How to get cited by ChatGPT, Perplexity and Gemini

What does it take to get cited by an AI answer engine?

An AI answer engine such as ChatGPT search, Perplexity, Gemini or Google AI Overviews works in two steps. It retrieves a handful of pages that look relevant to the question, then writes an answer and links some of them. To be cited you have to survive both steps: your page has to be retrievable, and it has to contain a passage the model can lift and attribute.

That gives you three levers. Access: can the crawler read the page at all? Fit: does the page answer the question in the words people use? Trust: do other sources already talk about the brand? The evidence so far says trust weighs most, access is a hard gate, and fit decides which of your pages wins.

Which signals matter most in 2026?

The strongest public evidence points away from tricks and towards reputation. A round-up of the 2026 studies by Geoptimizer reports that Muck Rack found 84% of links cited by AI engines were earned media, that Discovered Labs measured AI-perceived domain authority at roughly six times the influence of the strongest single page-level feature, and that Ahrefs found branded web mentions correlated with AI visibility at 0.664, well ahead of backlinks at 0.218.

The same round-up lists what did not move citations in matched tests: adding schema markup and publishing an llms.txt file. Google says the same about its own products. Its May 2026 guide to generative AI features states that special markup or AI-only files are not required, because those features run on the core Search ranking and quality systems.

How do you make your pages readable to AI crawlers?

Most AI crawlers fetch raw HTML and stop. Vercel and MERJ tracked hundreds of millions of GPTBot and Claude requests and found no evidence that either executed JavaScript. If your key content only appears after a script runs, those crawlers see an empty shell. Render the words, headings and links on the server, or prerender them at build time.

Then check access. OpenAI documents separate agents for search (OAI-SearchBot), model training (GPTBot) and user-triggered visits (ChatGPT-User). Blocking the training bot does not have to block search visibility, so decide each one on purpose in robots.txt rather than copying a blanket rule.

How should you write pages that get quoted?

Write for the question, not the keyword. Collect the questions buyers actually ask an assistant in your category, then make sure one page answers each of them in the first two or three sentences under a heading that repeats the question. Models lift short, self-contained passages; a page that buries the answer under an introduction loses to one that leads with it.

Make the passage specific. Numbers, named methods, dates and a clear owner give a model something to attribute. Keep the page about one job, but you do not need to chop it into fragments: Google says its systems can find the relevant passage inside a longer page.

• Lead each section with a one- or two-sentence answer.
• Use the buyer’s own wording in the heading.
• Add a dated, sourced number where you have one.
• Link the page from the rest of your site so crawlers find it.

How do you measure AI citations week to week?

Pick 30 to 100 questions a buyer would ask, run them through each engine you care about on a fixed day each week, and record whether your brand is named, which URL is cited, and which competitors appear. The share of answers that cite you is your AI share of voice. Track the trend, not single runs: answers vary from one run to the next.

This is the loop behind one of my own projects. At Imagera AI the brand appeared in about 8% of tracked AI answers. A weekly citation check across the answer engines, structured data and fixes to the pages that earn the mention took AI-answer citation share from ~8% to ~19%, tracked through to signups. The free AI share-of-voice tool on this site does the arithmetic for your own prompt set.

No, and that changes where you spend effort. Each engine retrieves from its own index and weighs sources differently, so a page that Perplexity cites for a question may not appear in ChatGPT’s answer to the same question, and Google’s AI Overviews draw on Google’s own ranking systems. That is why the weekly check should record results per engine rather than as one blended number.

Start with the engine your buyers use most, look at the domains it cites for your category, and work out which of them you can realistically appear on. Then repeat for the next engine. Over a quarter this produces a short list of target sources per engine, which is far more useful than a generic list of “top sites”.

How do you earn the mentions AI engines trust?

Treat mentions as the main channel, not a side effect. Make a list of the places your buyers already read before they ask an assistant: industry newsletters, review sites, comparison articles, community threads, podcasts, partner directories and the documentation of tools you integrate with. Each one that names your brand with a sentence of context is a source an engine can retrieve and quote.

Give those places a reason to mention you. Publish original numbers from your own work, offer a clear expert view on a question people argue about, contribute a useful answer in a community thread without a sales pitch, and make it easy for partners to describe what you do by giving them a short, accurate paragraph and a link. Original data travels furthest, because other writers need something to cite.

Keep the description of your brand consistent everywhere. When a review site, a partner page and your own home page all say the same thing about what you do and who it is for, engines have an easier time connecting the mentions to one entity.

• Publish one piece of original data a quarter and pitch it to the newsletters your buyers read.
• Answer real questions in the communities your buyers use, under your own name.
• Keep a short, accurate brand description that partners and reviewers can reuse.
• Ask happy customers for a review on the sites assistants cite in your category.

What should you stop doing?

Stop expecting a file or a tag to do the work. Keep schema where it helps search features and accessibility, but do not count on it for citations. Do not publish hundreds of near-identical pages hoping one gets picked up; Google’s scaled content abuse policy targets exactly that, and the September 2026 spam update hit templated programmatic pages hard. Spend the time on content worth quoting and on getting other people to mention you.