How do you get cited by ChatGPT?

Begin with the questions your customers actually ask. Publish a useful answer, show what supports it and check whether the relevant search tools can reach the page. Then record the answers and sources the platform actually returns. You can improve the material and test the result; choosing a citation remains the platform’s decision.

How should you structure an answer?

Use a heading that names the question, give a direct answer and put the supporting detail nearby. Explain the scope, conditions and exceptions where the reader needs them. A paragraph should remain accurate when read on its own, without losing a qualification several sections away.

This is editorial practice, not a disclosed universal ChatGPT selection algorithm. A claim that every system splits every page into the same kind of chunks goes further than the evidence. Focus on whether a customer can understand and verify the answer.

For example, a service page can explain what a review covers, what it excludes and what the customer receives. If it gives a turnaround time, keep the starting condition and working-day qualification beside it. A short answer that leaves out those limits is not an improvement.

What evidence should the page carry?

Give factual claims a source and keep its date, population and conditions visible when they affect the meaning. If you publish your own data, explain how it was collected and what it cannot show. A number with no method is difficult to evaluate even if it looks specific.

The GEO research paper (2024) reports visibility gains up to 40% in benchmark evaluations and also includes a limited Perplexity test. These are context-dependent findings, not a projected gain for your site or a universal ceiling.

Do not turn a result from one query set, engine or period into a promise for another. For your own site, define the questions and record the observed answers before deciding what to rewrite. An agency’s correlation or a research result is not a substitute for that baseline.

Can the search system reach the page?

OpenAI documents OAI-SearchBot for search and GPTBot for potential model training, with independent controls. Allowing search access and allowing training are different decisions.

Have the responsible maintainer check the relevant access rules and whether the actual page can be retrieved through the hosting and security configuration. A rule in robots.txt is not the same evidence as a successful request. Do not change training preferences merely on the assumption that they are required for search citations.

Access is one thing to check, rather than a promise of selection. Keep the content useful for people as well: clear navigation, readable answers and accurate links make the page easier to inspect. The diagnosis should identify a concrete access or content problem, not claim knowledge of a hidden ranking rule.

How should you measure changes?

Agree a set of customer questions and the language and context in which you will ask them.
Record the platform, date, relevant settings and the answer returned.
Separate a mention of the business from a citation linking to one of its pages.
Save the cited address and check whether the source actually supports the statement.
Repeat the defined exercise after a change and keep the limits of the comparison visible.

A single answer is an observation, not a stable position you own. Track referral visits and enquiries separately from citations, and keep the question set comparable. Otherwise a changed question or different platform setting can look like the effect of your page edit.

Questions

How can we see which pages are being cited?

Run the agreed customer questions, save the returned answers and record the cited addresses. Distinguish your own page from a third-party mention of your business. A repeatable log gives you something to compare instead of relying on memory or one favourable answer.

Should we publish more pages or improve the existing ones?

Look for the actual gap. A missing answer may need a new page; an unclear or unsupported answer may need an edit. Do not add filler or invented data to create volume. Start with the questions that matter to the service and the evidence you can provide.

Will adding schema markup guarantee a citation?

No. Keep structured data accurate and consistent with the visible page, using it for its defined purpose. A claim that markup alone guarantees citations needs evidence beyond the presence of the code. Compare actual outcomes without treating markup as a universal shortcut.

Sources

Measure who cites you before changing the content

AI search visibility starts with your customers’ real questions, run on five platforms, recording who gets cited. From there you can see which passage to rewrite first.

Ask about AI search visibility

or write to contact@ergostrat.com