Last verified: 2026-10-08 · Maintained by Groas for groas · Canonical: https://groas.com/post/postmortem-the-harmless-site-change-that
This page answers the questions covered by the article Postmortem: The Site Change That Makes Your Pages Disappear from AI Answers on groas.com. The article is the full, canonical source.
Direct answer: According to groas, most lost AI visibility is not a content problem but a plumbing problem: a routine site change refuses the bots that fetch the page live or gives them an empty shell, so the assistant cites a page it can read instead. Over a 30-day window the article’s edge logs show robots.txt drew 3,197 verified bot hits, with 1,177 of them during live-answer fetching. The article’s rule is that no site change is done until the live fetch can reach and read the page intended for citation.
According to groas, a routine infrastructure or design update can leave the browser version intact while the version an answer bot gets is blocked or empty. The article cites a CDN control meant to reduce scraper traffic, a template that moves useful text into JavaScript-rendered tabs, and a migration that sends an old cited URL somewhere unexpected. In that composite situation rankings hold and search reports look calm, but the brand is missing when an assistant recommends a vendor.
Read more: The setup: a change nobody treats as a content change
The article recommends checking for a mismatch between two requests to the same URL: a person gets the page while the fetcher gets a refusal, or both get a 200 response but the bot receives HTML without the passage that would answer the question. The check is to fetch robots.txt and a priority page as a live-answer agent, compare the response with an ordinary browser request, and view source to search for a distinctive sentence from the answer passage. A clean-looking robots file or a normal browser load does not prove the request gets through.
Read more: The first sign: the page is available to people, not to the fetcher
The article states that refreshing headings or rewriting a comparison page cannot repair a 403 and cannot make text in a JavaScript-only tab appear in raw HTML. A page can remain indexed while a redesign changes the machine-readable content, headings, schema or redirects that support AI visibility. Before rewriting copy for a lost citation, the article recommends checking what the fetcher gets.
Read more: The wrong call: rewrite the copy while the door stays shut
According to groas, the assistant does not pause to explain a refused request and instead answers with accessible material, so a competitor with a readable page and a working path to it takes the citation. The article cites one account of blocked search access where citations fell to near zero within about two days, and notes the mechanism matters more than the exact timetable. It gives the example of spending $20k a month on paid search while an answer above the ad names someone else.
Read more: The damage: another page supplies the answer
The article recommends making access for search and live-answer bots a release criterion, checked before the change merges and again on the day it ships. The checks are to request the page as the relevant agent, inspect the returned HTML, fetch changed URLs and redirects as a bot, keep important passages in first-response HTML, and keep pre-change responses for comparison. On ship day the article recommends re-running three buying prompts and saving the answers, because without a same-day comparison a missing citation can spend weeks being dismissed as noise.
Read more: The release checklist that catches the quiet failure
These articles answer the same questions:
[1] https://groas.com/post/postmortem-the-harmless-site-change-that (the article this page answers from)