SEO audit report: turn a crawl into decisions somebody can actually make

An SEO audit report is not a list of everything a crawler noticed. It is the smaller set of findings that deserve a decision, written so the person who owns the website can decide without reopening your tools.

That sounds obvious until a “report” arrives as 63 tabs, 4,000 warnings and a traffic-light score. The data may be correct. The work is still unfinished because nobody has said which pattern matters, what caused it or what should happen next.

The structure of a useful SEO audit report

Use five layers. A reader should be able to stop after any layer and still understand what the preceding pages told them.

1. Scope and limits

Name what you inspected: hostname, subdomains, markets, page types, data sources and date range. Then name what you did not inspect. “Crawled 8,420 indexable HTML URLs on the UK storefront; staging, paid landing pages and JavaScript behind login were excluded” is more useful than “full technical audit.”

Limits protect the conclusion. If the crawl ran without authentication, say so. If conversion data was unavailable, do not label a fix “high revenue impact.” If server logs were not supplied, do not claim how often Googlebot visits a URL pattern.

2. Executive summary

Keep this to the decisions:

• the three to five patterns with the largest likely effect;
• what is already healthy and should not be rebuilt;
• the first action, its owner and its dependency;
• any uncertainty that could change the order.

Executives do not need a softened version of every technical finding. They need the shape of the problem and the trade-off.

3. Prioritised findings

Each finding should answer the same questions:

Field	What belongs there
Observation	What you measured, without a cause attached
Evidence	URLs, counts, screenshots, exports or response samples
Interpretation	The likely cause, clearly labelled as interpretation
Impact	The user or search outcome this can affect
Action	One implementable change
Reach	Which templates and how many valuable pages
Confidence	How sure you are that the diagnosis and fix are right
Effort	A range agreed with the likely owner, not guessed in isolation
Owner	Team or role responsible for the next move
Verification	The test that will prove the change shipped correctly

This format prevents a familiar slide from appearing: “Missing meta descriptions — high priority.” Missing where? How many pages? Are they indexed? Does Google already rewrite the affected snippets? Is the fix a template rule or 6,000 manual edits? Without those answers, “high” is a colour.

4. Roadmap

Group findings by dependency, not by the order your crawler reports them. A clean sequence might be:

1. restore crawling and indexability;
2. resolve canonical and duplicate URL patterns;
3. fix the shared template;
4. improve the priority page set;
5. measure the result before expanding the work.

When two recommendations touch the same template, ship them together. When a migration will replace that template next month, decide whether the interim fix is worth doing at all. The roadmap is where an audit becomes operational.

5. Appendix

Put definitions, full URL samples, crawl settings and raw exports here. The evidence remains available without forcing every reader through it. Link each finding to its relevant appendix row so an implementer can move from summary to detail without searching.

How to prioritise without pretending the numbers are precise

A practical priority model uses four inputs:

Impact. What could improve: eligibility to index, discoverability, snippet quality, page experience or conversion? “SEO impact” is too vague.

Reach. Does the pattern affect one article, the product template or the whole domain? Count affected pages, then identify how many of those pages are valuable or intended for search.

Confidence. Did you reproduce the problem and isolate the cause, or do you merely see a correlation? A lower-confidence finding may deserve investigation rather than implementation.

Effort. Ask the owner. Changing a title rule may be a five-minute configuration change on one platform and a release across four services on another.

Do not multiply made-up scores to create an impressive decimal. “High impact, high confidence, one shared template, small engineering change” communicates more than 8.7.

Write observations before causes

Consider these two lines:

Google cannot render the product pages because JavaScript is broken.

In a fresh browser and in the rendered-HTML test, product descriptions remain empty until a consent choice is made. The raw HTML contains no description.

The first line jumps from symptom to actor to cause. The second records something another person can reproduce. The cause may be consent configuration, hydration, an API failure or geo-specific delivery. Investigation can now test those possibilities instead of inheriting a guess as fact.

This writing habit makes reports feel more human because it removes theatrical certainty. A credible consultant is allowed to say what is known, what is likely and what needs one more test.

What to leave out

Remove findings that are technically true but unactionable, irrelevant to the site’s goals or unsupported by the data. In particular:

• do not treat every crawler warning as a defect;
• do not recommend schema types the page does not qualify for;
• do not prescribe a character count as though search snippets were fixed boxes;
• do not call an issue “duplicate content penalty” without evidence of one;
• do not turn every page into a target page;
• do not forecast traffic from a health score.

A report earns trust partly through what it refuses to overstate.

The one-page finding template

Use this sentence pattern before expanding a finding:

We observed [measured pattern] across [affected valuable pages]. This can affect [specific outcome] because [mechanism]. We recommend [one action], owned by [team]. Confidence is [level] because [evidence or uncertainty]. Verify it by [test].

If you cannot fill a bracket, the finding is not ready for the report. Go back to the audit, ask for the business context or downgrade the claim.

Start with a real crawl rather than a blank document. The Visibility100x audit gives you the evidence layer; the report is the judgment you add on top.