How to Get Your Business Cited by ChatGPT and AI Search

Generative engines retrieve passages, not pages, and they cite a handful of sources. Here is what actually decides whether your business is one of them.

A growing share of the questions your buyers used to type into Google now get asked of an assistant instead, and the assistant answers directly. It names a few sources. Most of the time it names three or four.

That is a much harsher funnel than a page of ten blue links, and it is the reason generative engine optimization has become its own body of work rather than a rebrand of SEO. This explains what actually decides whether you are one of the sources cited.

The unit of retrieval is the passage, not the page

This is the single most useful thing to understand, and it is where most existing content fails.

A classic search engine ranks documents. A generative engine retrieves chunks of text, assembles an answer from several of them, and attributes what it used. It is not asking which page is most authoritative on a topic. It is asking which specific passage answers this specific question cleanly enough to quote.

So a 3,000-word guide that covers a question thoroughly across six scattered paragraphs can lose to a 300-word section that answers it in one place. The long guide is better content by most editorial standards. It is worse content for retrieval, because there is no clean span to lift.

The practical change is structural rather than editorial. Ask a real question in a heading, answer it immediately and completely in the following paragraph, and only then expand. Do not build to the answer. Lead with it.

Models have to be able to resolve who you are

An assistant will not confidently attribute a claim to a business it cannot identify. Entity clarity is doing quiet work here.

That means consistent naming across your site and your profiles, structured data that states what the organization is and what it does, and an about page that reads like a factual record rather than a mission statement. Schema markup is not a ranking trick in this context. It is how a machine resolves ambiguity about which entity your content belongs to.

The failure case is a business whose name is slightly different in four places, whose services are described in marketing language that maps to no recognizable category, and whose site carries no structured data at all. There is nothing for a model to anchor to.

Specifics get cited and generalities do not

Content that asserts something concrete and checkable is far more likely to be used than content that is broadly true of everyone.

A sentence like "response time matters for local service businesses" is unusable, because it is a platitude the model already knows and does not need a source for. A sentence that explains the specific mechanism, with the conditions under which it holds and when it does not, is quotable, because it adds something.

This is why thin, hedged, keyword-padded content performs badly in AI search even when it ranks acceptably in classic search. There is nothing in it worth attributing. If you can delete your business name from a paragraph and it would be equally true of any competitor, that paragraph will not earn you a citation.

The crawler has to be able to read the page

A surprising number of sites fail at this step, and it is invisible from inside a browser.

If your content only exists after JavaScript runs, anything that does not execute scripts sees an empty shell. Check what your pages actually return by fetching the raw HTML rather than looking at the rendered view, because the rendered view will tell you everything is fine when it is not. The crawlers behind AI answers are a mix of systems with different capabilities, and the safe assumption is that some of them read raw HTML and nothing more.

Beyond that, check that your robots rules do not block the crawlers you want, and that the pages you most want cited are reachable through ordinary internal links rather than only through a sitemap. A page nothing links to is a page with no standing.

This does not replace technical SEO

There is a temptation to treat AI search as a separate channel with its own playbook. It is not, yet.

Generative engines still lean heavily on the classic index to decide what to retrieve. A site with crawl problems, duplicate content, or no authority does not get cited simply because its passages are well structured. The work sits on top of technical and content SEO rather than beside it, and if the foundation is weak that is where to start.

The reasonable framing is that everything that made content retrievable before still applies, and a new requirement has been added: the content has to be answerable in a liftable span, by an entity a machine can identify.

Does a citation actually bring you anything?

This deserves a straight answer, because the honest one is more complicated than the pitch usually allows.

Sometimes a citation produces a click and a visitor who arrives already half-convinced, because a trusted assistant just named you. That is a genuinely good visitor, often better than an equivalent organic click.

Sometimes the answer is complete enough that nobody clicks anything. The assistant has solved the problem, your name appeared next to the solution, and that is all you get. Whether that is worth anything depends on how your category works. For a considered purchase where people research a shortlist over weeks, being named repeatedly is worth a great deal even without the click. For an impulse or immediate-need purchase, it is worth considerably less than a visit.

There is also a volume reality worth stating plainly. For most businesses, AI search is still a small fraction of total search demand. The reason to work on it now is that the content and structural changes it requires are the same ones that help you in classic search, so you are not making a bet that only pays off in one scenario. Be suspicious of anyone presenting it as a channel you are currently losing large amounts of traffic to.

The sensible position is to treat it as a growing surface worth being present on, built with work that has independent value, rather than as an emergency.

How to tell whether any of this is working

You cannot check this in Search Console, which is the part that catches people out. Citation in an AI answer is not an impression and does not appear in your reporting.

The method that does work is unglamorous. Fix a set of questions your buyers actually ask, in their words. Run them through the major assistants and record who gets cited for each. Do the work. Run the same set again on a schedule and compare. It is a manual baseline rather than a dashboard, but it is a real measurement, and it is the only way to tell movement from wishful thinking.

Keep the question set stable. Changing the prompts between runs makes the comparison meaningless, which is the most common mistake in this kind of testing.

Where to start

If you want a short list: pick the ten questions that matter most commercially, find out who gets cited for them today, and then rework the pages that should own those answers so each question is answered cleanly in one place, with specifics, under a heading that asks the question.

Add the structured data if it is missing. Confirm the pages render without JavaScript. Then re-test.

That sequence is the substance of generative engine optimization, and most of it is work you would want done anyway.

If you would rather have someone run the baseline for you, tell us what your business does and we will show you what the assistants currently say about your category and who they credit.