How to conduct a technical SEO site audit

Step 1 to 3: scope, data and crawl

1. Fix the scope in writingName the domain, subdomains, languages and page types. State what is out of scope, such as the checkout or a separate blog platform. Agree who owns each part, because findings need an owner.
2. Collect first-party dataExport the Pages report, the Sitemaps report, the Core Web Vitals report and the Crawl stats from Search Console. Pull the last 90 days of organic landing pages from your analytics tool. If server logs are available, request 30 days.
3. Crawl the siteRun a crawler with JavaScript rendering on for at least a sample, set the user agent to Googlebot smartphone, and respect the real robots.txt. Save the crawl so you can compare it with a later one.

Do the crawl twice for sites that depend on scripts: once with rendering off, once on. The difference between the two crawls is your rendering risk.

Step 4: compare the lists

Most findings come from disagreements between data sources. Build one sheet with a row for every URL and a column for each source.

Step 5: review a sample of every template

A site of 5,000 URLs usually has fewer than 15 templates. Open one or two examples of each in a browser, in view-source, and in URL Inspection live test. A defect in a template appears on every page that uses it, so each template review covers hundreds of URLs.

• Home, category, product or service, article, tag or archive, search results, pagination, contact, 404.
• For each, record the status code, canonical, robots meta, title and heading patterns, structured data type, and whether main content appears without scripts.
• Check the mobile version. Google indexes the mobile rendering first, so content hidden on mobile is easy to lose.

Step 6: score findings so the list is useful

An audit with 140 findings and no order gets filed away. Give each finding three numbers and sort by the result.

• Reach: share of important URLs affected, scored 1 to 5.
• Severity: how directly it blocks crawling, indexing or understanding, scored 1 to 5.
• Effort: developer time, scored 1 (a setting) to 5 (a rebuild).

Priority score = (reach x severity) / effort. Here is an illustrative result for an invented ecommerce site.

The canonical error scores 25 and goes first even though it is a one-line change. Sorting by score exposes cheap, high-impact fixes that a flat list buries.

Step 7: write the report

A report should be usable by a developer who has never met you. For each finding include the issue in one sentence, three example URLs, why it matters, the exact fix, how to test it and who owns it. Put the ten highest-scoring findings on page one.

• Executive summary: five lines, in plain language.
• Top ten fixes, with owners and effort.
• Full findings table, sorted by score.
• Appendix: crawl settings, date, tools and data sources, so the audit can be repeated.

Pitfalls in audits

• Auditing only what the crawler finds. Compare with Search Console and logs or you miss orphan and hidden URLs.
• Reporting tool warnings as findings. A crawler flags every missing meta description, which rarely matters as much as one noindex tag.
• No before and after. Save the crawl and the Search Console export so you can prove what changed.
• Skipping the fix test. Mark a finding done only after a recrawl confirms it.

You can use our SEO audit checklist as the sample-review sheet, or see how an SEO audit service is run with a client.