Copilot at work compares vendors from the buyer’s files as well as the web
When a buyer asks Microsoft 365 Copilot to compare vendors, part of the answer comes from their inbox. What that means for your pages and sales material.
· 4 min read
The Obility blog
Field notes for AI search.
Practical guides to better questions, clearer content and more useful measurement.
Hiding your price does not keep it out of AI answers
AI engines quote a figure for contact-sales software from listings, rivals and estimates. What to publish instead of a price list, and how to check it.
· 4 min read
IndexNow tells Bing a page changed. It does not get the page cited
What IndexNow reaches in AI search, from Copilot to ChatGPT and Google, why its value shows up when a fact changes, and how to check it worked.
· 4 min read
AI answers favour big brands when the question asks for one
The study behind “big brand bias” asked engines for the most popular brands. Where a smaller company can win AI answers, and how to measure it.
· 4 min read
Opting out of AI Overviews removes your pages, not the answer buyers see
Google’s Search Console control can exclude your site from AI Overviews and AI Mode. What it removes, what it leaves running, and when a brand should use it.
· 4 min read
A ChatGPT plugin will not get your brand into ChatGPT’s shortlist
OpenAI’s own plugin rules say who finds a ChatGPT plugin and what its metadata may not do. Why a plugin serves customers, not new buyers.
· 4 min read
Backlinks get your pages into AI answers, but mentions get you named
Links help a page get retrieved by AI search. The answer then names whatever that page says. How to judge a link opportunity for AI answers.
· 4 min read
LinkedIn posts can reach ChatGPT answers but not Perplexity’s index
LinkedIn, X and Medium each decide which AI crawlers may read your posts. What their rules allow today, and where to put content you want cited.
· 4 min read
ChatGPT traffic converts better in vendor claims than in the largest study
Claims that ChatGPT visitors convert 4.4 or 23 times better rest on narrow data. What the largest study found, and how to test your own traffic.
· 4 min read
AI answers shortlist you under the words buyers use, not the category you coined
If you describe yourself with a term buyers never type, AI assistants may know you and still leave you off the shortlist. How to test it and fix it.
· 4 min read
When another company shares your name, AI answers have to guess which one you mean
AI answers can describe a namesake, or blend two companies into one. How the mix-up happens, how it skews your tracking, and what a page can do about it.
· 4 min read
Share of voice in AI answers moves with the brands you choose to count
The same AI answers can give you a 19 percent or a 14 percent share of voice. How the competitor list and counting rules set the number, and how to report it.
· 4 min read
The language of the question decides which brands an AI answer names
Ask ChatGPT for a supplier in Norwegian and in English and you can get two different shortlists. What two studies found, and how to measure each market.
· 4 min read
Snippet limits on your pages also limit what AI answers can use
Google applies nosnippet and max-snippet to AI Overviews and AI Mode, and Bing applies NOCACHE and NOARCHIVE to Copilot. How to find these rules and decide.
· 4 min read
When your brand drops out of AI answers, check what the engine changed first
A sudden drop in ChatGPT or Google AI answers often starts on the engine’s side. How to date the change, rule out access and sources, then touch content.
· 4 min read
AI engines cut your page into passages, so write sentences that survive the cut
Answer engines split pages into passages. What Google, Microsoft and Anthropic say about chunking, and how to write passages that make sense alone.
· 4 min read
Answers pulled through an API differ from the ones buyers see in the app
Many AI visibility tools collect answers through developer APIs. Vendor docs and a 13,779 answer test show why those differ, and how to check a tool.
· 4 min read
AI answers draw on search indexes that already filter scaled AI content
Google and Bing now state how they treat AI-written pages, and both rules reach AI answers. What is allowed, what gets filtered, and how to use AI safely.
· 4 min read
Ranking yourself first on your own “best” list gets you cited, until the ranking under it slips
Self-ranked “best X” lists do appear in ChatGPT and AI Overviews sources. What the evidence shows, why the tactic is fragile, and what to publish instead.
· 4 min read
AI agents visit your site in a real browser, so verify them by signature
Why ChatGPT’s agent can pass as a human visitor in analytics, how signed requests prove who sent them, and how to count agent visits honestly.
