How to Get Cited by ChatGPT: An Agency Playbook

Advice on how to get cited by ChatGPT is usually written for a brand owner with one website. An agency has a different problem: twenty client sites, a team that implements, and clients who want to know why a competitor is named and they are not. The tactics are the same. What changes is the order, the split of work, and how you check that it worked.

To get a client cited by ChatGPT, Gemini, Perplexity, and Google AI Overviews, work five levers:

Access: make sure the engines’ search crawlers can reach the site.
Entity signals: the same name, category, and location everywhere.
Page structure: a direct answer high on the page, clear headings, and FAQ blocks.
Schema: markup that matches what the page says.
Citation sources: a place on the directories, lists, and review sites the engines already cite, with the reviews and profile signals that come with them.

Then re-test the same buyer prompts every month and adjust. These levers map to the five gap areas used in an audit (page answers, structured data, entity consistency, third-party citation sources, and reviews and profile signals), with access as the precondition. The risk for agencies is not the tactics. It is giving each account a different prompt set and a different report, so no one can tell which fix worked.

How the work splits across client accounts

Every client gets the same structure: a fixed set of 10–20 buyer prompts, run on the same four engines, with every result dated. What makes it work at agency scale is splitting planning from implementation:

Planning: building the prompt set, running the engines, diagnosing the gaps, and ranking the fixes for each client.
Implementation: page rewrites, schema, profile corrections, and outreach to citation sources, done by the agency’s writers, developers, and SEO team.
Checking: re-testing every month and adjusting the plan.

Keep the same prompt set template, gap categories, and report format across accounts. The levers below are the implementation team’s task list.

Mention vs citation: what you are actually trying to win

AI answers reward a business in two different ways:

A mention is the business’s name in the answer text. “For emergency plumbing in Parramatta, options include…” is a mention.
A citation is the business’s website linked as a source. It shows the engine used the client’s own page to build the answer, and it is the one that can send a click.

The two do not always come together. An engine can name a business it learned about from a directory or a review site without citing the business’s site at all. It can also cite a client’s page to explain a topic and then recommend a competitor. Track both, prompt by prompt and engine by engine. They are linked: off-site presence can make it more likely that the engine treats the business as a real option in the category, and on-page answers give it a passage worth citing. Off-site presence may need to be in place before on-page answers earn citations.

How AI assistants choose sources

They search before they answer. For many questions, assistants run web searches and build the answer from what they retrieve. Google says AI Overviews and AI Mode may use a “query fan-out” technique: issuing multiple related searches across subtopics and data sources to develop a response. A buyer who asks one question may trigger several searches behind it, and your client’s page has to be found for one of them.

They can use more than one search source. OpenAI says ChatGPT search may share queries with third-party search providers such as Bing. A client that is invisible in Bing may have a gap in ChatGPT, even if its Google rankings are fine.

They lean on third-party lists. For “best” and “near me” prompts, engines often draw on directories, review platforms, best-of articles, and forum threads rather than any single business website. If the client is not on the lists the engines read, it is hard for the engines to recommend it.

They need to be allowed in. Check access first, before any other work:

OAI-SearchBot surfaces websites in ChatGPT’s search features. OpenAI says sites that opt out will not be shown in ChatGPT search answers.
PerplexityBot surfaces and links websites in Perplexity’s results.
Google: a page must be indexed and eligible to show with a snippet in Search to appear as a link in AI Overviews or AI Mode.
Bing: because ChatGPT search can draw on Bing, confirm the client’s key pages are indexed there.

The practical conclusion: getting cited is search visibility, entity clarity, quotable pages, and third-party presence, all at once. No single change does it.

Entity signals: consistent name, category, and location

Engines can draw on many sources to work out what a business is. When those sources disagree, the engine has less reason to be confident, and it may name someone else. The failure is usually narrow: a client listed under the wrong primary category, or with an old service area, can still be described when someone asks about it by name, yet drop out of the category prompts that matter, such as “best plumber in [suburb].”

Check these across the client’s website, Google Business Profile, Bing Places, the main directories in its market, and its social profiles:

Business name. Use one form of the name, legal or trading, and use it everywhere.
Primary category. The same core category. A business listed as “plumber” in one place and “gas fitter” in another can read as two different businesses.
Location and service area. The same address or service area, with no old suburbs or past addresses left on stale listings.
Core services. The services the client wants to be known for, described the same way on the site and on the profiles.

This is unglamorous work, and it needs no new content.

Page structure: direct answer blocks, clear headings, and FAQ

An engine that retrieves a page still has to find a passage it can use. Pages that bury the answer under an introduction, a company history, and a sales pitch are hard to quote.

For every page that should win a buyer prompt:

Open with the answer. The first two to three sentences should answer the question the page targets, in plain language, with the specifics: the service, the area, and what the client actually does.
Use headings that match how buyers ask. “How much does blocked drain repair cost in Sydney?” is a better heading than “Our Pricing Philosophy.”
Add a short FAQ. Real buyer questions with direct answers, drawn from what customers actually ask the client and what shows up in People Also Ask.
State quotable facts. Service areas, the process, what is included, and what is not. Specifics are quotable; slogans are not.
Keep one obvious page per question. Three thin pages on the same service compete with each other and give the engine no clear source.

