How hard is it to change everyone’s current search habits? OpenAI’s chances to “beat” Google really depend on how they can effectively change people’s given “search instincts”. For early adopters and young users, this may be easy, but are some users just so entrenched in using Google? 🧵 1/5

A balancing act for the AI-search generation is that of Understanding vs Steering This short piece by Dan is one which overlapped some of my own thinking recently, but I wanted to call attention to it because it presents a relationship which many may only spend time understanding one side of 🧵 1/5

Turns out I’ve been building a few tools recently, it’s a bit of a random collection, but if you work in SEO and find yourself having to report on performance, chances are one of these may be useful to you: 🧵 1/3

OpenAI, Google, and the search data question. Rumours have been circling in SEO for weeks that OpenAI may be tapping into Google data to power its own search (and possibly SonicBerry). 🧵 1/5

I’ve been rebuilding my Metric Chain in the AI Search era. Not out of vanity, but because it's been requested by a few and I am growing increasingly aware that this measurement piece is something that presents a greater issue for SEO as a channel than the scary unknowns of AI search itself. 🧵 1/5

How should you chunk your content for AI? Here’s my take on it…

Ever wanted to try and build a search engine? Maybe, maybe not. But if you want to learn more about how search engines work and don’t have the time to try it all for yourself, Wilson has you covered. It’s hard to really summarise the level of detail in this one, so I’d suggest you read it. 🧵 1/2

If you want to delve deeper into the foundational knowledge of AI Search, Duane is still one of the best to follow in this space. This particularly interesting read, Duane looks at semantic density and semantic overlap with regards to optimising for AI Search. 🧵 1/4

LinkedIn is often plagued by bad data-viz. My biggest (recent) pet peeve? Using a secondary axis and cherry-picking a scale to make a correlation where there isn’t one. Not all of that is intentional; sometimes it is accidental, or just a knowledge gap. 🧵 1/3

How are AI Bots shaping the web? Here’s another data source that is really interesting. Paul has done a deep dive into HTTP Archive’s data - something that I’m super-close to right now as we’re preparing the 2026 Chapter of the Web Almanac! 🧵 1/3

We should use Google Trends data more, as the rest of our click/attribution data will degrade because of AI search. - Google Trends scales each query independently to 0–100. Compare up to 5 terms together, but anything more breaks down - Introducing a workaround - Trend-Stitching (see link) 🧵 1/2

Time to dust off trends data *again* Google Trends isn’t just for curiosity any more, it’s a reality check when our other metrics start to look shaky. This is the biggest challenge AI search brings us. 🧵 1/4

Take to the Skies 09/17/25 www.youtube.com/watch?v=t0GP... Join the adventure! 🔗https://bit.ly/LostSkies #LostSkies #IndieDevs

Does it matter that AI search is fundamentally different to “traditional search?” Yes & no First off, the more you understand about the underlying mechanism and methods, the better. I'm NOT putting people off learning more, and I’m not saying that those who have taken the time have wasted it 🧵 1/5

Does content substance impact rankings? In short, no. Watch the video to find out why.

“SEO is dead/dying/killed by AI” is fast becoming a new meme headline in the tech press. Not for the first time, and certainly not the last! One of the latest in this “series” - buff.ly/QLHzviH 🧵 1/8

In a world of LLMs and infinitely scalable content, companies like Google and OpenAI must develop robust methods to surface the most relevant content Google’s history is deep - PageRank was the first major breakthrough, followed by signals like quality, authority, relevance, and freshness 🧵 1/7

How “substantial” is ranking content? I recently tested Dejan’s Content Substance Classifier - an AI model designed to spot “fluff” (empty writing) vs. “substance” (meaningful content) - on over 500 keywords across multiple niches. The findings were… telling: 🧵 1/4

Will GEO replace SEO? Yes, I’m back with my “it depends” videos, and this is a timely one.

Who’s going to be first to crack “query fan-out”? My money's on Dan and the Dejan team. While the SEO industry gets its head around what this process is and what it means in practice, one of the key “actionable” pieces of info is how we can predict/anticipate this fan-out specifically. 🧵 1/6

As discussed with @panvictor.bsky.social and @chris-green.net before, while you can see the "RAG probability" in ChatGPT network data, it seems the threshold is wildly different on free vs paid (maybe a cost saving thing). Even things like that are interesting outside of model v model.

👀 ChatGPT Is Not Replacing Google, It’s Expanding Search 👇 A research from Luke Harsel from @semrushofficial.bsky.social shows ChatGPT adoption did not reduce Google Search usage. In fact, there was even a slight increase in avg Google Search usage. Check it out www.semrush.com/blog/google-...

So, we’re getting more confident that SEO isn’t dead (again), right? I joke, but it’s becoming clearer that “traditional” search results are important for more than just “traditional” search results (for now), which has calmed quite a few people down. Just to warn you, IT DOESN’T STOP THERE 🧵 1/7

Is ‘SEO is free traffic’ one of the most damaging myths in SEO? Maybe - add it to a long list. What I like about Nick’s blog on the subject is how stark this argument is when you consider that beyond the effort SEO needs, what is more damaging is how the effort is attributed to success. 🧵 1/5

Is ChatGPT getting closer to Google for Search Results? Looks possible. Has it dropped Bing? Quite likely! Josh recently dropped another fascinating piece of data around how often citation sources for ChatGPT results overlap with Google vs Bing’s. 🧵 1/5

Why Splitting SEO and GEO Functions Doesn’t Make Sense 👇 by @chris-green.net - going through how: "Splitting SEO and what some are calling “GEO” is, in most cases, a new solution in search of an old problem. " A must read: www.chris-green.net/post/splitti...

