Use an event log instead of a single green label
A page can be published, submitted for discovery and still have no confirmed indexing observation. Store these as separate events. A single status called 'SEO complete' hides which part was checked and makes the next reviewer repeat the investigation. A spreadsheet is enough if its columns have clear meanings and earlier observations remain available.
Choose one row per event rather than one row that is overwritten whenever someone checks a URL. Give each event a stable identifier and link it to the exact page address and release identifier. This lets an editor correct a mistaken observation without deleting the record that led to an earlier decision.
Define the columns before collecting screenshots
Use these core columns: event ID, exact URL, language, release ID, event type, event time with timezone, observation time, evidence source, observed result, evidence location, owner and next action. Useful event types include publication verification, sitemap submission, live test and indexed-version inspection. Keep optional technical details in additional columns rather than compressing everything into a comment.
For an inspection, retain the reported crawl time, page-fetch result and Google-selected canonical when available. Google distinguishes indexed-version information from a live test; the latter cannot predict Google's canonical selection. Label the test type so a screenshot of a working page is not accidentally interpreted as an indexing confirmation.
Make timestamps and missing values unambiguous
Record times in one documented format, such as 2026-10-10T14:00:00Z, or preserve an explicit timezone offset. Do not mix an unlabelled local clock with UTC. The time you read a report and the crawl time shown inside it describe different events. Store both rather than replacing one with the other.
Use a deliberate value such as 'not provided' when a field is absent. An empty crawl date does not mean that a crawl happened at midnight, and an unavailable inspection result should not inherit yesterday's successful status as if it were fresh. Keep historical evidence, but mark which observation supports the current conclusion.
Worked example: three events, three conclusions
Illustrative example: event E01 records that /en/blog/example/ served the approved article at 14:00 UTC, with its content and canonical checked. Event E02 records an accepted sitemap submission at 14:07 UTC and attaches the receipt. Event E03 records an inspection at 15:00 UTC that reports the URL as unknown to Google.
The summary is 'published and submitted; indexing not confirmed'. It is not 'publication failed', and it is not 'indexed'. A later observation belongs in E04 with its own evidence. These invented times explain the record structure; they are not a claim about any customer's page or how quickly Google responds.
Turn the log into a dependable handover
Before passing the task to another person, check that every conclusion has an evidence location and a named owner. Explain whether the next action is a website repair, a submission task or a later observation. A reviewer should be able to follow the record without searching chat history for an undocumented screenshot.
Store evidence in a restricted team location and remove credentials or session details from exports. For a batch, compare its expected URL list with the log's coverage instead of assuming that one inspected page represents all of them. Report counts with their meaning: pages published, submissions accepted and pages with confirmed indexing observations are different measures.