Technical SEO Audit Checklist: 30 Checks We Run Before and After a Website Launch

A technical SEO audit is not a score out of 100. It answers one question: can Google, Bing and AI assistants find, read and trust every page that should bring you enquiries? This is the checklist we use on client sites, and the problems it found on our own site after we relaunched it in seven languages.

Crawling and indexing

1. robots.txt allows every public section and blocks only what should stay private: admin, checkout steps, internal search results.
2. The XML sitemap lists only canonical, indexable URLs that return 200. No redirects, no noindex pages, no parameters.
3. The sitemap is submitted in Google Search Console and Bing Webmaster Tools, and Search Console shows zero sitemap errors.
4. Index coverage: compare the number of URLs in the sitemap with the number Google has indexed. A gap is normal for a new site, but it should shrink week by week.
5. Key pages are inspected with the URL Inspection tool: indexed, crawled as mobile, user-declared canonical equals Google's canonical.
6. "Crawled - currently not indexed" pages are reviewed one by one. Google fetched them and said no, which usually means thin or duplicate content.
7. Nothing important depends on JavaScript to appear. View the raw HTML: titles, headings, body text and links should be there before any script runs. AI crawlers often do not run JavaScript at all.

What we found on our own site: a week after relaunching unlockd.art in seven languages, Google had indexed 188 of 460 sitemap URLs, with 70 discovered but not yet crawled. Nothing was broken; a young site simply gets crawled slowly. The fix is not technical: internal links to the pages that matter and links from other sites.

Status codes and redirects

8. Every old URL redirects with a single 301 (or 308) to its closest new page, not to the homepage. Test the full list from the old sitemap, not a sample.
9. No redirect chains (A to B to C) and no loops.
10. Removed pages return 410 or 404, not a soft 404 that shows "page not found" with status 200.
11. One version of the domain: http redirects to https, and either www or the bare domain redirects to the other.
12. Links from your Google Business Profile and social profiles point to live pages. A profile link that ends in a 404 loses visitors on the first click.

Duplicates and canonicals

13. Every page has a self-referencing canonical, and paginated or filtered views point to the right one.
14. Tracking parameters (utm, gclid) do not create indexable duplicates.
15. No two pages target the same query. In Search Console, filter by a query and check how many of your pages appear. When we checked ours, the brand name was still split across the old /en URLs and the new root pages, a leftover of the migration that disappears as Google recrawls the redirects.

International sites

16. hreflang is reciprocal: every language version lists all the others and itself, plus an x-default.
17. Each language has its own canonical, never pointing to the English page.
18. No automatic redirect by IP or browser language. Offer a visible language switcher instead; crawlers mostly come from the US and would never see your other languages.
19. No untranslated pages under a language folder. English text under /de/ is a duplicate, not a German page.

Speed and mobile

20. Core Web Vitals pass on mobile in the field data, not only in a lab test: LCP under 2.5 s, CLS close to zero, INP under 200 ms.
21. The main heading and text render without waiting for a hero video, a slider or a font. Below-the-fold images load lazily.
22. The page is built for phones. A missing viewport tag still shows a shrunken desktop page on a phone. We found that on many company sites during our outreach research.
23. HTTPS works without warnings. Note that modern Chrome upgrades http links to https automatically when the site supports it, so an http link alone is not a security warning; a broken certificate is.

On-page technical basics

24. One unique title per page, under about 60 characters, with the query people actually type.
25. One unique meta description, under about 160 characters.
26. One H1 per page that says what the page is about.
27. Descriptive alt text on meaningful images, empty alt on decorative ones.
28. No leftover template text. Search the rendered pages for "Lorem ipsum" and sample phone numbers. It happens on live sites more often than you would think.

Structured data and AI search

29. Structured data matches the visible page: Organization, Service, Product, Article, FAQ or BreadcrumbList where they fit, validated without errors.
30. AI assistants can read and cite you: server-rendered HTML, clear answers near the top of key pages, consistent company facts everywhere, and an llms.txt that points to your most important pages. More in how to get your business recommended by ChatGPT.

When to run the checklist

Run the full list before a launch or migration, and again two to four weeks after, when Google has recrawled the site. If you are moving platforms, start with our WordPress to Next.js migration checklist and the guide to redesigning without losing rankings.

If you would rather have it done for you, our paid website audit covers technical SEO, conversion and UI/UX in 3 days, with a recorded video walkthrough and a prioritised plan. The fee is €390 and is credited in full if you continue with a project.