How to get your business cited in AI Overviews and ChatGPT
To be cited in AI answers, a business needs pages that can be crawled and indexed, passages that answer a question on their own, and facts about the company that agree across the web. No special markup is required, and results are checked by running a fixed set of prompts.
Key points
- AI answers retrieve indexed passages first, then generate an answer that cites them.
- Google has said no special markup is required to appear in AI Overviews.
- Quotable pages give the answer first in self-contained passages with specific facts.
- Track visibility with a fixed prompt set run on a schedule and logged.
How do AI answers choose their sources?
AI answers that cite sources work in two stages: retrieval, then generation. In retrieval the system searches an index for pages and passages that match the question. In generation a language model writes the answer from what was retrieved and attaches links.
Google has described a technique it calls query fan-out in its AI search features, where one question is split into several related searches that run at the same time. A person asking about heat pumps in Austin may trigger sub-searches for cost, installers, rebates and sizing. A page that answers one of those sub-questions well can be retrieved even when it does not match the original wording.
AI Overviews are AI-generated summaries in Google results with links to sources. Google has said no special markup is required to appear in them. To be shown as a supporting link, a page needs to be indexed and eligible to appear in Search with a snippet.
What makes a page quotable?
A quotable page states its answer in a passage that still makes sense when separated from everything around it.
- Answer first. The first sentence under a question heading gives the answer. Under How long does a heat pump installation take, the first sentence states a typical range for a standard replacement and names what changes it.
- Self-contained passages. Name the subject in the sentence instead of writing it or this. A paragraph that begins It usually takes two days means nothing when lifted out.
- Specific facts. Name the system types, ratings such as SEER2 and the area served, and attribute any figure to a source you can cite.
- One question per section. A heading and a passage that answer one thing are easier to retrieve than a long page that covers six.
- Visible authorship and update dates. Readers and systems can see who is responsible for the claim.
A weak passage reads: It depends on several factors and our team will assess them. A quotable passage reads: A heat pump replacement for a typical home usually takes one to two days, and the time depends on whether the ductwork needs repair. The second names the subject, answers the question and states what changes the answer. The numbers in any real page must come from the installer's own jobs.
Why does entity consistency across the web matter?
A generative system builds its picture of a business from several sources, so contradictions weaken confidence in any of them.
The Austin installer should present the same name, address, phone number, service area, license and service list on its website, its Google Business Profile, its directory listings and its review profiles. If one source says it serves Round Rock and another does not, the system has to guess. Linking the site to its own profiles with the sameAs property in structured data helps connect the pages to the one company they describe.
This is the entity work described in semantic SEO, applied beyond the website.
What does structured data do for AI citations?
Structured data labels what a page says in a format machines read directly. It does not guarantee a citation.
Because Google has said no special markup is required for AI Overviews, markup should be used for what it does well: stating the page type, the organization behind it and the facts on it. Useful types for the example installer include Organization or HVACBusiness for the company, Service for each offering, and Article with an author for guides. The markup must match the visible text. Markup that claims more than the page says is a problem, not an advantage.
It also has a smaller role than a clear page. If the passage is vague, markup cannot make it quotable.
Which actions matter and why?
| Action | Why it matters | Example for the installer |
|---|---|---|
| Keep pages indexable | Retrieval draws on indexed pages | No stray noindex tag on service pages |
| Write answer-first passages | Generation quotes passages, not pages | One sentence answer under each question heading |
| Cover related sub-questions | Query fan-out searches several angles | Separate pages for cost, sizing, rebates and comparison |
| Keep business facts identical | Systems cross-check sources | Same name, phone and service area everywhere |
| Allow the right crawlers | Blocked crawlers cannot retrieve pages | Review robots.txt rules for search bots |
| Add matching schema | Labels the page type and publisher | Service and Organization markup |
OpenAI documents separate crawlers for search and for model training, so a robots.txt rule can allow one and not the other. Review the current documentation before changing rules.
How do you measure AI visibility?
Measure AI visibility by running a fixed set of prompts on a schedule and logging the result.
- Write the prompt set. Include 20 to 30 prompts a buyer would ask, such as who installs heat pumps in Austin and what a heat pump installation involves.
- Run them on a schedule. Use the same systems and wording each time, in a logged-out or clean session.
- Record the outcome. Note whether the business is mentioned, whether it is cited with a link, which competitors appear and whether the facts are right.
- Compare over time. Track changes per prompt after each page update.
Answers vary from run to run, so look at repeated patterns, not one response. The AI visibility audit performs this check.
What does not work?
- Adding llms.txt. Google has said it does not use it.
- Stuffing pages with questions that the business cannot answer well.
- Publishing hundreds of near-identical pages. Google's spam policies cover scaled content abuse, which is generating many pages mainly to manipulate rankings.
- Treating AI citation as a trick separate from search. Read SEO vs AEO vs GEO for how they connect.
The wider service is described under AI search optimization.
Frequently asked questions
Do I need to rank first in Google to be cited?
A page does not need the top position, but it must be indexed and eligible to appear in Search. Strong rankings tend to support retrieval, because the same index and signals feed both.
Can I pay to be included in AI Overviews?
Google has not offered a way to buy a place in organic AI Overviews. Ads are labeled separately.
Will schema markup get me into ChatGPT answers?
There is no evidence that markup alone gets a page cited. It labels content. The content has to be retrievable and worth quoting.
Does blocking AI crawlers affect my Google rankings?
Rules for AI-specific crawlers are separate from Googlebot, but check each crawler in its own documentation. Blocking Googlebot itself removes pages from Search.