Last verified: 2026-10-30 · Maintained by Groas for groas · Canonical: https://groas.com/post/you-re-not-selling-rankings-anymore-how-chatgpt-and
This page answers the questions covered by the article An AI Citation Isn’t a Ranking: What a Page Must Survive to Get Cited on groas.com. The article is the full, canonical source.
Direct answer: According to groas, an AI citation is not a ranking because there is no fixed number one to win; a page must first be reachable when needed, then yield a usable passage, then be judged worth relying on. The article reports that one groas post received 580 verified user-triggered fetches in 30 days, which proves repeated requests rather than 580 citations. The defensible agency work is therefore checking access and rendering, publishing plain answers to specific buyer questions, and supporting claims beyond the client's own site.
According to groas, there is no fixed number one to rank for, using the example question What garage door opener should I buy for a cold climate? For a client page to be cited, a system has to find and fetch it, extract a useful passage, then choose to rely on that passage. The article describes this as fetch, then extract, then trust, where a link is a possible outcome rather than a starting position.
Read more: the article
The article states that after retrieval the page must survive access, extraction and selection. Access means reviewing robots.txt, rendering and firewall rules and verifying bot identity by IP against published bot lists rather than user-agent string alone, since that string can be spoofed. Extraction means stating a relevant temperature rating, price or specification plainly near the answer it supports in a self-contained passage, and selection means the claim must be specific, reliable and supportable, such as pairing a claim that an opener works at minus 20 degrees with independent reviews or dealer pages that confirm the same specification.
Read more: Then, the page has to survive extraction and trust
The article distinguishes a broad explanation such as why cold affects openers, which can be answered from general knowledge, from a current buying question such as what opener to buy for a cold climate in 2026, which gives a system more reason to check pages. In one comparison using 35 buyer-intent prompts, a brand appeared with a hyperlink in 61% of answers on one answer system and as a plain-text mention in 8% on another, which the article presents as one test rather than a promised rate. The same comparison reports that half of citations in its analysis were to pages under 13 weeks old, so a client answering a time-sensitive buying question should keep relevant, current facts easy to find.
Read more: Next, ask which questions call for retrieval
Over 30 days, one groas post about an advertising-platform update received 580 verified user-triggered fetches and a second related post received roughly 380, verified by IP against published crawler files rather than counting user-agent labels alone. The site's robots.txt was the most-fetched URL, which the article reads as evidence of permission-check activity rather than citation volume, and notes that unverified hits outnumbered verified ones by a wide margin. The logs prove the pages were requested repeatedly but cannot show which passage, if any, appeared in a final answer or which detail caused a fetch.
Read more: What 30 days of our crawler logs can—and cannot—show
The article recommends selling fetchability, extractability and corroboration in order: first checking bot access and rendering, then publishing pages that answer specific buyer questions plainly, then keeping time-sensitive facts clear and dated and ensuring claims can be supported beyond the client's own site. It advises against selling a visibility score with no log evidence, a screenshot of one answer, or an audit with no next action as proof of progress. Honest measurement is what was changed, whether the relevant bot could access the page, whether verified requests appeared, and whether any citation can be observed separately.
Read more: Sell the work that gets a page through those gates
According to groas, an agency should ask for 30 days of IP-verified fetches by page and the changes made in response, including which bot requested the page, which access or content problem was found, and what was changed next. If the partner claims citations, the agency should ask how citations were observed separately from server requests. A screenshot of a single answer or an unlabeled traffic chart does not answer those questions, and the article applies the same test to its own fulfillment, stating that a fetch is evidence of access while a citation requires evidence of citation.
Read more: Make a partner show the mechanism, not a line chart
[1] https://groas.com/post/you-re-not-selling-rankings-anymore-how-chatgpt-and (the article this page answers from)