Using AI to complete coding tasks could potentially take 19% longer than not. Vibe coding is dead, right? Some are suggesting that experienced developers/engineers can work more quickly and efficiently using an LLM to “copilot” throughout the process. But this study suggests otherwise. 🧵 1/4

“Google doesn’t owe you sh*t.” Nick LeRoy June 2025. Not a popular view, but one that can be helpful, in some ways, for framing the challenge ahead of us now. A lot of the arguments Nick makes may be familiar, and you may not like them... 🧵 1/2

If generative search is going to eat up the top of the funnel, shouldn't I avoid putting content there? Nope. Although it is easy to see why people might start feeling that way, attribution issues, the change to AIOs, and the soon-to-be AI mode, along with ChatGPT, etc., 🧵 1/4

The Average Position metric in Google Search Console (GSC) is often misunderstood or misused. For many, it's seen as an indicator of SEO success, but without proper context, it can lead to misleading conclusions. By filtering your data intelligently within GSC or platforms like Looker Studio 🧵 1/2

Will MCPs solve all your issues? Are they the solution to your problems? Probably not, at least, not how most are testing them right now. Don't get me wrong, I think MCPs have HUGE potential, and I've been testing some that really do have potential to enable some powerful applications. 🧵 1/5

You may have heard “query fan out” recently. It’s likely to become more ubiquitous as time goes on, as marketers will see the impacts of AI Mode and want to get the better of it. Right now, the secret sauce behind this is only talked about - as far as I’m aware, nobody has cracked it yet. 🧵 1/5

It's surprisingly easy to set yourself up to fail at SEO, despite the best of intentions. Overselling expectations, undervaluing the process, and obsessing over the wrong KPIs - these common pitfalls quickly come to mind. 🧵 1/2

Everything in SEO feels uncomfortable right now. Or at least you'd be forgiven for thinking that - I’m quite enjoying it! Ahrefs have clearly been having fun with the product roadmap and content recently, and this is possibly one of the most "SEOy" things I've seen recently. 🧵 1/4

“Vector Embeddings” discussions nearly always mention “cosine similarity.” The core idea: turn content into vector embeddings and use cosine similarity as a distance metric to quantify semantic relatedness. But Massimiliano highlighted a crucial nuance in a recent LinkedIn post: 🧵 1/5

“My Chat GPT rank tracker isn't accurate” is something I've been hearing, but it shouldn’t be that surprising “Answers from LLMs are based on probability & creative variance (temperature) amongst other things. This means the same question from the same person can receive different answers” 🧵 1/5

Are you "Multimodal ready?" If the answer is "no" or "what is that”, I suggest you take a look at this article. Optimising content/websites is a given for SEOs, but what could our day jobs look like when we start optimising product packaging for image search, or our logos, our stock imagery? 🧵 1/2

Rankings are moving to answer synthesis - it sounds daunting if you've been heavily invested in rankings for the last 5,10,15+ years! Dawn's SMX slides provide a (naturally) detailed and informative account of how we can make this transition. Some of my notes/thoughts from it: 🧵 1/4

Are we generating a fluffy web? Building off a recent blog by Dan (and playing with one of his trained models), I'm taking a look at how fluffy SERPs really are. It's not quite ready yet - but coming soon. If you're interested and want to take a look before I'm ready to publish, let me know!

“How do you become a trusted source in the eyes of machines?” If this question isn’t taking up some of your time right now, it should be from here on! In another blinder, Duance has shared a continuation of his thoughts on “Chunked, Retrieved, Synthesized.” Read the links as a pair... 🧵 1/3

Worried about AI Overviews hallucinating? RAG might worsen it More context can boost confidence but reduce a model’s ability to say “I don’t know,” making hallucinations more likely Google’s trying to fix it but “safe & trustworthy” AI in search feels a long way off

To get the most out of AI search, we have to better understand the user's journey 🕵️♂️ Yes, that means things like "personas" might start being mentioned in SEO circles again, as much as that may pain some of you. 🧵 1/5

Hey All, I'm so pumped that our Campfire Chat returns tomorrow at 8AM MST with Dan Petrovic!!! Join us on YouTube at youtube.com/live/3c3SzWn...

There's almost too much to summarise from this one. Massimiliano has pulled a blinder in his deconstruction of Google's AI Search processes as part of the antitrust case. I advise you to try to work through it, although it is a long and complex read, but the things I found most important: 🧵 1/6

Will hype of AI's capabilities kill or severely hamper it? Not directly related to SEO, but it offers valuable lessons. Like SEO, AI in science has been heavily (over)sold as giving transformative results, and whilst this can be true, the hype can often overwhelm all instances where it isn't 🧵 1/6

Lots has been written about Google's new AI Mode, but it often isn't super relatable to those new to the underpinning concepts. We need to be prepared for our content to be a chunk in the journey of the customer - a fundamental mindset shift for SEOs

We analysed 8,489,351 English PAA results with @alsoasked.com over the last few weeks and found a whopping 12.6% of answers in these boxes are now AI-generated by Google 😱 Will update monthly. Maybe represents opportunity where content answering “full” intent does not exist? 🤔

While you read this, Google probably ran 2-3 search quality tests. In 2023 alone: > 17,000 experiments > 700,000 search quality tests (via RankHouse blog) Search is constantly evolving. Are we adapting fast enough or just reacting to what works today? More here:

AI Overviews may reduce clicks but raise impressions— a trend Google has hinted at. Are SEOs ready if content helps convert later, but we lose attribution? The challenge isn’t impressions vs conversions - it’s proving value when we see only part of the picture

Working in search and unsure what job title you want in 2025 and beyond? I have you covered.* *Just a bit of fun, it's still all SEO really!