Website audit for SEO: attach evidence to each issue

A useful website audit leads to testable decisions. A list of 300 warnings without URLs, evidence, or consequences does not tell a team what to fix first. This method is small enough to run but precise enough to separate a sitewide template defect from an isolated page issue.

Define the question and baseline

Write down what the audit must explain: falling organic clicks, missing indexed pages, errors after a migration, fewer enquiries, or a broken purchase journey. Each question requires different evidence. Record the period being examined and a comparison period that accounts for seasonality.

Gather the sitemap, analytics landing pages, Search Console page and query data, page templates, pages with external links, and important conversions. If a metric is unavailable, mark it unknown. Missing data should not silently become zero or a confident conclusion.

Choose a representative sample

Select at least one URL of each type your site actually uses: home, service, category, product, article, pagination, filtered results, and an old redirected page. Add the pages that receive most clicks, enquiries, or links. Eight carefully selected URLs may reveal a shared template problem; they cannot prove every other page is healthy.

For each URL, record the full address, page type, HTTP status, final destination after redirects, canonical, sitemap presence, internal links, and page goal. A simple spreadsheet is enough if the tests can be repeated.

Test access, indexing signals, and canonicals

Request each page and inspect its actual response: 200, permanent redirect, 404, or server error. A permanent redirect should end on a relevant page; a chain or loop needs attention. Check whether robots.txt blocks crawling, whether noindex is present, and whether the canonical points to the intended URL.

The sitemap should list canonical URLs you want discovered. Inclusion does not guarantee indexing. A page missing from the sitemap can still be found through internal links. Compare the sitemap with your internal links and Search Console data. Google's sitemap documentation explains its discovery role.

Where several URL variants show the same content, select a consistent preferred version for internal links, canonicals, and the sitemap. Google documents canonicalization signals. A canonical is a preference signal, not a cure for an empty or misleading page.

Check whether the page answers the search

A technically accessible page can still fail its reader. Read the title, opening, practical detail, and next step. Does it explain the promised subject, state relevant limits, and offer useful evidence such as a process, a verifiable example, a screenshot, or an appropriate primary source?

Compare a few queries producing impressions with the page they lead to. If a pricing page never explains why the price varies, ten more generic paragraphs will not answer the user's question. The content itself needs to address it. Google's SEO starter guide covers understandable site structure and people-focused content.

Also look for near-duplicate pages created for minor keyword variations. A stronger central page may serve the intent better, while genuinely different needs may warrant separate pages. Do not decide solely from an automated similarity score.

Turn findings into assigned decisions

For each issue, record affected URLs, reproduction steps, evidence, likely effect on visitors or search, scope, owner, proposed fix, and a closure test. Separate observations from hypotheses. “Canonical points to the old URL on 24 products” is testable. “Google penalized us” requires much more evidence.

Prioritize access and conversion failures on valuable pages, then shared template defects, then editorial improvements. A minor warning on a page with no traffic or business role should not outrank a broken enquiry form on a sought-after service page.

Fictional example: a crawler flags 300 filter URLs. Sampling shows that two combinations attract useful visits and the others offer no distinct content. The team improves the two useful pages and their internal links first, then defines a consistent rule for the remaining variants.

Retest and measure

After a fix, repeat the original test on the same URL. If a template changed, test several pages using that template. Save the release date and the new result. Then track clicks, impressions, enquiries, or sales for the affected pages over comparable periods. An increase does not prove the fix alone caused it; other conditions can change.

The final deliverable is not a score out of 100. It is a decision table with evidence and an owner for each correction. Start with a representative sample, then expand wherever it uncovers a shared defect.

Example issue record

Imagine a shop category listed in the sitemap whose canonical points to an old category that now redirects. The issue record should keep both URLs, their HTTP responses, the canonical visible in the source HTML, and the sitemap entry. The suspected impact is that a search engine may select a different address from the one the team intends to promote. The proposed fix is to set the canonical to the final category and make internal links and the sitemap use that same address.

The closure test goes beyond seeing a change in the CMS editor. Request the public URL, inspect its response and HTML, and check another category using the same template. Only then look for a change in Search Console. Search Console data may lag: the HTTP and HTML checks confirm the technical fix, while search impact still needs observation.

An article losing clicks without a technical error needs a different issue record. Keep the queries that declined, comparison dates, editorial changes, and competing pages that answer the same question. A useful next action may be a better answer to the reader's question, rather than another meta tag or redirect.

Decide when to expand the sample

Sampling is a way to detect patterns, not a shortcut for declaring the whole site clean. If two tested product pages share a wrong canonical, inspect the product template and sample more products across collections. If one manually edited article has a missing heading, check its edit history before assuming every article is affected.

Use Search Console's page examples to locate clusters, then reproduce the problem in a browser or HTTP client. Keep a list of tested URLs and the reason each one was selected. That makes the audit repeatable when the site changes next month.