Most brands are invisible in AI answers for a boring reason. Nobody ever published the sentence the machine could quote. The model is not hiding you. It has nothing of yours worth repeating.
The average brand gets named in about 17% of AI answers about its own category. That is the baseline, and beating it is a sequence of five moves done in order. None of them requires buying anything.
Where do AI answers come from?
When someone asks ChatGPT, Claude, Gemini or Perplexity a buying question, the engine assembles an answer from two places: what it remembers from training, and what it retrieves live from the web. Both routes reward the same thing. A clear, self-contained, verifiable claim on a page the engine can reach and trusts.
So the job is not persuading a model. The job is being the best available source for a question it keeps getting asked. The distance between the questions your buyers ask and the answers that name you is your citation gap, and everything below is how you close it.
Which questions should you target first?
Start from the questions your buyers ask right before they find you, phrased the way a person types at 10pm. Best tool for a 40-person team. Alternatives to the market leader. Is this category worth paying for. Ask each one yourself across the four big engines and write down two things: who gets named, and which pages get cited.
The cited pages matter more than the answer. They are the supply chain your buyers' answers get built from, and they tell you exactly where a mention of you would have to live. I walked through the two-minute version of this in Go ask ChatGPT about your own category. Target the questions where the current answer is vague, generic, or missing a source. Those are open goals. A question where three competitors already get cited on every engine is a longer campaign.
What does a quotable page look like?
One question per page, and the direct answer inside the first fifty words, before any wind-up. A model quotes the way a journalist on deadline quotes: it lifts the cleanest self-contained passage that answers the question and moves on. Give it one.
Structure the rest of the page as the follow-up questions a real buyer would ask, each answered in a short section that survives being read alone. Then load the page with things worth lifting. A number with a date and a source. A named example. A definition sharp enough to repeat at a dinner table. Ahrefs found AI-referred visitors were 0.5% of traffic and 12.1% of signups, and that stat travels precisely because it is specific, sourced and surprising. Fog does not get cited. I made the longer argument in Stop writing to rank. Write to be quoted.
Can the engines actually read your site?
This is the unglamorous gate. Your robots.txt has to allow the AI crawlers by name, because several of them do not hide behind the wildcard rule. Your content has to exist in the HTML the crawler receives, not appear later through scripts. Your pages should say plainly who you are, in schema and in text, with the same description everywhere it appears.
An afternoon covers all of it, and skipping that afternoon quietly zeroes out everything else on this page. Blocking the crawlers, which some companies still do on principle, just means the engines describe you using other people's words. You stay in the conversation either way. You only lose your seat at it.
Why do you need mentions on pages you don't own?
Open any AI answer to a buying question and read the citations. Most of them are pages the named brands do not control: comparison articles, review platforms, Reddit threads, industry publications. An engine deciding whether to recommend you behaves like a careful buyer. It checks what everyone else says about you before repeating your claims.
So the second half of the work happens off your site. Show up with real answers in the communities where your buyers ask. Get reviewed where your category gets reviewed. Get named in the publications and listicles the engines already cite for your questions, which you know, because you collected them in the research step. Your own site makes you quotable. Third parties make you credible.
When should you start measuring?
Last. A tracker pointed at a brand with no quotable answers is a scale in an empty kitchen, and I have already said what I think of paying for the scale. Re-ask your target questions monthly, by hand or with tooling once the prompt count justifies it, and expect drift: the engines disagree with each other and with their own last month. Eight models agree on a category's top pick less than half the time. That volatility is the opportunity. Shelf space is still being assigned.
Questions people ask about this
How long does it take to get mentioned in AI answers?
Weeks, for the engines that search the live web. Perplexity, Google's AI Overviews and AI Mode re-crawl constantly and can cite a new page within weeks of publishing. ChatGPT is slower on questions it answers from memory and faster on questions it browses for. Plan in quarters, check monthly.
Is getting mentioned by AI different from SEO?
The foundations overlap: crawlable pages, clear structure, real authority. What changed is the unit of success. SEO optimises for the click on a results page. AI answers often end without a click, so the win is being named and cited inside the answer itself.
Do you need an AEO tool to get mentioned in AI answers?
Not to start. Asking the engines your buyers' real questions by hand is free and tells you where you stand today. A tracking tool earns its subscription after you have published answers worth tracking, when the job becomes watching movement across hundreds of prompts.
Does blocking AI crawlers protect your content?
It removes your voice from the answer while your competitors stay in it. The engines still learn about you from third-party pages you don't control. Blocking the crawlers only guarantees that when your category comes up, someone else supplies the words.
Alex Muresan does go-to-market distribution for founders who'd rather be right early than safe and late. Get in touch →
If you need help getting better outcomes in AI answers, that is my day job: I run growth at Reach, where this loop is the product. usereach.ai →
Part of Native Organic Distribution, the motion these posts describe. The glossary →