How to get cited by AI: a cross-engine checklist
AI answer engines cite pages they are allowed to reach, can find through the index they search, and can quote without guessing. Getting cited across Google's AI features, ChatGPT, Perplexity, Gemini, Copilot and Claude comes down to three jobs: open the door to each engine's search agent, publish something worth citing that other pages lack, and make sure the third-party pages those engines already cite describe you accurately.
The on-page writing habits (answer first, question headings, schema, freshness) are covered in our citation playbook on the blog. This page covers what that post leaves out: the full crawler picture, the kinds of assets that attract citations, and the off-site work that shapes what AI says about you when it does not cite you at all.
Open every door: the crawler matrix
| Engine | Agent that must reach you | Index behind the answers | Training agent, controlled separately |
|---|---|---|---|
| Google AI Overviews, AI Mode | Googlebot | Google Search | Google-Extended (no effect on Search) |
| Gemini app | Googlebot, plus Google-Extended allowed | Google Search | Google-Extended (same token) |
| ChatGPT search | OAI-SearchBot, ChatGPT-User | OpenAI's crawl plus search partners | GPTBot |
| Perplexity | PerplexityBot, Perplexity-User | Perplexity's own index | None named |
| Microsoft Copilot | Bingbot | Bing | None named |
| Claude | Claude-SearchBot, Claude-User | Anthropic's crawl, Brave Search | ClaudeBot |
Check three places for each row: robots.txt, your CDN or firewall, and any security plugin that rate-limits or challenges bots. A block at any of the three stops the engine even if the other two allow it. The per-engine pages go into detail, starting with ChatGPT and Perplexity, and RankWave AI's crawler directory lists every agent's token.
Publish assets worth citing
When ten pages give the same advice, an engine has no particular reason to cite yours. Give it one. Assets that tend to earn citations are ones that contain information nobody else has published:
- Original data. Prices you charge and when they changed, results from your own tests, anonymized numbers from your own customers, a yearly survey. A roofing company publishing its real average repair costs by roof type has a citable page; a generic "roof repair cost guide" does not.
- Definitive references. Compatibility charts, size guides, specification tables, glossaries for your niche with precise definitions.
- Tools with a visible method. A calculator is more citable when the page explains the formula and assumptions in text beside it.
- Clear policies. Returns, warranties, eligibility rules and service areas stated in full, so answers about your business quote you rather than guess.
- Named expertise. Commentary from a person with real credentials, identified by name and role on the page.
Mark the date on anything that changes, and cite your own sources. An engine checking a claim can follow the trail.
Get mentioned where AI already looks
Answers often cite third-party pages about you, or about your category, instead of your own site. Run your prompt panel and collect every third-party URL the engines cite for your topics: review platforms, "best of" roundups, directories, comparison sites, forums, Wikipedia. Then work through the list:
- Correct your listings. Wrong hours, old pricing or a missing category on a directory or review site will be repeated in answers.
- Offer facts to roundup authors. If a well-cited "best X" article omits you or has stale details, send the editor accurate, checkable information. Do not pay for undisclosed placement.
- Earn reviews honestly on the platforms that appear in answers for your category.
- Take part in communities where your customers ask questions, as a named representative, answering the question rather than posting links.
This is slow work, and none of it can be automated without looking like spam. It is also the part of AI visibility that competitors most often skip.
Keep your facts consistent everywhere
Models combine what they retrieve with what they learned earlier, so contradictions across the web show up in answers. Pick one version of your business name, description, locations and key facts, and use it on your site, your Business Profile, social profiles and directories. On the site itself, Organization schema with sameAs links ties those profiles together; Hydrogen SEO builds that list from its Schema and Social screens, and the JSON-LD validator checks the output.
Then measure. The measurement page covers the prompt panel and referral tracking. Expect slow movement: engines recrawl on their own schedules, answers vary between runs, and established sources hold their places. You are improving your odds across many answers, not securing any single one.
Common questions
How long until AI tools cite a new page?
It depends on when each engine recrawls the page and whether it ranks for the related searches. Some engines read pages live, others rely on their index, so allow several weeks before judging.
Does paying for listicle placements help?
Paid placements that are not disclosed break the trust that makes those pages worth citing, and may breach advertising rules. Offering accurate facts to editors is the safer route.
Can I ask ChatGPT to remember my site so it cites me?
No. Memory and instructions affect only your own account. Other users' answers depend on the engine's search results and the model, not on what you told it.