Keep three records for the same passage
Your publication record should show what changed and when it became publicly available. Save the exact URL and released HTML, the relevant text before and after the edit, and the release time with its timezone. A file hash can help identify the saved artifact later; it says nothing about whether Google received it.
Google's URL Inspection indexed result describes its recorded indexed version; the live test fetches the current page. Record the last-crawl time, Google-selected canonical and, when available, the HTML under View crawled page. The canonical is Google's chosen representative URL among similar pages. A live test cannot establish that choice. Google reports last crawl in your local time. Google's URL Inspection documentation
Finally, preserve the AI question and full answer with the collection time, engine and exact cited URLs. Keep the run settings used for the comparison. A citation to the right URL still needs to be read: does the answer reproduce the revised rule, describe the old rule, or say something the page never supported?
These records answer different questions. The release copy establishes your edit; the crawl evidence can show what Google fetched; the saved answer establishes what the assistant returned. None supplies a missing record from another step.
Compare the words, then the timestamps
Find the changed passage in the saved crawled HTML, when that HTML is available. Check nearby qualifications as well. A new headline with an old eligibility table is not a match to the complete revision you intended to test.
If the last crawl predates release, that record cannot verify the later edit. If the timestamp follows release but the passage is unavailable, record the version as unverified. Timing alone does not show which words arrived: a deployment can be partial, or a cache can serve an earlier response. Those are possibilities to investigate, not diagnoses.
Google can select a different canonical from the one you declare. Canonical annotations and redirects are signals; sitemap inclusion is a weaker signal. If the selected URL differs, compare the relevant content there before treating the requested URL as the representative copy. Google's canonical guidance
Use an evidence label that survives the next review
The following is an invented service-page example, not a Search Console screenshot or a measured result. All times use the same timezone. The revision changes eligibility from “existing customers only” to “new and existing customers.”
Scroll horizontally to read all columns. Use the focused table region with the arrow keys.
| Saved evidence | What to record for the proposed retest | What remains unresolved |
|---|---|---|
| Release Monday 09:00; crawl Sunday 23:00 contains the old eligibility rule | Earlier passage observed in the saved crawl | Whether Google later fetched the revision |
| Release Monday 09:00; crawl Monday 11:00; crawled HTML unavailable | Post-release crawl reported; passage version unverified | Which eligibility text was fetched |
| Release Monday 09:00; crawl Monday 11:00 contains the new rule and its qualifications | Revised passage observed in the saved crawl | Whether the answer system retrieved, used or cited it |
| Current live test contains the new rule; saved crawl still has the old rule | Current fetch verified; saved crawl remains earlier | When the indexed evidence will reflect the revision |
Treat these as evidence labels, not four scores to average. They let the next person see why a run can describe current answer behavior without establishing the effect of a particular edit.
Google says a successful live test does not guarantee indexing, and an indexed status does not guarantee appearance in results. Screenshots from URL Inspection are available for successful live tests, not the indexed version. Keep that distinction in captions if you save a screenshot as evidence. URL Inspection result limits
A submission receipt cannot close the version gap
If recrawling is appropriate, use the supported request route and keep the receipt separately. Google says crawling can take days to weeks, requests do not guarantee inclusion, and repeated requests do not make the process faster. There is no universal waiting period after which a revision can be assumed processed. Google's recrawl guidance
Likewise, a sitemap's lastmod describes a significant page update; it is not Google's crawl timestamp. Google asks for accurate, verifiable dates and treats sitemap submission as a hint. Retain your publication date without relabeling it “indexed on.” Google's sitemap guidance
State what the next measurement can answer
If the revised passage is still unverified, you can postpone a revision-specific evaluation or run an explicitly labeled snapshot of current answers. Choose deliberately. Do not describe the latter as evidence that the edit worked or failed.
If the revised passage is present in the saved crawl, the measurement still needs its own evidence. Record whether the answer cited the page, mentioned the company, recommended it, or did none of those. Read the cited passage and surrounding claim before assigning accuracy. An answer that cites your page while misstating eligibility is not a successful explanation of your service.
For Google AI Overviews and AI Mode, Google documents indexing and eligibility for a search snippet as requirements for appearing as a supporting link. Meeting those requirements does not guarantee selection. A Google crawl record also does not establish what another provider indexed or retrieved. Google's AI-feature eligibility guidance
Preserve the same questions and settings where a comparison is intended, and report any changes or unavailable answers separately. Even with a verified revision, an answer changing afterward does not by itself establish that your edit caused it. The question-set comparison guide covers that separate measurement decision; the citation audit covers broader access problems.