How to appear in ChatGPT search: a website checklist

To give your business a chance of appearing in ChatGPT search, make the relevant public pages accessible, clear and worth using as sources. You cannot buy a setting that guarantees a recommendation. The practical work is removing access problems, publishing reliable information and checking what actually happens.

This guide concerns ChatGPT using web search. A conversational answer that did not search the web is a different observation. Record that distinction before treating an answer as evidence about your website.

Start with the right crawler

OpenAI’s crawler documentation distinguishes search discovery from model training. A blanket block on every OpenAI agent can therefore conflict with a business’s search objectives.

OpenAI says OAI-SearchBot and GPTBot settings are independent. SearchBot permission enables access; it does not establish that a page was selected, cited or recommended. Robots rules are also not a way to protect confidential information.

A six-step website check

1. Choose the page that should answer the question. Start with one commercially important service and its strongest supporting guide. Write down the intended buyer, location and decision. “What does this service cost?” is more actionable than “make us visible everywhere”.
2. Open the final URL without signing in. Check that it loads on mobile, returns the intended page and does not depend on a cookie choice to reveal the main explanation. Note redirect chains, error messages and content hidden inside an image.
3. Review crawler and firewall access together. Ask the developer to check robots.txt, response headers, bot protection and access logs. Use OpenAI’s published verification information; a user-agent label alone can be copied. Fix a specific block rather than turning off website security.
4. Make the offer understandable on the page. State who it serves, what is included, geographic coverage, likely price drivers, exclusions and the next step. Link to relevant supporting information with descriptive text. Resolve conflicting contact details or outdated service descriptions.
5. Add evidence a buyer could inspect. Useful evidence includes a named author, a dated explanation of a process, an original worked example or a genuine project with permission to publish. Separate your experience from external research, and link to the source of a numerical claim.
6. Record the change and test again. Save the old problem, the fix, the affected URL and the publication date. Recheck accessibility before drawing conclusions from another search sample.

What useful content looks like

Consider a hypothetical engineering consultancy selling factory energy assessments. “Leading experts delivering innovative solutions” gives a potential customer little to compare. A stronger page explains the equipment reviewed, information needed before a visit, deliverables, exclusions and how the estimate changes with site size.

A supporting example could show how the consultancy assesses one anonymised process, with permission and a clear explanation of assumptions. It should not invent savings or imply a result applies to every factory. This improves the buying experience even when no AI service cites the page.

Choose questions from sales conversations, support requests and search data where available. Avoid publishing a separate article for every slight variation of the same question. One maintained explanation is easier to keep accurate than twenty nearly identical pages.

Run a small, repeatable search sample

Use a fixed set of realistic buyer questions, including at least some that do not name your company. Record the date, product, location context, wording and whether web search was used. Repeat the same questions in a later round; one favourable answer is only one observation.

For each answer, distinguish a brand mention from a link to your website. Save the actual cited URL and check whether it supports the answer. Mark a failed run as unavailable rather than “not mentioned”. OpenAI’s web-search guidance explains the search feature; it is not a promise that every conversation searches or that a particular business will appear.

Use the AI visibility tracking template to keep those observations comparable. For a wider review, follow the AI visibility audit checklist. Website visits and enquiries need a separate record, explained in our guide to measuring AI referrals and enquiries.

What to ask an AEO supplier

Ask for the pages they will improve, the access problem they have verified and the evidence they will report. Reject a proposal whose entire deliverable is allowing a crawler or generating more articles. If you want an agreed starting scope, our answer engine optimisation service begins with a documented visibility audit and prioritised recommendations.

Official guidance checked on 27 October 2026.