How do AI answer engines decide which sources to cite?
Retrieval-augmented generation picks passages, not pages. Understanding the four steps between a question and a citation explains most of what people find mysterious about AI visibility.
Insights
Insights
Notes on how answer engines pick their sources, how new markets are actually tested, and why good deals stall. Every figure carries the study it came from.
Getting found when buyers ask AI
How retrieval works, where AI answers get their information, and what makes a page eligible to be cited. See the service →
-
-
What actually works in AEO, and what does not
Three tactics have real evidence: pages readable without JavaScript, sources and statistics in the text, and consistent listings. Three are fiction.
-
How do you measure AI visibility honestly?
Run fixed buyer questions on each engine on a schedule, and record four things: named or not, cited or not, which source, and who was named instead.
-
Which sources should you fix first if AI is not naming you?
Yext found 44% of AI citations came from brands' own sites and 42% from listings and profiles (October 2025). Fix crawlability first, then your profiles.
-
How do you write a page an AI engine can actually quote?
Engines quote a chunk, not the page. Each section has to answer its heading in the first sentence, name its sources inline and run about 40 to 150...
-
Why are buyers asking AI before they ask you?
Buyers shortlist in AI assistants, and absence leaves no trace in analytics. Yet Yext found 86% of AI citations come from sources a brand can influence.
-
What to ask an AEO agency before you hire one
Nine questions that separate a real answer engine optimisation practice from a renamed SEO package, and what a weak answer to each one sounds like.
-
Schema and sameAs: how a model decides your company exists
Before an engine can recommend you it has to resolve which company you are. That is an entity problem, and structured data with a complete sameAs array is...
-
Half of what AI cites about you is not on your website
Yext analysed 6.8 million AI citations and found 42% came from listings and profiles against 44% from first-party sites. If half the surface area sits outside your CMS,...
-
Ranking top ten used to mean getting cited. Ahrefs says that halved in eight months.
In July 2025, 76% of pages cited in Google AI Overviews also ranked in the top ten. By March 2026, Ahrefs measured 38%. What that change does and...
-
Most AI crawlers do not run your JavaScript. Here is how to check in one command.
Vercel and MERJ found that none of the major AI crawlers execute JavaScript. If your content only appears after JavaScript runs, most answer engines never see it. Here...
Entering a new market
What an entry costs before the first order, how to test a segment without hiring into it, and why technical products fail for non-technical reasons. See the service →
-
Why do technical products fail in new markets for non-technical reasons?
The product was never the variable. Translated propositions, a different buying process, missing listings and buyer risk cause most of the failures.
-
How do you test a country or segment before committing headcount?
Four cheap tests, each able to stop the project: name 30 to 50 target companies, verify demand, test two propositions, then score the quality of replies.
-
What does a market entry actually cost before the first order?
Mostly a year and a salary, and the salary is the smaller half. The costs that decide it, such as ramp time and lost focus, rarely reach the...
Turning interest into signed deals
Why most B2B deals are lost to nobody, who actually decides, and what to change when a deal goes quiet. See the service →
-
What should you change when a deal goes quiet?
Work out which kind of quiet it is first. The JOLT Effect research found 40–60% of B2B deals are lost to indecision, which needs less risk, not a...
-
Who actually decides in a B2B purchase, and why do they stall?
Gartner found in May 2025 that 74% of B2B buying teams show unhealthy conflict. Six positions decide a purchase, and two of them do most of the stalling.
-
Why are most B2B deals lost to nobody?
The JOLT Effect research, from 2.5 million recorded sales conversations, found 40–60% of B2B deals are lost to indecision. Lower the risk, not the price.
Find out what the engines currently say about you
Send me your website and a few words on what you sell. I run a set of buyer questions across ChatGPT, Perplexity, Gemini, Claude and Google's AI Overviews, and send back a recorded walkthrough of what came out: where you appeared, where you did not, who was named instead, and whether AI crawlers can read your pages at all. No charge, no obligation, and you keep the findings whether or not you decide to hire me.
I run these myself, so there is a queue. Expect a few working days rather than an instant report.