Citation-readiness audit: eligible is not cited

A citation-readiness audit is the card you run before you buy a GEO retainer. One URL. Four checks. A fail is useful. A vibe is not.

What is a citation-readiness audit? A written fail/pass on whether an answer engine could name you from that URL, and whether you actually checked. Eligibility is not a citation. A crawl is not your name in the answer.

How to get cited by ChatGPT is the week of work. Why ChatGPT does not name you is the diagnosis after a miss. This page is the card you fill in fifteen minutes.

The four checks

Run them in order. Stopping early is a result.

1. Door. Can OAI-SearchBot fetch the live URL? OpenAI’s crawler docs say that bot surfaces sites in ChatGPT Search. Sites that opt out “will not be shown in ChatGPT search answers,” though they can still appear as navigational links. The publishers FAQ is blunter: summaries and snippets require you not to block it. GPTBot is training. ChatGPT-User is a person-triggered fetch and does not decide Search inclusion. If a vendor slide collapses those names, throw the slide out.

2. First screen. Fetch the URL the way a crawler does, not the way a browser paints it. This site’s audit heuristic: one checkable claim in the server-delivered HTML, with a person or organisation sitting on that claim. OpenAI does not document a universal “first screen” citation rule. Treat client-only copy as unverified for the audit.

3. Named test. Turn Search on. Ask the question you actually want to win. A logged-in memory chat with no sources is the wrong product. One run is not a result.

4. Log. Write three columns: named, linked but unnamed, used without attribution. Absence is a door or discovery miss. A source link without your name is a weak entity cue. Apparent reuse without a name is not proof of copying.

If you cannot produce the four answers on paper, you do not have an audit. You have a feeling that the competitor is “better at GEO.”

What a fail means

Fail the door: fix robots.txt and wait. OpenAI says a Search-side robots change can take about 24 hours. Then retest.

Fail the first screen: move one dated, sourced sentence above the fold in the HTML. Attach a name. Do not add a fake FAQPage. Google ended FAQ rich results on 7 May 2026.

Fail the named test: you tested the wrong product, or the page is eligible and still not the capsule. That is the diagnosis page, not a schema plugin.

Fail the log: you cannot tell which miss you have. Start logging before you change copy.

Google’s generative reports in Search Console measure Google. They do not measure ChatGPT. llms.txt is ignored by Google Search. Do not treat a green GSC tile as a citation.

What this page is not

It is not a ranking guarantee. It is not a ChatGPT algorithm. It is not the nine-step how-to and not the full CITATION expansion.

The Findability Engine owns that expansion. The book is in production and not for sale. You can already fail a URL.

Write the four answers on a card. Run them on the URL you think should win. If the card is empty, you are not ready to complain that ChatGPT named someone else.