Schema that supports entity clarity

Schema does not get a page cited on its own, and Google says no special schema.org markup is required to appear in AI Overviews or AI Mode. What schema does is remove ambiguity about who the business is and what the page says. For example, LocalBusiness markup whose areaServed lists the same suburbs as the client’s service area pages tells a crawler the same thing the page and the profile do.

Organization or LocalBusiness markup with the same name, address, service area, and contact details used in the entity check above.
FAQ content on pages with real buyer questions, with markup as an optional extra.
Service details that match the services described on the page.

The rule that matters most: structured data must match the visible text on the page. Google lists this among the fundamentals that still apply to AI features. Markup that claims something the page does not say is a liability, not a signal.

Citation sources: directories, best-of lists, reviews, and forums

This lever is easy to under-invest in, and for local service businesses it can hold the biggest gaps.

Start with evidence, not a generic directory list. Run the client’s buyer prompts and record which sources each engine cites. Then:

List every source that appears more than once across the client’s prompts: directories, review platforms, best-of articles, local publications, and forum or Reddit threads.
Check whether the client is on each one, and whether the listing is complete and accurate.
Fix the listings the client already has before chasing new ones. An incomplete profile on a source the engines already trust is the cheapest gap to close. Cheap is not the same as fast: a corrected listing may not show up in answers until the engines retrieve the updated page, so expect it to surface over the following re-tests.
Earn places on the lists the client is missing from. For best-of articles, that usually means contacting the publisher with a reason to include the client, not paying for a placement.
Build review depth on the platforms the engines cite, through the client’s normal customer follow-up.

Do not fabricate any of it. Fake reviews, invented testimonials, and planted forum posts breach platform rules and put both the client and the agency at risk.

Local service businesses: the extra levers

Local service businesses in Australia and Canada have a few levers that national brands do not.

Google Business Profile completeness. Categories, services, service areas, hours, and photos can all shape how Google understands the business. Keep it in line with the website.
Bing Places. Because ChatGPT search can draw on Bing, a claimed and accurate Bing Places listing is worth checking.
Market-specific directories and review sites. Australia and Canada each have their own trade marketplaces and review platforms, such as hipages and ProductReview in Australia or HomeStars in Canada. Do not assume which ones matter: check which local sources the engines actually cite for that client’s category and market, and start there.
Suburb and city specifics. Prompts are often local (“in Parramatta,” “near Kitchener”), so pages should name the areas the client actually serves. Do not create thin pages for suburbs the client does not serve.
Licences and associations. Where the client holds a trade licence or belongs to an industry association, state it clearly on the site, because it is a specific, checkable fact an engine can use. Only state what the client can show.

Re-test and adjust

Every change above is a hypothesis until the prompts are re-run. Re-test the same buyer prompt set every month, on each engine, from a clean test profile, and compare against the baseline.

Look at the trend across the set, not one answer. AI answers vary between runs.
Match changes to fixes. If entity corrections went live in week two and the client started appearing in Gemini answers in the next re-test, note it. If nothing moved, look again at the cited-sources column.
Expect engines to move differently. A fix that moves Perplexity may not move ChatGPT, because each engine can draw on different sources.

For the full measurement setup and how to report it to clients, see the guide to measuring AI visibility.

FAQ

How do you get a client cited by ChatGPT?

Make the client’s business consistent across its website and profiles, open key pages with a direct answer to the buyer’s question, add schema that matches the page, make sure crawlers such as OAI-SearchBot are not blocked, and get listed on the directories and review sites ChatGPT already cites for the category.

Does ChatGPT use Google or Bing?

OpenAI says ChatGPT search may share queries with third-party search providers such as Bing. That is why Bing indexing and a Bing Places listing are worth checking for any client you want cited.

How long does it take to get cited by ChatGPT?

There is no fixed timeline. Entity and page fixes can show up in the next monthly re-test or take several months, and off-site work such as earning places on best-of lists usually takes longer. Report progress as a trend across the prompt set.

Do I need schema to be cited by AI?

Not on its own. Google says no special schema is needed for AI Overviews or AI Mode. Schema helps by making the business’s identity and the page’s content unambiguous, and it must match the visible text.

Can an agency guarantee ChatGPT citations?

No. No one controls what an AI engine says, and answers change between runs. An agency can fix the gaps that make a client hard to cite and report the trend.

Your team implements; SearchKata sets the plan

If your clients are asking how to get cited by ChatGPT, the work splits cleanly. Your team implements. SearchKata builds the prompt set, diagnoses the gaps, writes the plan, and checks the work with a monthly re-test. See the SearchKata method for how it runs, or learn more about white-label AEO for SEO agencies.