Research Guides
Did the SEO Fix Reach Visitors? Separate Deployment, Cache and Search Evidence
A corrected title in an editor is not yet a corrected public page. A corrected public page is not yet a refreshed search result. Mixing those milestones turns a simple verification task into repeated edits, unnecessary cache purges and vague claims that “Google is ignoring the fix.”
Use three separate questions: Was the intended change released? Which response reached this visitor? Has search processed the changed page? This manual worksheet helps a small team answer each question with evidence. It does not depend on an automated Colquhoun.AI scan or claim that a research integration is available.
Start with a small, observable change
Before checking anything, write the expected result exactly. “SEO updated” is too broad. “The public HTML title on /guide/ changes from X to Y, while the canonical remains Z” gives the reviewer something definite to compare.
Record the approved deployment identifier, target URL, expected title or other field, release time and rollback reference. If you cannot establish which release is active, mark deployment unverified. A cache explanation should not substitute for missing release evidence.
Keep customer sessions and private pages out of this exercise. Use a public URL and never paste cookies, authorization headers or personal account responses into a shared audit.
Treat the first public response as a specimen
Save the response body and headers from the same request, with the observation time and final URL. Record any redirects rather than silently treating the starting and final URLs as identical.
The Age header is measured in seconds and describes a response's age in proxy-cache terms; it is not the age of the article or the date when a human last edited it. A missing Age header does not establish the whole delivery path. Use it as one clue, not a verdict. MDN: Age
Also record Cache-Control, ETag, Last-Modified and any documented provider cache-status header when present. Preserve the exact values in your private evidence. Do not guess the meaning of a vendor-specific status from its name.
Read cache instructions literally
no-cache permits storage but requires validation before reuse; no-store tells caches not to store the response. They are not interchangeable. A request with Cache-Control: no-cache asks for validation, rather than proving that every cache was bypassed. Browser history restoration can also behave differently from a new network request. MDN: Cache-Control
For an authorized diagnostic check, compare an ordinary navigation with a fresh network request whose conditions you record. Do not immediately replace the site's cache policy. A broad “disable caching everywhere” change can create a separate performance problem without identifying why this page looked old.
A 304 Not Modified validation response allows a suitable stored response to be updated and reused. It is not a new page body you can inspect by itself. Record the stored representation and validator involved before interpreting that response as evidence about the intended edit. RFC 9111, section 4.3.3
Use a three-column acceptance record
| Checkpoint | Evidence to retain | What it does not prove |
|---|---|---|
| Intended release active | Deployment identifier and release record | Every visitor received it |
| Public response corrected | Timestamped body, headers, final URL and expected field | Google has recrawled it |
| Search processing observed | Dated Search Console observation for the relevant URL | A particular ranking outcome |
Add a fourth column for the owner of the next action. The developer checks the release. The delivery-platform owner investigates inconsistent responses. The search reviewer checks search evidence. One person can fill all three roles, but the questions should remain separate.
A fictional example: one title, two observations
At 09:00, a team releases a new guide title. At 09:05, a browser still displays the old wording. A separate recorded response contains the new title.
Do not write “fixed globally” or “deployment failed.” Write: “New title observed in response B; old title observed in browser A; delivery conditions differ and need comparison.” Check whether the two observations used the same final URL, logged-out state and actual network response. Preserve both examples before making changes.
If a targeted cache action is authorized and justified, record exactly what was invalidated and retest that URL. Keep the before-and-after evidence. Avoid broad purges as a reflex, and do not change account or payment caching while investigating a public article.
Close the search task separately
Google says recrawling may take days to weeks, and a crawl request does not guarantee immediate inclusion or inclusion at all. Repeated requests for the same URL do not make it crawl faster. Use URL Inspection or an appropriate sitemap workflow rather than treating a deployment as a search-update deadline. Google Search Central: request recrawling
A useful completion note is specific: “Approved release verified; corrected public HTML observed at this time; search processing still pending.” That is more actionable than a green “SEO fixed” badge, because it tells the next reviewer both what is known and what remains open.