Those who know me know I have a "love-hate relationship" with prompt tracking. Done right, it can be insightful, so I came up with 3 frameworks and a free skill to help you choose, get started and write your prompts 👇 oritmutznik.com/ai-search-op… The hate: The same prompt rarely gives the same answer twice (@SparkToro saw the same brand list less than 1% of the time). The "love": Read as a trend, it shows who gets named in your place, how you’re described and which pages get cited. So each of my frameworks answers a different question: 1️⃣ By buyer expertise: Which buyers does the AI recommend you to, from first-timer to expert? 💡 For products that serve first-timers and experts alike. 2️⃣ By funnel stage: At which stages of the buying journey does your brand appear? 💡 The fastest first read with retention included. 3️⃣ Both, in a buyer journey matrix: Where exactly does your brand appear, by buyer level and stage? 💡 For the one product that carries the business. 🧰 The free skill: Attach it to any LLM, answer 3 quick rounds of questions and get a framework plus 25 to 50 prompts, real queries first. 💡 Run the set by hand first as a stress test of your framework of choice and then move it into your tracking tool. Includes methods, inspiration and research from @marktraphagen, @randfish , @lilyraynyc , @suganthan , @chris_nectiv , @ChouinardJC and Jason Packer 🙏 Check it out👇🏼 oritmutznik.com/ai-search-op… #ai #seo #aeo #geo #aisearch #prompttracking #aivisibility

We have come full circle #seo

All well and good. But a lot of these problems are self-inflicted and can be sorted with good old-fashioned technical SEO. For example, Profound's own pricing page has a features table that's... not a table. Which means that when it's fetched by a bot/agent that's not layout-aware, it can end up as a jumbled mess of markdown, with missing feature flags, duplicated sections, and values disconnected from their labels. Not very machine readable. Third screenshot here is a fetch of their pricing page via Exa, which is used by a lot of agents and I assume they care about. Other extraction pipelines may encounter similar problems. Yes, there are genuine cases where this (unrendered content) can be an issue. But this particular example is just a straightforward technical SEO problem. So before you get to fancy solutions and serving different documents to different user agents, start by... making sure your main document is well structured. Happy to help Profound with the fix, although I assume there are people on their team that can handle it. I do see this regularly on Ecommerce sites, so they're not alone. But probably quite important that your pricing page is machine readable, particularly if you're marketing a product around all that. Notes: 1. Image 4 is what ChatGPT was able to reconstruct from the Exa return. 2. ChatGPT's own fetch did return the missing fields. Sidenote: sometimes a table should be a table 🤷

Those who know me know I have a "love-hate relationship" with prompt tracking. Done right, it can be insightful, so I came up with 3 frameworks and a free skill to help you choose, get started and write your prompts 👇 oritmutznik.com/ai-search-op… The hate: The same prompt rarely gives the same answer twice (@SparkToro saw the same brand list less than 1% of the time). The "love": Read as a trend, it shows who gets named in your place, how you’re described and which pages get cited. So each of my frameworks answers a different question: 1️⃣ By buyer expertise: Which buyers does the AI recommend you to, from first-timer to expert? 💡 For products that serve first-timers and experts alike. 2️⃣ By funnel stage: At which stages of the buying journey does your brand appear? 💡 The fastest first read with retention included. 3️⃣ Both, in a buyer journey matrix: Where exactly does your brand appear, by buyer level and stage? 💡 For the one product that carries the business. 🧰 The free skill: Attach it to any LLM, answer 3 quick rounds of questions and get a framework plus 25 to 50 prompts, real queries first. 💡 Run the set by hand first as a stress test of your framework of choice and then move it into your tracking tool. Includes methods, inspiration and research from @marktraphagen, @randfish , @lilyraynyc , @suganthan , @chris_nectiv , @ChouinardJC and Jason Packer 🙏 Check it out👇🏼 oritmutznik.com/ai-search-op… #ai #seo #aeo #geo #aisearch #prompttracking #aivisibility

