Work in E-commerce SEO? I can guarantee there’s data you’re missing from Google Merchant Centre, OR there is data you could be making more of. The New Merchant API is easier to work with, more modular, and offers new and enhanced features. Of particular interest: 🧵 1/2

Still on the lookout for KPIs for an Age of AI Search? I am! “AI tracking is a waste of time” or “Tool providers are doing a bad job” are really easy positions to take, but are you investing time/energy to find a better way? John Lovett from Seer has added more thoughts to this process. 🧵 1/4

Thrilled to receive my copy of this one, it’s been a few years since I last contributed to SEO in __ from David/Majestic - and what a whopper this is! I think this year’s is probably the most important. Why? Well, 2025 has been full of surprises, hype, excitement, mystery and worry for many 🧵 1/2

We’ve all seen it. Google’s AI Overviews not knowing the days of the week or ChatGPT failing to count the number of letters in a word… or counting letters at all! Does this mean LLMs are dumb, though? 🧵 1/7

SEOs have been mildly obsessed with knowledge graphs for a while, but mostly because we want to ensure that our websites/brands/clients are well represented in Google’s Knowledge Graph, which makes sense. 🧵 1/11

Agents (or agentic workflows) have potential to change a lot I'd argue that building “agentic” workflows will be a significant part of our jobs soon. In a world where it is relatively easy to build a workflow, those people who can do it effectively and efficiently ARE going to have the edge. 🧵 1/3

Attention workflow builders, and why not EVERY part of your new workflow needs to be using an LLM. Having tried several approaches with some “interesting” constraints and really tough success criteria, it’s really clear that A LOT of workflows out there really aren’t worth their time. 🧵 1/8

Morning after Search Central Live 2025. I have a head full of ideas for the new shopping API, a crap-load of questions and met a load of (new and old) people who make me glad to be in this weird, dysfunctional, fascinating industry. Thanks to Martin, John and the team for putting this on!

This is something I can’t wait to dig into further. It feels like Nick's been rummaging around in my head to find a topic that frequently causes me stress 🤯 “Brand” value can be quantified in “SEO data”, but it’s not always a clear or obvious picture as to the impact of SEO/Search Teams. 🧵 1/4

So, where is your Bing Strategy? Bing still drives significantly more traffic to websites than AI chat experiences today. This is almost a meme response to AI/LLM hype. 🧵 1/5

This is neat, smart and well-timed 👏 I’ve been doing a lot of hand-waving around feed optimisation recently, and this tool from Martin Aberastegue makes this much easier. If you wondered whether your existing merchant feeds are “ready” for OpenAI, this will help you understand that 🧵 1/2

Is ChatGPT-5 sending fewer visitors because it’s better than the previous version? Potentially. Interesting thought. “GPT-4.5… sometimes created URLs or citations that looked real but were not always accurate… GPT-5 stopped doing this. 🧵 1/4

SEOs could be testing more, with very few exceptions. There are sometimes good reasons not to test, but knowing when, where & how is really important. If I were learning SEO testing from anyone in the biz, SearchPilot is my pick - and there's a free course -

A post-click world is a scary place 😱 It’s also a dystopian future that it’s easy to get excited about, but probably won’t be what we think it will be…or will it? A sub 1% CTR from ChatGPT results is pretty sobering. Even as a thought exercise, but what if that’s as good as it gets? 🧵 1/3

Here’s a very common - yet important - mistake people make We talk about LLMs (Large Language Models) a lot - they are becoming the centre of “AI Search” - but if we REALLY want to understand how results are generated and how to optimise for them. We have to get a bit more precise than that 🧵 1/7

ChatGPT Web search evocation update I’ve been adding to buff.ly/j3N1nD4 as data comes out, and spoken to SE Ranking, who've been good enough to share their data Out of 100k keywords, they returned citation sources 15.77% of the time (from web search), but that number risks losing nuance 🧵 1/3

Here’s a data source to bookmark and watch closely 👀 Aleyda’s findings here aren’t super-surprising, but they are still interesting. AI traffic is low, relative to Organic, and it impacts certain sites/industries more than others. But this is my favourite: 🧵 1/3

Wonder how algorithm updates would play out in AI Search? Less of a worry for newer members of the SEO community, as algorithm updates are less well telegraphed and understood. If you’ve been in the industry a while, they’ll be formative experiences and a “boogie man” you warn kids about. 🧵 1/8

We’re deep into the final edits of the SEO chapter for the Web Almanac of 2025 (nearly complete!), and it’s been a great learning experience. I have had the opportunity to work alongside some really dedicated and talented SEOs, as well as the wider HTTP Archive team, who are incredible. 🧵 1/3

We’re in desperate need of better data from Google. I’ve felt this ever since Google stopped showing queries in Google Analytics (not provided), but it's now borderline critical. The thing is, this lack of data doesn’t benefit anyone. 🧵 1/7

E-commerce, listen up - how many of you have clocked this yet? OpenAI’s developer docs for their merchant feed has this interesting attribute: - q_and_a - A recommended (not required) attribute, with the simple description “FAQ Content” buff.ly/BHwtyQE 🧵 1/4

Agents are coming for your traffic and revenue. This shouldn’t be considered a bad thing! Except… if you don’t know whether agents can correctly navigate around your site or not. 🧵 1/4

Want an easy way to understand if a ChatBot prompt is going to give you a good output or not? Read on! It’s no secret that I’m not a fan of prompt libraries. Not because I think the concept is flawed, but because I think the majority of prompts shared in them aren’t actually that good. 🧵 1/5

OpenAI’s publication of its product feed schema was something many were waiting for. The feeling that e-commerce feels a little helpless with AI (at least right now) is waning as the per-for-exposure conversations and the mechanics that’ll make them work are starting to become clearer. 🧵 1/7

