SEO Audit: The Complete Guide

Most SEO audits fail for the same reason. Somebody runs a crawler, exports 340 issues sorted by severity as defined by the tool, formats it into a deck, and hands it over. Nothing in it is wrong. Nothing in it gets done either, because there is no way for the person receiving it to tell which four things actually matter.

A useful audit ends with a short, ranked list of changes and a reason each one is on the list. Everything below is in service of getting to that list. This guide walks each layer in the order I would actually check it, then gives you the process, the checklist and the reporting format.

What an SEO audit is

An SEO audit is a structured diagnosis of why a site performs the way it does in search, and what would change that. It is diagnostic rather than prescriptive. The output is a prioritised set of hypotheses backed by evidence, not a to-do list generated by a tool's default rules.

The distinction matters because tools do not know your business. A crawler flagging 200 pages with missing meta descriptions is technically correct and usually irrelevant. Three of those pages might be your highest-intent commercial pages, which is the finding. The other 197 are noise.

Why it matters

Three situations where an audit pays for itself, and one where it does not.

Traffic has dropped and nobody knows why. The most common trigger, and the one where an audit is genuinely diagnostic. The answer is almost always in the last 90 days of changes.

You are about to invest in content or links. Publishing into a site with indexation problems is spending money to build on sand. Audit first, in that order.

Growth has flatlined despite consistent effort. Usually a structural problem: cannibalisation, thin content diluting the domain, or an authority ceiling. None of these show up in a content calendar review.

Where it does not pay: a brand new site with 12 pages. You do not need an audit, you need content and links. Do not let an agency sell you a diagnosis of a site that has nothing to diagnose.

Start with the data, not the crawler

Before you crawl anything, spend an hour in the analytics. This is the step most audits skip and it is the one that tells you where to look.

• Search Console, 16 month view. Clicks and impressions over time. Note the shape. Gradual decline, cliff edge, or plateau? A cliff means a specific event. A gradual slide usually means content decay or competitive displacement.
• Page-level comparison, last 3 months vs same period last year. Sort by clicks lost. Ten pages usually account for most of the decline.
• Query-level comparison. Same exercise. Look for queries where impressions held but clicks fell, which is the fingerprint of an AI Overview or a SERP feature landing on that cluster.
• Analytics landing pages by conversion, not by traffic. The pages that matter commercially are rarely the pages that get the most visits, and they deserve disproportionate audit attention.
• Deployment and content history. Ask what changed. Migrations, template updates, plugin changes, CMS upgrades, robots.txt edits. Cross-reference the dates against the traffic chart.

By the end of this hour you should have a hypothesis. Everything after this is testing it rather than fishing.

Technical SEO audit

Work the pipeline in the order search engines experience it, because a failure early on makes everything downstream irrelevant. Our technical SEO guide covers the mechanics of each stage in depth; what follows is the audit sequence.

Indexation first

Open the Page Indexing report in Search Console and read the exclusion reasons. This single report answers more audit questions than any crawler.

• "Crawled, currently not indexed" on pages you care about. Not a technical fault. A judgment about value, usually thin content or no internal links.
• "Discovered, currently not indexed" at scale. A crawl budget or site quality signal. Check server response times.
• "Excluded by noindex tag" on anything commercial. Check immediately. Staging noindex tags shipping to production is depressingly common.
• "Blocked by robots.txt" on indexable content. Same urgency.
• "Alternate page with proper canonical tag" in large volume. Verify the canonical targets are what you intended.
• "Duplicate, Google chose different canonical". Google disagreeing with you. Worth understanding why.

Then crawl

Run Screaming Frog or equivalent, connected to Search Console and your analytics so you get traffic data alongside crawl data. What to look at, in order:

1. Response codes. Internal links to 404s and 301s. The redirects are the more common and more ignored problem.
2. Redirect chains. Anything over one hop, especially in navigation.
3. Canonical consistency. Do internal links, sitemap entries and canonical tags all agree on the same URL? Disagreement is how you lose control of which version ranks.
4. Titles and H1s. Missing, duplicated, or truncated. Filter to pages with traffic first.
5. Crawl depth. Anything commercially important more than three clicks from the homepage.
6. Rendered vs raw HTML. Crawl with JavaScript rendering enabled and again without. Content appearing only in the first is a rendering risk.
7. Sitemap accuracy. Crawl the sitemap separately. Every URL in it should be canonical, indexable and return 200.

Then logs, if you can get them

Server logs are the only place you see what Googlebot actually requested. On large sites this consistently finds crawl budget being spent on parameterised URLs nobody knew existed. On small sites it is usually skippable.

On-page audit

On-page work is high leverage and quick, provided you audit the right pages. Take your top 30 pages by conversion or commercial intent and check each one properly, rather than checking 3,000 pages superficially.