· 4 min read
AI answers cannot read behind your form, so publish the findings openly
Why a form-gated report is nearly invisible to ChatGPT and Google’s AI answers, what Google allows for gated pages, and what to publish openly.
· 4 min read
Prompt volume for ChatGPT is modelled, not counted, so use it to compare topics
No AI assistant publishes how often a prompt is asked. How vendors build prompt volume figures, and what those numbers can and cannot support.
· 4 min read
The pages AI assistants invent on your site show what buyers expected to find
ChatGPT and other assistants send visitors to URLs on your site that never existed. Which ones to redirect, and what the dead ends tell you.
· 4 min read
Ads in AI answers sit beside the recommendation, so they cannot buy it
ChatGPT and Google now sell ads next to AI answers. What their own documentation says an ad can change, what it cannot, and which buyers ever see one.
· 4 min read
Your AI assistant knows where you work, so it is the wrong place to check your brand
ChatGPT, Claude and Google’s AI Mode remember who you are. Why that skews your own brand checks, and how to run one from a buyer’s starting point.
· 4 min read
Each AI assistant searches a different index, so check your site in each
What Google, Microsoft, OpenAI, Anthropic and Perplexity say about the search systems behind their AI answers, and how to check your pages are in each one.
· 4 min read
AI Overviews take clicks, so the summary becomes your first impression
What Pew Research measured about clicks when Google shows an AI Overview, what Google says in reply, and what to measure when the summary replaces the visit.
· 4 min read
Reddit feeds AI answers, which is why planted comments backfire
How Reddit threads reach ChatGPT and Google’s AI answers, why a seeded brand comment is read alongside its replies, and what to do with Reddit instead.
· 4 min read
A Wikipedia page is a symptom of AI visibility, not a shortcut to it
Why Wikipedia matters to language models, what its company notability rules exclude, and the independent coverage AI answers draw on either way.
· 4 min read
GA4’s AI Assistant channel counts referrers, not every AI visit
What Google Analytics’ AI Assistant channel matches, the ChatGPT visits it can miss, the flaw in Google’s sample regex, and the 13 May trend break.
· 4 min read
AI answers change on every run, so count how often you appear, not where
Why ChatGPT, Claude and Google AI rarely repeat a brand list, why position in an answer is noise, and how many runs a visibility rate needs.
· 4 min read
The famous 40 percent GEO lift was measured with made-up evidence
What the Princeton GEO study behind “add statistics and quotes for AI” actually tested, what its prompts asked for, and how to use the idea honestly.
· 4 min read
A newer date can sway an AI ranker, but only a real update holds up
Research shows language models favour recently dated text. What that means for refreshing pages, why date bumping is a weak bet, and how to test a refresh.
· 4 min read
AI crawlers read the HTML you send, not the page your visitors see
Most AI crawlers do not run JavaScript. Why a client-rendered pricing page can be blank to ChatGPT, Claude and Perplexity, and how to test yours.
· 4 min read
Search Console and Bing now count your AI citations, but not your mentions
What Google’s generative AI report and Bing’s AI Performance report actually measure, what they leave out, and how to read them without overclaiming.
· 4 min read
AI shopping answers pick products from data you only partly control
How ChatGPT shopping and Google AI Mode choose which products to show, what a product feed changes, and what it cannot change.
· 4 min read
When AI gets your company wrong, trace the error before you fix it
A wrong claim about your brand in ChatGPT, Gemini or AI Overviews comes from somewhere. How to find the source, match the fix to the cause and check it held.
· 4 min read
AI answers search for questions your buyer never typed
AI Mode and AI Overviews split one question into many searches. Why a top-ranking page goes uncited, and how to plan pages around the sub-questions.
· 4 min read
Blocking AI training bots does not have to hide you from AI answers
OpenAI, Anthropic, Perplexity and Google split crawling into training, search and live fetches. How to block one without losing the others.
· 4 min read
Build an AI search baseline you can actually use
A practical way to choose prompts, preserve evidence and avoid confusing an isolated answer with a trend.
· 4 min read
Turn an AI answer gap into a useful content brief
An editorial workflow for deciding what to update, what to write and what evidence a draft needs.
· 4 min read
Crawlers, citations and referrals are three different signals
Keep automated requests, answer observations and human visits separate when reviewing AI search performance.
· 3 min read