Time to get comfortable with the fact that things in SEO aren't fixed. By 'fixed,' I mean 'stable' or 'unchanging.' But there are a number of elements we need to be more comfortable with. What am I talking about this time? Search Grounding. 🧵 1/7

As AI is moving so quickly, how do you decide when targeting your strategies and initiatives? If you go too futuristic, you are asking for an investment in the future that may/may not be repaid. Not future-thinking enough, and you may be caught out if the change is overwhelming. 🧵 1/8

So you think all the worry about rank tracking data is new? It’s not, it’s old - like as old as “SEO is Dead”, and other SEO horror stories you keep the kids awake at night with. 🧵 1/4

This article is essential reading. Great work, JesScholz 👏 If you’re not preparing your website for agents yet, there’s a lot of good food for thought here. I’ve cribbed some key points from it, but you should also check it out if you get the chance. 🧵 1/3

Yass, a business-justifiable reason to visit Zurich! Coffee & hills, I'm coming for you (too!)

Where “traditional” search and AI search differ is important. Many people have, however, pointed to this difference as an inflexion point in SEO and, therefore, heralding the death of the “old way” of doing things. A more productive way to look at this difference is how Duane puts it here. 🧵 1/4

I really wouldn’t want to be a publisher right about now, but I’ve been thinking that for the last few years LLMs have been the cherry on top But reading how the Economist plans to define itself against AI/LLMs in the next three years IS super interesting: 🧵 1/4

Working with search data can be unforgiving. About two weeks after I started work on repositioning impressions as important metrics in an age of AI search, Google swoops in and does a proverbial table flip. 🧵 1/8

How to use the Streamlit Regex Tool: - Paste URLs directly into the "Enter URLs" text field (one URL per line) - Configure Regex Options: 1. Enter Domain: Provide the domain you want to strip (e.g., example.com) 2. Wildcard Options: Enable partial matching at the start or end 🧵 1/2

Since digging into Web.Search in ChatGPT 5 (SonicBerry), I’ve been testing how to explain this function and work with it. Others, such as Mark Williams-Cook, Britney Muller, Mihir Naik, Dan Petrovic etc., have all dug into this and shared their thoughts, which contributed to my thinking too. 🧵 1/7

“If you’re reading this, you’ve likely wrestled with time-series forecasting” Yes! “Perhaps you’ve implemented Facebook Prophet” … guilty. “Prophet excels at business time-series with strong seasonal patterns but requires manual tuning for each new dataset”. Ain’t that the truth! 🧵 1/3

Why chaining prompts can help you work with AI/LLMs. If you work with LLMs or have a specific focus to add value with the AI processes, one thing you may have hit against is what happens when you start using longer and longer prompts. 🧵 1/7

Where’s the best place to keep up with SEO news? And here’s another way to shortcut you to some of that least! I was thinking about the question I'm asked the most in podcasts and when I give talks/workshops, and one of them is how to stay informed about SEO and related news. 🧵 1/4

SEO’s intent state system is kind of broken. This was before AI Search, in my opinion. For IR/Data Science, the question of intent types has been long-discussed, but what usually bubbles up (across, diagonally?) into SEO is often over-simplified and more flawed than we’d expect. 🧵 1/8

After sharing a load of my SEO tools a few weeks back, one of the requests I got was for a more detailed guide on how to use Prophet… Here it is -

If you see me at @brightonseo.com, make sure you ask me for a sticker. We need to get this standard official 🐈

Essential reading for e-commerce in the age of AI There’s a lot of panic and general anxiety on how it impacts e-commerce - but the details are slowly emerging! In this blog, Chris (that one again), looks at Agentic commerce and how Google (and others) are preparing. 🧵 1/4

If you knew, with certainty, that organic clicks (through Google) would drop by xx% by the end of 2026, what would you do now? Quite an intimidating prospect (assuming you’re not expecting that already), but as a thought experiment, I think it is one we need to be training ourselves on. 🧵 1/2

Footers are back? I’m not sure they ever left, but I can agree that I’m glad I don’t see many sites with 50+ links in the footer anymore. That was once valuable for SEO, but now - thankfully - not so much. 🧵 1/3

Web Guide, AI Mode and AI Overviews - what next? Gianluca has created a super-detailed explainer into Google’s Web Guide AND set it against the already-familiar AI Overview and AI Mode “surfaces”. 🧵 1/4

I had some great feedback on the Google Trends Stitcher APP previously, but one of the biggest requests was to include DataForSEO as well. So I did - trends-stitcher.streamlit.app 🧵 1/2

Chris (another one) questions the reliability of X-robots headers in the world of LLMs/AI search. "If you’re serious about protecting your data, you need stronger measures: Explicit robots.txt blocks for AI crawlers that claim to respect them. Server-level blocks against suspicious IPs. 🧵 1/2

You’re fed up with chunking, right? Wait a sec! There are enough well-respected SEOs who've given sensible accounts for why obsessing over “chunking” your content could be a waste of time from an LLM visibility standpoint. Beyond just readable, “well chunked” content, I'm inclined to agree. 🧵 1/4

Do you have some time to provide some advice? One of the tools I started creating a while back was my People Also Asked (PAA) moments App - www.chris-green.net/post/user-mo... 🧵 1/4

“Agents” are all over your site. Well…maybe. But are you even sure how to track them? If you’re fortunate enough to be able to check your server access logs, you’ll see it happening. 🧵 1/5

AI Search is more volatile plays on my mind. What if you spend months achieving visibility for key events in a user’s journey, only for it to be lost again a few days/weeks later? www.tryprofound.com/blog/ai-sear... suggests this is more of a feature in AI search, which may be unavoidable. 🧵 1/5