international and hreflang in general, and that sort of has gone away with all the AI stuff. But it's still a problem. Like we're working with a couple of clients right now that are trying to implement it well.
Really long-winded way of asking, how are you seeing hreflang impact LLM results or maybe you're not at all.
Patrick Stox (38:21)
Yeah, not at all. They're not using it. In fact, I think in a lot of cases, like even if you're asking in another language, a lot of times LLMs the output will actually be in English. And guess what? English sites are going to get cited in that case. So these systems are still new. Like, they're not doing a lot of the things that search engines over the years have figured out. Like, still not crawling JavaScript for instance.
Jon Clark (38:44)
Right.
Patrick Stox (38:45)
Yeah with hreflang they I don't think they're swapping it at all like they have no system for that swap and re-ranking like in search I don't really you know that they have an opportunity for that, they would just try to get the right one for the right language but obviously that's not happening and that's because a large portion it's not always happening like sometimes it will respond in the right language other times it'll respond in English because again, there's probabilities involved
and a large portion of the training data was in English. But yeah, that is like a pet peeve of mine right now too. Almost all these like LLM or AI search systems are telling people like you can use BCP 47. That's like the three letter hreflang system. It's not actually supported. I had to like go to John Mueller cause like we were getting, you know, like three, four people a day on our support chat that were just like,
but all these are telling me like I can use this and I'm like, you can't like look, read the, here's the Google documentation. It says the ISO blah, blah. Like the two letter code. And I was like, Jon, can I get something on the record? Has anything changed here? So now like our support team has a screenshot from, from that, that they can share basically saying like, no, it doesn't actually work. You can't use it for this. And I'm like, I'm so tired of these systems telling people they can, they're telling them wrong.
Jon Clark (39:49)
You
Well, one thing I had never actually thought about until right now is
if the LLM model is not pulling from their training data. So for example, like, I don't know, what's the freezing, what, what degree is freezing, right? Like that's well known. They just pull it out and they're going out and doing a search and sort of, I don't know, reordering those results based on their own criteria. Are they only searching from the U S meaning if they get query in, I don't know,
Japanese, are they going to Google.jp or are they only searching from the US? Maybe you don't know, but I'd never thought about this before.
Patrick Stox (40:33)
⁓ It wouldn't matter,
Google actually changed their system. So, like, doesn't matter which version of Google you're on. They'll try to match it based on the language instead. Like, you used to have to go to Google.jp, but now on google.com you search Japanese, you're going to get Japanese results.
Jon Clark (40:40)
Got it.
Got it, got it, okay. So they're just not doing a good job of moving it over to the international variant of the URL, clearly.
Patrick Stox (40:54)
it's, it's, think even in Google, like we, we are seeing that in like AI overviews, AI mode, like they just don't have the same systems in their newer products that they have in traditional search to actually swap that. So in some cases, like you'll do a search and you'll get like five citations of the same website that are just all different countries.
Jon Clark (41:14)
Oh my gosh. Man. Well, I know we're bumping up against time. ⁓ Should we jump to rapid fire?
Patrick Stox (41:19)
Fine with me.
Jon Clark (41:20)
Let's do it. All right. This is always a fun topic. Subdomain or subfolder. What is your final word here in 2026? Nice. I liked your answer. I forget which podcast it was, but I thought the explanation was well, just interlink it well, treat it like you would a normal website. You should be okay. What do think the most underrated SEO skill is in sort of this AI world that we live in today?
Patrick Stox (41:27)
Either is fine.
I don't even know that it's an SEO skill, but like I in general just want creative people, people that want to solve problems because yeah, anyone can kind of cross into different skill sets. Now, like anyone can go build tools. Anyone can build skills to automate process. They just, just need to know. So I would say two things are important experience and creativity because like you, need to know like when
it's a bad output when something is wrong in the process. So like you, you have to be able to like build these processes if you're going to try and automate stuff and cross lines. And like, you have to be able to tell like when something's lying to you, like when, when I mentioned the AI searches, like use BCP 47.
Joe (42:27)
What's the strangest SEO bug you've uncovered that turned into a big lesson for you?
Jon Clark (42:28)
Right.
Patrick Stox (42:34)
things are like randomly they're like like zero width spaces and stuff and like robots.txt. I don't know, that I learned that much from those but um It's interesting to see how often that actually does occur on the web like we get stuff for it all the time and it's like why are you still crawling my site when, you know, I've got this blocked and wrote I've got you blocked and robots.txt and there's like random hidden character
in the robots.txt that makes that entire line invalid.
Jon Clark (43:02)
Yeah, those are fun to try to diagnose. What's the tool, or maybe it's a feature in a trust that you're most proud of?
