An SEO audit records an observation made in a particular way and at a particular time. It does not see a site “in general”: a scanner uses a specific user agent, IP, timeout, JavaScript mode, and crawl rules. A disputed finding may therefore be a genuine defect, a tool limitation, a temporary failure, or a rule applied incorrectly. The recheck should establish which explanation is supported by the evidence.
Turn the finding into a testable statement
“Page unavailable” is too broad. Record the exact URL, time, status code, and request conditions. “The scanner received 504 in its response at 14:32” can be tested. Clarify “no H1” as either “the source HTML contains no <h1> element” or “there is no visible H1 in the DOM after rendering.” These are different tests with different possible causes.
For each finding, define the expected state. Should a canonical exist on every page or only on indexable pages? Does the requirement for a unique description apply to product pages, or also to utility pagination? A check without a defined scope produces false errors even when the data is extracted accurately.
Save the original report excerpt, audit version, and run parameters. If a subsequent crawl has already overwritten the result, it will be harder to demonstrate a temporary failure.
Reproduce the network request
Start with HTTP, not the browser view. Retrieve the response headers and body without cookies, authentication, or a local cache. Then repeat the request with the scanner’s user agent, a search engine crawler’s user agent, and a regular browser’s user agent. Check DNS, the redirect chain, the final status code, response time, size, and content type.
A difference by user agent or IP points to a WAF, CDN, geographic rule, or anti-bot measure. The same 5xx confirms a server-side problem. If one request succeeds, that does not rule out an intermittent failure: review monitoring and logs around the time of the audit. The incident response plan describes how to locate the cause of unavailability.
Do not label a tag as missing if the scanner did not receive any HTML to analyze. In the case of a timeout or empty response, the correct finding status is “could not verify.” Missing data and a negative result are different states.
Compare three representations of the page
For HTML checks, save:
- the original HTTP response body;
- the DOM after JavaScript has run;
- the screen visible to the user.
An H1 may be present in the DOM but absent from the source; a canonical may be correct on the server and then replaced by client-side code; text may be displayed visually in a canvas without existing as indexable HTML. The report should make clear which representation the tool analyzes.
If content is loaded through an API, check for network errors, access policies, delays, and the state shown to a logged-out user. Sometimes a team sees data because it has an active session, while the scanner receives an empty template. For more context on these differences, see the material on mobile indexing: the check needs to cover the version actually available to the crawler, including the mobile version.
Check the meaning of the detected element
An element can be present technically and still be incorrect. Two H1s, one of them hidden in a mobile menu, do not make a correct structure. A canonical exists, but points to the home page instead of the product page. A description is present, but repeated in full across thousands of URLs. A simple substring search is only the first step.
Check the element’s location, value, visibility, uniqueness, and relevance to the page. For a redirect, inspect the full chain: the first response does not show the final destination. For robots, collect the rules from meta and X-Robots-Tag, identify which crawler each applies to, and account for the most restrictive applicable rule. A set of <link> confirms hreflang only when the reciprocal links are valid and their target pages are indexable.
Also ask whether the rule is useful. Some audits flag the absence of elements that are not required for a particular page type. A recommendation to add markup should not automatically be described as a blocking indexing issue.
Check the URL context
A URL may be utility, duplicate, canonical, redirecting, or intentionally blocked. Before fixing “noindex,” find out whether the URL should appear in search. Before adding a unique title to a filter page, decide whether the filter is a standalone landing page.
Compare the HTTP status code, robots, canonical, sitemap, and internal links. If all signals consistently exclude the URL, its absence from the index is not a defect. If the sitemap suggests that the URL should be indexed, the canonical points elsewhere, and navigation links to it through a redirect, the report has found a real conflict, even if an individual tag is technically correct.
The guide to canonical explains duplicate versions in more detail. Use it as a basis for your decision. The same tag on every URL does not create correct canonicalization.
Identify scanner limitations
Find out whether the audit runs JavaScript, how long it waits, whether it follows robots.txt, whether it follows redirects, the maximum HTML size it accepts, whether it scans subdomains, and whether it uses canonical when counting duplicates. A five-second timeout and blocked resources can explain a discrepancy with a manual browser check.
Review the run settings: crawl scope, URL limit, depth, and enabled exclusions. A page may not have been included in the audit because the limit was reached, even if a link to it exists. Checking one page and running a full crawl can also produce different conclusions about internal links and uniqueness.
If the tool provides a raw response or diagnostic log, attach it to the task. If not, reproduce the conditions as closely as possible yourself and clearly state the limits of your certainty. “Cannot reproduce from our network” is more accurate than “false positive” when the scanner’s conditions are unknown.
Assign a verdict using a standard scale
To keep the team from arguing over wording, use a set of outcomes:
- confirmed — the defect is reproducible and violates the expected rule;
- partially confirmed — the problem occurs only for some variants or intermittently;
- false positive — the site complies with the rule, and the cause lies in the analysis method;
- outdated result — the state changed after the audit;
- could not verify — there is no response, access, log, or reproducible set of conditions;
- not applicable — the rule does not apply to the URL’s purpose.
Add evidence to the verdict: the request command, saved HTML, DOM screenshot, log entry, or list of examples. “It opens for me” does not confirm consistent availability, and “the tool said so” does not confirm that an element is missing.
Fix the cause and repeat the same test
If the defect is confirmed, change its source: the template, data, server status, CDN settings, or link structure. After publishing the change, repeat the original test under conditions as similar as possible, then check other URLs of the same type. This prevents fixing a single example while leaving the underlying issue across many pages.
reChecker technical audit saves the results of standard page checks, while the full mode collects several URLs from the sitemap and internal links or accepts a specified list. Attach its report as another observation and compare it with the HTTP response, DOM, and URL purpose.
It is useful to summarize the final record in five fields:
| Field | What to save |
|---|---|
| Original statement | Exact check text and affected URL |
| Actual response | Status code, header, HTML, or DOM, with the date of the check |
| Reason for discrepancy | Site error, method limitation, or change since the audit |
| Action | Fix, rejection, or request for missing data |
| Retest | Result of the same request after the action taken |
If there was not enough data, keep the status “could not verify” and list the missing conditions. The next run can then be compared with a specific request and saved response.
Check the original result and the next run
In public report examples, check the date and URL tested; in the methodology, check the scope of the test. For your own project, save the original result before running the audit again: a fresh successful response does not erase a recorded failure or prove that its cause has been fixed.
In the public rechecker.ru example from 15:27 MSK on 3 October 2026, CLS 0,401 relates to one page, https://rechecker.ru/, and a lab run. When rechecking, save the new result and its conditions; if the original device or network is unknown, do not describe the difference as a confirmed effect of the fix. Field CLS and its impact on conversions require different data.
If the finding is confirmed, assign a priority and create a task for the developer with the same URL and acceptance criteria. After the fix, repeat the check using the technical audit. To reduce the risk of recurrence, choose a regular monitoring mode; verify search indexing separately in webmaster tools.