Why Doesn't My Site Show Up in AI Answers?
Why doesn’t my site show up in AI answers?
Strong rankings improve your chances of being cited by an AI model but don’t guarantee it, because ranking and citation are decided on different criteria. A search engine is choosing which page best answers a query. A model is deciding which sources it trusts enough to build an answer from, which usually means it needs your claim to be cleanly extractable and needs someone other than you to have described your company first.
Here is a situation I see constantly. A company holds position two or three for the terms that matter in their category. The dashboard looks healthy. Then someone on the team asks ChatGPT the exact question that page was built to answer, and the response names four competitors and not them.
Nothing broke. Rankings did not slip. The page is still there, still ranking, still getting a fraction of the traffic it used to. It is simply not being read by the system that now sits between the question and the answer.
Do organic rankings still matter for AI visibility?
Yes, as a strong signal rather than a guarantee — and this is where I want to be careful, because the numbers being passed around on this topic disagree with each other, and most of the people citing them are selling something.
AirOps found that roughly 60% of AI Overview citations come from URLs outside the top 20 organic results. That figure gets quoted constantly, usually as proof that SEO is finished. But seoClarity, analyzing 5.1 million citations across 362,000 AI Overviews, found the opposite emphasis — 56% of citations came from the top 20, and 94% of AI Overviews included at least one URL from the top 20. In the same study, pages ranking first were cited 43% of the time, dropping steadily to 7% at position 20.
Both can be true, and probably are. Different query sets, different time windows, different definitions of a citation. What survives the disagreement is the part everyone is measuring around: strong rankings meaningfully improve your odds and do not guarantee anything. Position one gets cited in fewer than half the answers it is eligible for. Something else is deciding the rest.
If you take one thing from the spread in that data, take this: nobody has a clean formula yet, and anyone selling you one is ahead of the evidence.
What actually decides whether a model cites you?
Rankings answer a question about your page. Citations answer a question about your company.
When a model assembles an answer, it is not scanning a results page and picking the top three. It is pulling fragments from sources it has reason to trust, on a topic it has already formed a representation of, and stitching them into something that reads as settled. Your page is a candidate for one of those fragments. Whether it gets used depends on whether the model can extract a clean claim from it and whether it has corroborating reason to believe you.
That second part is where most companies lose. Muck Rack’s May 2026 analysis of more than 25 million cited links across ChatGPT, Claude, and Gemini found that earned media accounts for 84% of all AI citations, while paid and advertorial content accounts for 0.3%. Journalism alone makes up 27% of cited sources, and that pattern has been stable across three editions of the study going back to July 2025.
The reflexive read is that this is a PR finding, and PR firms have been reading it that way loudly. I think that read is half right and expensive. What the number actually describes is corroboration. Models cite sources that are not you when they are establishing what you are, and they cite you when they are quoting a specific claim. If nothing outside your own domain has ever described your company, there is nothing for the first half of that process to work with, and you do not enter the answer at all — no matter how well the page ranks.
Two more findings from that study are worth more than the headline. Journalism citations span more than 20,000 distinct outlets, which means the trade publication your buyers actually read is a live target, not a consolation prize behind the tier-one outlet you cannot get. And more than half of journalism citations come from articles published within the past 12 months, with citation volume dropping sharply after the first six months. Coverage decays. A mention from 2023 is doing very little for you right now.
What should I fix first?
Four things, in this order. The order is the point — each one depends on the answer before it.
Find out where you actually stand
Before changing anything, run the questions your buyers ask through the models and write down what comes back. Who gets named. How they’re described. Which sources the answer pulls from.
That last part is the one people skip, and it’s the most useful. The sources cited in your category tell you exactly where the corroboration is coming from, which tells you where yours has to come from. It is usually not anyone’s blog.
Say the same thing everywhere
The single most common failure I find is a company that describes itself five different ways across its homepage, its G2 listing, its docs, and its founder’s podcast appearances.
A model reading that cannot determine what you are, so it declines to place you anywhere. One short, stable description, repeated without variation, is worth more than a quarter of new content.
Make yourself extractable, then get described by someone else
Two halves of the same job.
If your best page buries the answer under three paragraphs of setup, states your claim only as a benefit, and never defines the term it is built around, it is hard to quote. Lead with the direct answer. Define your terms explicitly. Use headings that are the question a person would type. This is unglamorous and it is the cheapest lift available.
Then the corroboration. Not a wire release — BuzzStream’s 2026 study found press releases via wire syndication account for 0.04% of AI citations. Original data, a genuinely useful expert comment to a reporter working your beat, a partner writing about the work. The bar is that a third party publishes a sentence describing what you do.
Measure the thing you care about
Rankings are a proxy that no longer maps cleanly to visibility. Pick a set of the questions your buyers actually ask, run them across the models on a schedule, and record whether you appear, how you are described, and who appears next to you.
It is unsexy and it is the only way to know whether any of this is working.
The uncomfortable version
Most of the work above is not new. Clear writing, consistent positioning, earning third-party credibility, and measuring what matters were good ideas in 2015. What has changed is the penalty for skipping them.
An unclear page used to still rank, and the buyer would sort it out. Now an unclear page gets read by a system that will not sort it out, and the buyer never sees you at all. The margin for being vague has closed.
Being better doesn’t help if you’re not in the answer.
Frequently asked
Do organic rankings still matter for AI visibility? Yes, as a strong signal rather than a guarantee. Studies disagree on the size of the effect — estimates of how many AI Overview citations come from outside the top 20 organic results range from roughly 44% to 60% — but they agree that ranking well improves your odds and that a large share of citations go to pages that do not rank highly.
What is the difference between SEO and AEO? SEO optimizes for position on a results page. AEO, or answer engine optimization, optimizes for being cited inside a generated answer. The tactics overlap on technical fundamentals and diverge on extractability, source corroboration, and measurement.
Why does earned media matter so much for AI citations? Because models use third-party sources to establish what a company is before they will describe it. Muck Rack’s May 2026 study found earned media accounts for 84% of AI citations across ChatGPT, Claude, and Gemini, while paid content accounts for 0.3%.
How do I get cited by ChatGPT? Three conditions have to be met: the model can access and parse your content, your site describes what you are in consistent language, and sources it trusts in your category already reference you. Most companies fail the first or second and spend their budget on the third.
How do I measure whether my company appears in AI answers? Define a set of buyer questions, run them across the major models on a regular schedule, and track appearance rate, how your company is described, and which competitors appear alongside you. Keeping the questions consistent over time matters more than how many of them you track.
Sources: AirOps, State of AI Search 2026. seoClarity, 2026. Muck Rack, May 2026. BuzzStream, 2026.
Want to know where you stand? Book a call.