Patrick Stox (43:09)
Oh God, there's so many. I think AI Content Helper actually. I really, like we built that in a time before, you know, chunking was a known topic, but like it was basically built for that from the start. Like we could see this coming and funny enough, it was built like years ahead. It's built for what SEOs
Jon Clark (43:20)
Yeah, right.
Patrick Stox (43:30)
need, really needed at the time, but like didn't know that they needed and, and it holds up so well in the modern era. We added like the highlighting and stuff for the different chunks, which is basically your, your topics and everything. And the scoring in that is pretty similar to the scoring system we actually use in the search engine. Like that's how we were thinking about it when we built it is like, what is a good answer? How do we measure that? How do we define that? How can we show people that?
Jon Clark (43:56)
Got it, got it. ⁓
Joe (43:57)
What's a
tough question you got at a conference that informed maybe how you think differently about the tool at Ahrefs?
Patrick Stox (44:06)
I'm just going to not answer that one, I'm sorry. I don't, I don't know. No, people don't usually ask me the tool necessarily. The conferences is usually about whatever I'm talking about, which I'm not super self-promotional in talks.
Jon Clark (44:08)
Huh.
Joe (44:08)
Wow, tough question.
Tell us a tough question you got.
Jon Clark (44:19)
Got it. What's a Google patent, maybe a general document that you think every SEO should read?
Patrick Stox (44:25)
The webmaster guidelines. I think so many SEOs don't even know they exist, They've never actually read them or checked to see if updates have been made. Like one of the first places I sent people was like, go to the Ahrefs Starter Guide, Google Starter Guide, read the rest of their documentation.
Jon Clark (44:27)
Good.
Yeah, I think a great AI use case is just to set up something that monitors for any change across those, basically across that folder or that particular area that you're most interested in, just to stay ahead of what's being changed there. ⁓
Patrick Stox (44:54)
I think we have a project
even in Ahrefs like we are always on audit and it's just like constant and it has a feed of like website changes.
Jon Clark (44:59)
yeah.
⁓ That's smart. I'm going to write that down. right, last question. Do you have a sense for when Ahrefs will compensate for the number equals issue that Google broke for basically every tool? Are you guys thinking through a plan for that, or do you have a rough timeline of when that'll be available?
Patrick Stox (45:21)
So we had brought it back for all of rank tracker and at one point in site explore, I think for everything above, trying to think what threshold we got to 180, 280. The problem is like, we are constantly battling for this. I don't know if y'all saw like Google even hired someone, an anti-scraping engineer. This is tough at our scale.
Jon Clark (45:41)
Right.
Patrick Stox (45:43)
it is literally a constant battle. Literally, I think last week, two weeks ago, they changed something again. I think for some results now, the most we can get is like 30, brutal honesty. I don't think anyone will have a hundred after this year.
Jon Clark (45:57)
What do think the cap will be?
Patrick Stox (45:58)
Right now it looks like 30. But yeah, we see so many tests, so many things they're doing every, like every week there's something else that changes, something that some other way they're combating us. There's probably 35, 40 different identified methods internally that they are for blocking in various ways. yeah, at our scale, I just don't know.
Jon Clark (45:58)
Like 20? 30? Got it.
Patrick Stox (46:25)
We are doing our best. I promise you there is an entire team. Our SERPs team is literally fighting this battle bring it back. But I just, I don't know that we can.
Jon Clark (46:35)
I know it's definitely a complex issue. I mean, for everybody, even on the Google side, right? There's costs associated with all these tools hitting them and things like that too. So it's not a, you're right. It's not an easy problem to fix.
Patrick Stox (46:48)
I honestly don't think they even care about the SEO tools. Like, they never really had. They went like years without any crackdown on SEO tools. The ones that are scraping, the third-party scrapers who are then providing it to the AI assistants. That is who I am convinced they are combating with this.
Jon Clark (46:51)
It's the LLM tools, right?
Yeah.
Yeah, makes sense.
Joe (47:09)
We should have started with this question so we could have gone deeper on it, Jon, It rapid fire.
Jon Clark (47:12)
I know, I know.
Well, we'll just have to Patrick back again. This has been a real pleasure. Like I said, I hope we do get to do this again, maybe next season. But please let our listeners know where they can find you online and if you're gonna be speaking anywhere next.
Patrick Stox (47:25)
I'm Patrick Stox. You can always find me at the local Triangle SEO Meetup in Raleigh, North Carolina. And next speaking gig, I think is two weeks? Sydney SEO Conference.
Jon Clark (47:38)
Oh, fun. Nice. That'll be a good show. Well, thanks again for joining us on the Page 2 Podcast. And for those listening, if you enjoyed the show, please remember to subscribe, rate and review. We'll see you next week. Bye bye.