DeepSeek just opened its search index to SEOs and developers. Instead of just posting about the news like everyone else, I decided to actually test it. 110 API calls later, I’d spent a grand total of $0.08. The API exposes authority scores, reranking scores and the sources it retrieves. DeepSeek also claims to have protections against SEO and GEO spam. So I put that to the test. My website got an authority score of 1 out of 4. A “Free Robux Generator” PDF on a government gov site got 4 out of 4. The highest authority tier. Sir, I’d like to speak to your manager. 😬 Jokes aside, this is a significant release for SEOs. We can now query DeepSeek’s own search engine and inspect its sources, scores and rankings directly. I’ve written up my preliminary findings, including some questionable rankings, spam results and how its fetcher accesses websites. suganthan.com/blog/deepseek-…

if that ain't proof idk what is

The site I’m working on saw its average CTR in France drop from 1.63% to .53% and clicks drop by 80% after AI Overviews were launched there. 🫠

Looking to buy #Raye tickets at #RoyalAlbertHall on the 23rd of Oct 2026 - tickets sold out in a sec, mostly to ai bots and scalpers. Everyone I spoke to was a scammer, sending photoshopped/ai generated ticket proofs. Please can a human sell me a ticket?🙏🏼🙏🏼🙏🏼

am loving Salience so far @carlhendy - top page crawled by LLMS (so far...) 🥁 robots.txt!

Hey @TicketmasterUK if there are no tickets, why am I still in the queue?! Why do you waste people's time? Trying to get tickets in the last 3 days where AI bots and scalpers always win😭 This queue definitely does not contain hope🫠 #raye @raye

Before you report an unconfirmed Google update or panic over a major drop in your rank tracker: check if you were served a poisoned SERP. An interesting new study from @nozzleio looked at 27.5M SERPs across 49 countries and found Google has started actively poisoning results served to scrapers. Instead of blocking requests with 429s or CAPTCHAs (which just tells scrapers to rotate IPs and retry), Google serves a fully rendered, completely normal-looking SERP. What this is resulting in is that the organic results are quietly swapped out with irrelevant data. Some key observations from the study: • 8% of all SERPs were poisoned post-Sept 13 (peaking at ~27% between Sept 27–29) • Organic results kept only ~21% of their usual sites (vs. 88% normal) • Tracked client sites fell out of the top 10 on 58.6% of poisoned SERPs • Social sites like Reddit (+8 pts), Facebook (+5.5 pts), and LinkedIn surged to fill the gaps • Interestingly, paid modules (Ads and Shopping) remained almost completely intact, while AI Overviews and PAA dropped significantly • Desktop pagination was heavily truncated (stopping at page 4 or earlier on deep queries) This is essentially Google's countermeasure against automated AI scrapers and RAG grounding pipelines, but rank trackers are inevitably being impacted. If your client's ranking report shows a significant decline overnight, check GSC average position first. If GSC hasn't displayed a major deviation, it is likely that it is a data anomaly rather than a change in your ranking Read the full write-up and data: nozzle.io/research/poisoned-…

What can I say about @nickwilsdon 😞He always made me laugh and made sure I'm comfortable in any situation, he made so many others laugh too with his contagious laughter and superior humour. He always tried to find a way for me to make it work at @TorquePartners if I ever take the plunge to independent consulting (and was actively simplifying what it would take, guiding me through the steps). I'll miss his stories about edge SEO, his chickens and quiet life in Ireland. Such a kind young soul. Speaking about him in past tense is surreal. You'll be terribly missed by anyone who ever crossed paths with you, Nick💔 My heart goes out to your family and loved ones. Rest in peace🥲

RT @suganthan: Still can’t believe you’re gone Nick. 💔 You were always so generous with your time, always happy to share advice and help o…

I’m deeply sad today as so many in the SEO community in the UK, Europe and beyond about the passing of a wonderful member of the search community, excellent friend and wonderful human, @nickwilsdon , with whom we shared so many moments through the years. He was so funny, kind, supportive, smart… what an amazing human. You will be deeply missed, Nick. I can’t still believe you’re gone. You’ll be always remembered… thank you for all the beautiful moments. My deepest condolences to his wife Anna and kids, and his team at Torque Partnership, the awesome consultancy he successfully built and was so proud of. ❤️ whatever you need, I’m here. A big hug in heaven, Nick 🙏 RIP, my friend.

JUST IN: “Goth MAGA” Republican Kelley Dennison, 27, could make history as the youngest woman ever elected to Congress. 2% chance.