Why isn't my site showing up in AI search?
Before working through the reasons below, run a GEO audit, a scan tells you which specific signals are failing on your site, so you can skip the causes that don't apply.
There are 5 common reasons a site never shows up in AI search, and almost all of them are mechanical, not mysterious. You block the crawlers, your content only renders in JavaScript, you aren't in Google's index, the page is too thin to quote, or there's no self-contained answer near the top. Google's own position, in Google's AI optimization guide, is that the same fundamentals apply. The rule underneath all of it: “you can't be quoted by a machine that can't read you.”
1. You block the crawlers in robots.txt, often by accident
This is the single most common cause, and most owners do it to themselves. To be cited in ChatGPT, Perplexity, or Claude, you have to let their retrieval bots fetch the page. Those are separate user-agents: OAI-SearchBot for ChatGPT search, PerplexityBot for Perplexity, and Claude-SearchBot for Claude. To appear in Google AI Overviews, you have to let plain Googlebot in, because Overviews are built on Google's normal index.
What goes wrong: an owner reads a scare headline about AI training, decides to "keep the AI out," and adds a blanket Disallow: / for every bot they can name. That removes the site from the answers entirely. Training and retrieval are different jobs done by different bots. You can refuse the training crawler and still welcome the one that cites you.
How to check: open yoursite.com/robots.txt in a browser and read it line by line. Look for any Disallow: / under the search bots above, or a wildcard User-agent: * block that catches them. If the retrieval bots are blocked, that alone explains the silence. Or check your AI crawler access with one scan and get both layers, robots.txt and the edge firewall, read together.
2. Your content only renders with JavaScript
Retrieval crawlers take the raw HTML your server returns. They generally do not run your JavaScript, wait for a framework to boot, and hydrate the page the way a browser does. So if your headline, your answer, and your body copy only appear after JavaScript runs, the crawler sees an empty shell and quotes a competitor whose answer is sitting in the HTML.
A page can look flawless to you and be nearly blank to a bot. This is one of the failures we catch most often on otherwise well-built sites, because modern frameworks default to client-side rendering and nobody told the owner what that costs.
How to check: load your page with JavaScript disabled in your browser settings, or right-click and view the raw page source. Search that source for a sentence you can see on the rendered page. If your real content is not there as plain text, the crawler can't see it either.
3. You aren't indexed by Google
Google AI Overviews are drawn from Google's existing index. There is no separate AI index. If your page can't surface in ordinary Google Search, it can't surface in an Overview either. So an indexing problem is also an AI-search problem, and it has to be fixed first.
Pages drop out of the index for ordinary reasons: a stray noindex tag, a canonical pointing at another URL, a site that's too new to be crawled yet, or a thin page Google chose not to keep. None of it is AI-specific, but all of it ends the same way.
How to check: search Google for site:yoursite.com and see whether your pages come back. For the real answer, open Google Search Console, use the URL Inspection tool on the exact page, and confirm it says the URL is on Google. If it isn't indexed, that is the first thing to fix.
4. The page is thin or commodity
Even a perfectly crawlable, indexed page gets skipped if there is nothing specific worth quoting. AI engines pull passages that say something. A page that restates the obvious in generic language, the same paragraph a hundred other sites already published, gives the engine no reason to choose you over any of them.
The fix is not length for its own sake. It is substance: unique, first-hand, specific content. A real number, a named tool, a step someone actually performed, a result with a date attached. Specifics are what make a passage quotable, because they carry information the generic version doesn't.
How to check: read your page and ask what a person could not get from any other page on the topic. If the honest answer is "nothing," that's the gap. Add the detail only you have.
5. There's no self-contained answer near the top
Engines quote passages that answer the question directly, on their own, without the reader needing the rest of the page. If your answer is buried below three paragraphs of preamble, or split across the page, the engine has nothing clean to lift. The most quotable shape is a short, complete answer placed right under a heading that matches the question.
So lead with the answer in the first paragraph, in plain language, before any windup. Put a specific number or a named entity in it. Use a clear, question-style heading so the engine can match it to what the person typed. This page is built that way on purpose: the answer block sits first, with a number and a source.
How to check: read only your first paragraph. If it does not answer the page's main question by itself, rewrite it until it does.
