A technical SEO audit that finds revenue, not warnings
Most audits produce a spreadsheet of 400 issues and no priorities. This is the sequence we use to separate the findings that cost money from the ones that only look untidy.
Key takeaways
- Audit in order: reachability, canonical identity, rendering, then performance.
- The gap between submitted and indexed pages is usually the single most valuable finding.
- Check that AI crawler user agents are not blocked; tightening robots.txt often blocks them by accident.
- Prioritise findings by the revenue of the affected pages, not by a tool's severity colour.
A technical SEO audit should answer three questions in order: can search engines and AI crawlers reach the pages that make money, do those pages resolve to one canonical address, and does the page render what the crawler needs without waiting for JavaScript. Everything else is secondary. An audit that lists 400 undifferentiated issues has not done the job.
One: reachability
Start at robots.txt and work forwards. Confirm nothing commercially important is disallowed, then confirm the same for the AI crawlers you want citing you, which are governed by their own user-agent tokens and are frequently blocked by accident when someone tightens a rule. Check that the XML sitemap is submitted, downloaded without errors, and contains only canonical, indexable, 200-status URLs. A sitemap listing redirects and noindex pages teaches Google to trust it less.
Then check what is actually indexed against what should be. Two numbers matter: pages submitted and pages indexed. A large gap is the most valuable finding in most audits, because it usually points at one systemic cause rather than hundreds of individual faults.
Two: identity
- One canonical host. Pick www or bare domain, redirect the other with a 301, and make sure internal links point at the winner rather than relying on the redirect.
- One canonical protocol and one trailing-slash convention. Every variant that resolves with a 200 is a duplicate competing with itself.
- Self-referencing canonical tags on indexable pages, and correct hreflang pairs on multilingual sites. Hreflang must be reciprocal; a one-way declaration is ignored.
- Parameters, filters and pagination handled deliberately. Faceted navigation is the most common source of crawl budget waste on e-commerce sites.
Three: rendering
Fetch a representative page as Googlebot and read the rendered HTML. The questions are whether the main content, the internal links and the structured data are present without user interaction, and whether the title and description in the rendered output match what you intended. Client-side rendered navigation that only exists after hydration is invisible to a crawler that does not execute it, and most AI crawlers do not.
This is also where you verify structured data. Validate it, but more importantly check that the entities in it are consistent across the site. Conflicting definitions of the same organisation on different pages are worse than no structured data at all.
Four: performance and stability
Use field data from the Chrome User Experience Report rather than a lab score. Check mobile before desktop. Look for the specific failure patterns rather than a global number: which template fails, on which device class, at which metric. A single bad template repeated across 3,000 URLs looks like a site-wide problem and is actually one fix.
How to prioritise the output
Sort every finding by the revenue of the pages it affects, then by the effort to fix. A canonical error on a category page that drives sales outranks a missing alt attribute on a blog image from 2021, no matter what the audit tool's severity colour says. Deliver a shortlist of the five things that will change the number, and put everything else in an appendix.
An audit is not a list of everything wrong. It is an argument about what to fix first.
Sources