Are we at peak AI content? Honestly, that question is almost impossible to answer with certainty (since things change so quickly and measuring it is hard), but Graphite’s update to their highly important study suggests it could be. That got me thinking - 🧵 1/6

Optimising pages for crawling has been the name of the game for SEO for years. We’re going through another key shift here at the moment. The target audience for our activities continues to expand. This Web.dev article (H/T Mark Williams Cook for putting it on my radar), 🧵 1/7

“Search would be an agent manager” for future users trying to complete a task. I don’t know about you, but a kind of “agent manager” which helps people solve tasks is a LONG way from the “search engines” we were raised on. 🧵 1/6

What does a truly optimised AI search experience look like? We’re a long way from that now. We know that LLMs (large language models) with RAG are the main components of AI search. 🧵 1/7

This has probably done the most to further blow apart my assumption about “tracking AI” to date. I'll explain The easiest way to “explain” why AI tracking is worthless/rubbish has been to suggest that the probabilistic nature of the process makes the results too noisy and therefore worthless 🧵 1/7

Most SEOs treat data feeds like someone else's problem. @chris-green.net is coming to Antalya to explain exactly why that's costing you. He's built a reputation for catching the moment your data starts lying to you. Grab your ticket: searchnstuff.co.uk/search-n-stuff-antalya-conference-2026/

More info on how and when Shopify data will be introduced into “agentic storefronts” - just the key points from a recent announcement email: - Shopify Catalog automatically structures and enriches your product data, and will distribute it to AI channels where buyers are actively shopping 🧵 1/6

Packt have just told me there is a limited time offer on Amazon for pre-orders for my forthcoming book, The AI SEO Playbook and asked me to share. I am increasingly intrigued as I write this book and guarantee you will not find the info I am writing on any regurgitating AI blah blah blogs or books.

It was no surprise that Shopify was the first to really offer agentic commerce at scale. Will it work? We need to get a real feel for the conversion rates and ROI of these channels for retailers. 🧵 1/4

@johnmu.com Hey John - we had this weird one, we saw a period where the robots.txt (and other files) were not fetchable. But the response we were getting was "Not Fetched - N/A". Can you help us understand what that means? The docs don't have this error type (support.google.com/webmasters/a...)

What content is the most “AI-friendly”? One of my most-asked AI-search related questions (just behind “can AI 'see' my content?). Research like this from Wix is useful for giving us a window into how we can identify this. 🧵 1/8

I’ve done a lot of thinking about prompt-tracking and measuring AI recently - most of which has made it onto LinkedIn already, so I’ll keep this one brief. Tom has done a great job at distilling the key issues people have when approaching prompt tracking and how to tackle them. 🧵 1/3

How do you say “I don’t know” without people thinking you’re not very good? Great question! During this age of AI search, if you have a detailed conversation and DON’T say “I don’t know” at some point, chances are you’re maybe a bit deluded - OR a genius. Bengu really nails this with this: 🧵 1/5

I enjoyed this one. What happens when we abandon “do-all” processes and instead adopt a series of principles and apply them with our own experience/creativity? Amaka - who I had a blast working with on the 2025 Web Almanac - nails it when she says: “principles, not fixed steps.” 🧵 1/6

Visibility scores have always been a bit hit-or-miss for me. On the face of it, a % score which expresses how visible you are (relative to what's possible), is a useful stat. Whether for “traditional” rank tracking or prompt tracking, it gives us a simple way to show how well we’re doing. 🧵 1/7

What are the forces shaping SEO most profoundly? This is something I spent a lot of time digging into as part of my work on the SEO Chapter of the Web Almanac. buff.ly/fBtlcg7 🧵 1/5

Are AIOs the main reason you’re losing clicks? AI overviews are big and potentially impact clicks - that’s widely known and feared, and some industries are suffering heavily from this already. 🧵 1/7

I’m on a panel on GEO and AI Search at Similarweb on April 28. Expect insights from Ethan Smith @dixonjones.bsky.social @sagibson.co.uk @chris-green.net and Rice Tong. Drinks, networking, and the future of search. 🍻 Free tickets www.tickettailor.com/events/cjeve...

Links in visibility in “traditional” search and “AI search” keep coming thick and fast. Meaning, you can’t safely invest in SEO without impacting GEO/AEO or the other way around. We should be waaaay beyond arguing this by now, but it’s just not going away. 🧵 1/4

Where is the value in AI prompt tracking? Given how costly most AI prompt tracking platforms are, I’m pretty sure most users aren’t seeing a positive ROI from them. Let’s break this down: - Daily/weekly monitoring of overall visibility is so noisy it has the potential to waste time 🧵 1/7

I’ve been talking about spam in AI results a lot recently. Almost certainly because I’m encountering more people offering “GEO services” that are broadly based on ranking well and misinforming AI models. 🧵 1/8

For years, I felt like keyword research often started in the wrong place - an SEO tool. Granted, the tools had the data, and depending on your objective, there is a huge amount of value here. 🧵 1/6

Another HTML vs Markdown showdown for AI consultants to contend with. Do LLMs ever see the full HTML when provided a URL? David has run some tests that provide super-interesting outputs and food for thought for those trying to understand whether their website is “AI-ready.” 🧵 1/4

This won’t be a popular one for all, but it is thought-provoking. Devs and engineers are rightfully resistant to vibe coding, and analysts might not enjoy the thought of being replaced by AI Agents either. But hear me out. Wil’s blog “I have 10 questions a day & only ask you 1. 🧵 1/7

How I keep up to date with SEO news has changed in the last 10+ years The main influence here is how I identify the valuable voices in the industry, and how they choose to circulate is a huge component. The painful metamorphosis of Twitter to X is possibly the biggest single contributor here 🧵 1/4