What you do not need to chase
There is a lot of folklore in this space, so be careful where you spend effort. Google's official AI guide is blunt about it: there is no separate framework for AI search, the same SEO fundamentals apply, and the page has to be crawlable and indexable before anything else matters. Google has also said on the record that it does not use llms.txt, so do not treat that file as a requirement or a shortcut. The work that pays off is the work that always paid off. Be reachable, be readable without JavaScript, be indexed, and be the most specific answer on the page.
Perplexity: let the right crawler in
Open your robots.txt and allow Perplexity's bots explicitly. This is the line most owners get wrong, usually by accident:
# Let Perplexity index and cite you User-agent: PerplexityBot Allow: / # Let Perplexity fetch links users open or share User-agent: Perplexity-User Allow: /
Review PerplexityBot permissions and verified bot traffic if Perplexity visibility matters to you. Its documentation also describes a separate user-triggered fetcher. An allowed request does not guarantee a citation. Check both agents and their published IP ranges.
Gemini: the one Gemini-specific switch
Google-Extended is a robots.txt token, not a crawler you will see in logs. Disallowing it tells Google not to use your content for training future Gemini models or for grounding Gemini answers. It does not remove you from Search, and per the documentation it is not a Search ranking signal in either direction. But the grounding half matters: because grounding is inside Google-Extended's documented scope, blocking it can keep your pages out of grounded Gemini citations even while your Search rankings stay untouched. Background in our Google-Extended glossary entry. If you want Gemini citations, leave it allowed.
Claude: let the right bots in
The robots.txt mistake that costs Claude citations is the same one that costs ChatGPT citations: a blanket rule written to block AI training that also blocks search and user fetches. The 3 Anthropic agents do different work. Blocking ClaudeBot is a training opt-out, a legitimate choice that does not protect visibility. Blocking the search agent removes you from consideration. Blocking the user agent stops Claude from opening your page when a human asks about you specifically, which is the strangest own goal in the list, since that human is often a customer mid-decision.
Read your robots.txt now and look for any rule matching Claude agents, including wildcards above them. Name each bot separately and give each its own rule, the same per-bot pattern shown in our robots.txt walkthrough. Then verify against your server logs that the agents you allowed are actually fetching.
Why there is no magic AI sentence
There are 2 ways your business ends up in a ChatGPT answer, and neither has a slot for a self-declared ranking claim. The first is training: the model learned about you from crawled data, blended with everything else it read, on OpenAI's schedule, not yours. Adding a sentence today does not re-teach the model this week. The second is retrieval: ChatGPT search runs live queries against an index, pulls pages that answer the question, and cites them. Retrieval ranks pages on whether they answer the query, not on how confidently they describe themselves. An LLM does not have a keywords meta tag. We spent years telling owners that Google ignores self-praise; the machines changed, the rule did not.
There is also a simpler way to see it. If one sentence moved rankings, every competitor in your market would paste the same sentence within a month, and you would all be “the leading provider.” A signal everyone can copy for free is not a signal.
The machine listens to what other people say about you
In December 2025, Ahrefs studied 75,000 brands to find which factors track with showing up in AI answers. Branded web mentions correlated with AI visibility at 0.66 to 0.71 across platforms. Backlinks, the currency of the last two decades of SEO, came in around 0.19. In the study's words: "YouTube mentions show the strongest correlation with AI visibility (~0.737), outperforming every other factor across ChatGPT, AI Mode, and AI Overviews" (ahrefs.com).
Correlation is not a recipe, and Ahrefs says as much. But the direction is useful, and for an owner without a link-building budget it is oddly good news. What moves the needle is being named, accurately, in places machines read: review sites, local press, industry lists, YouTube. And the mentions have to agree, with each other and with your own site. A machine that finds three versions of your name and two of your address trusts none of them.
Check all five in one scan
You can walk through every reason above by hand, or you can check your AI citation status free to see which engines are citing you and which are not. When you are ready to understand the fundamentals, start with what is AEO, or paste your link into AuditLamp and let us check the five causes at once. We read your site the way the engines do, then tell you whether the retrieval bots can reach you, whether your answer survives without JavaScript, whether you look indexable, and how quotable your top passage is. We print the failures in fix order, in plain language, so you know exactly what to change first.