The world of AI search needs more testing. Okay, perfect testing is hard; causation does not equal correlation; and just because something is happening now doesn’t mean it’ll happen in the future. These are common objections against testing, but they shouldn’t stop. 🧵 1/5

“Is this optimised for AI?” or “Can AI see this?” are two of the most frequent questions I field at the moment. For non-experienced SEOs, this is a tricky question to answer. Even some experienced SEOs aren’t super sure - a lot of information posted is conflicting. 🧵 1/3

A while back, I went down a rabbit hole using the AlsoAsked API I was actually building an AI content workflow, but I had a feeling in the back of my head, “What if answering PAA questions drives better results?” 🧵 1/6

Recently circulated on LI was the regex “^(?:\S+\s+){9,}\S+$”, which can help people identify queries with over 9 words in their Google Search Console data. This is a neat proxy for AI-based queries (AI mode, AI overview, or search grounding). 🧵 1/5

Can an AI Chat Bot “just read an image” when it views a page? No. Just because AI models are “Multi modal” doesn’t mean they’re used in the RAG pipeline process, and I ran a small experiment to test this. 👉 buff.ly/Of7XdVY 🧵 1/2

AI Search (GEO/AEO) has a spam problem. That was inevitable. We’ve seen this pattern before: • A new surface appears • Early tactics exploit it • Scale accelerates • Monetisation pressures build • Algorithm updates recalibrate trust 🧵 1/3

Search in e-commerce is undergoing a huge facelift. Optimising Google product feeds to leverage organic shopping features has been a best practice for a few years. But we’re standing on the precipice of this all changing again: 🧵 1/6

👏 @chris-green.net did an amazing study showing how answering more PAAs can correlate to better rankings, and goes into more depth about what these findings mean on video with @alsoasked.com . Interview: www.youtube.com/watch?v=_9df... Full study: alsoasked.com/insights/paa...

There’s still so much talk of aggressively scaling content 💬 More content equals more success seems to be the simple calculus. And it works… until it really doesn’t. 🧵 1/4

“You Can't Generate What You Can't Retrieve” is a foundational message for AI Search. Retrieval is the bedrock of everything, and Dawn has hit the nail on the head (again). Slides are never a great substitute for being in the room, but this is a slide deck you’ll still learn from. 🧵 1/3

We love tools, and awesome tools are even better 👏 Suganthan has done a great job with this list, with contributions from across the industry. Curating and maintaining this is a huge effort, so I suggest you follow it on GitHub for future reference. Thanks, Suganthan

We’re no longer playing a zero-sum game. In so many ways, this is an opportunity; in others, it’s a massive stress-headache waiting to happen. 🧵 1/7

Just so people get a feeling how much 2MB HTML actually is, I made this (3MB+) monster: experiments.geekonaut.de/3mb-behemoth I know - nobody asked but here we are. You are welcome.

I review more technical audits than I write these days. From the teams I work with, to third-party SEO resources (mainly agencies), there is almost always one observation I make. Where is the prioritisation? OR That prioritisation feels wrong. 🧵 1/9

Whilst this piece is framed as a GEO-myth-spotting safari, I found the most useful part to be the lessons we should take when approaching any particular area. - Challenging biases - Lazy thinking - Being willfully ignorant - Looking for easily-fitting or convenient answers 🧵 1/6

This is pretty big: Pages that fully answer more PAA questions tend to rank higher. This is one of the conclusions of a brilliant independent study completed by @chris-green.net and published on the @alsoasked.com site today. alsoasked.com/insights/paa...

This is great from Harry 👏 We’re starting to see more mature, less BS approaches to SEO/AEO/GEO. Disambiguation is similar to what I’ve been encouraging. Interpretability focuses on how we ensure that we are being understood and represented as we want to be. 🧵 1/2

Will ChatGPT paid ads be a worthwhile investment for brands? Right now, it’s hard to see how we can prove it. Estimated average CPM at $60 (which is super high relative to other paid platforms), and yet there is no clear way to understand conversions generated from those placements. 🧵 1/7

What replaces clicks in a post-click world? If “SEO is dead” and “AEO/GEO is the new SEO”, it’s ironic that Google Impressions are one of the best indicators Don’t expect to hot-swap clicks for impressions without questions. Time to start educating on impressions and why they’re important.

So you’re happy your content works for users, but are you sure it works for AI systems too? Okay, so this is something that I wish we didn’t need to worry about - but we do. It comes worrying close to the “writing for SEO” activity that is responsible for at least some of my grey hairs. 🧵 1/9

AI Search has driven an evolution in your SEO function, not - I’ve argued many times - a replacement of it. But for those who were quicker at launching a new function (net new or an extension of existing SEO), here’s likely what you have seen in the first year: 🧵 1/7

If you want an example of why 2026 is the year of better bot management, this is it! In this specific example, Grok is asked to fetch a URL, and we see 16 requests across 5 different user-agents and 12 IPs. 🧵 1/6

I use AI but… I wouldn’t connect it to my Google Drive I wouldn’t use a browser powered by it I wouldn’t auto-allow Cursor to run any/all commands in terminal I wouldn’t take health advice without question I wouldn’t let my kids access unsupervised I wouldn’t trust references without review 🧵 1/4

Final shout for Issue 1 of This Station Is Non-Operational, a newsletter for human computer users forced to live under the yolk of big tech. buttondown.com/thisstationi...

Been a busy week, so busy I barely clocked this one go out - www.searchenginejournal.com/what-the-lat... Thanks @shelleywalsh.bsky.social for the interview and the cracking write up too!