For each page:

• Does the page match the intent of its target query? Search the query, look at what ranks, and ask whether your page is the same type of thing. A service page competing against buying guides will not win no matter how well optimised it is.
• Is the primary keyword in the title, H1 and first 100 words? Basic, still frequently missing.
• Is the title written for clicks as well as relevance? Check CTR in Search Console against the average for that position. Below average CTR at a good position is a title problem, and it is the fastest win in SEO.
• One H1, logical H2 and H3 hierarchy. Headings should read as an outline of the page.
• Is there a direct answer near the top? Increasingly what determines snippet and AI citation eligibility.
• Images with descriptive alt text and sensible file sizes.
• Schema present, valid and matching visible content.
• A clear next action. An optimised page with no conversion path is a vanity ranking.

Cannibalisation

Check this deliberately, because it hides well. In Search Console, filter by a query and look at which pages receive impressions for it. Two or more of your own URLs appearing for the same query, or a ranking URL that flips between two pages over time, means you are competing with yourself.

The fix is a decision, not a tweak: consolidate the pages into one and redirect, differentiate them so they target genuinely different intents, or noindex the weaker one. Choose one. Leaving both live and hoping is the default outcome and the wrong one.

Content audit

The content audit is where most sites find their largest untapped gains, and also the part people resist because the conclusion is often "delete this".

Export every URL with its clicks, impressions, average position, conversions, word count and last modified date. Then sort every page into one of four buckets:

Signal	Bucket	Action
Ranking positions 4 to 15, decent impressions	Improve	Highest priority. Closest to real gains.
Good traffic, declining trend, old date	Refresh	Update facts, add depth, republish.
Two pages on near-identical topics	Consolidate	Merge into one, redirect the weaker.
No clicks, no impressions, no links, over a year old	Remove	Noindex or delete with a redirect.

A few honest notes on this. The "improve" bucket is where your quarter should go, because moving position 8 to position 3 on an existing page beats publishing something new almost every time. The "remove" bucket feels like destroying work, but a site carrying 400 pages of which 250 have never earned a click looks materially different to a search engine than a site with 150 pages that all perform. And check for links before deleting anything, because a page with no traffic and three referring domains should be redirected, not removed.

Our content strategy guide covers the refresh cycle and how to structure what replaces the pages you cut.

Internal linking audit

Internal linking is the most underused lever in SEO, largely because no tool tells you what a good structure looks like for your specific site.

What to check:

• Orphan pages. Pages with no internal links pointing at them. Compare your crawl against your sitemap and analytics to find pages that exist but are unreachable.
• Link distribution against importance. Export inbound internal links per page and compare against commercial value. If your pricing page has four internal links and a 2022 blog post has sixty, that is backwards.
• Anchor text. "Click here" and bare URLs tell search engines nothing. Descriptive anchors do.
• Links to redirects. Every internal link should point at the final URL.
• Cluster integrity. Do supporting articles link up to their pillar page, and does the pillar link back down? Most sites do the first and forget the second.
• Click depth from the homepage, for anything that matters.

The fix is usually adding 30 to 50 contextual links in a single afternoon, which is a rare combination of cheap and effective. Our SEO strategy guide sets out the hub and spoke structure worth aiming at.

Backlink audit

Two questions only: is your profile strong enough to compete for what you are targeting, and is there anything in it that could hurt you?

On strength. Compare referring domains against the three sites currently ranking for your primary commercial terms. If they have five times your referring domains, your ranking ceiling is real and no amount of on-page work moves it. That is a finding worth putting in the report, because it changes what content is worth commissioning.

On risk. Google's algorithms discount most low-quality links automatically, so disavow files are far less necessary than they were a decade ago. The cases still worth attention: an obvious paid link scheme from a previous agency, a sudden spike of irrelevant links from a single country or niche, or sitewide footer links across dozens of unrelated domains. Absent those, do not disavow. People damage themselves with disavow files far more often than they rescue themselves.

On opportunity. Check broken backlinks pointing at 404s on your site, which is free authority recovered with a redirect. Then look at competitors' referring domains for patterns you could realistically replicate, which feeds directly into the work in our digital PR guide.

UX and Core Web Vitals

Use the Core Web Vitals report in Search Console for field data, and treat Lighthouse purely as a debugging tool. Lab scores make poor targets.

Check the three metrics on mobile specifically, since mobile is what gets indexed and mobile is where the failures are. LCP under 2.5 seconds, INP under 200 milliseconds, CLS under 0.1. Moving from poor to good is worth a sprint. Moving from good to marginally better usually is not.

Beyond the metrics, a handful of experience checks that affect rankings indirectly through behaviour: intrusive interstitials on entry, content requiring interaction to read, tap targets too small on mobile, and forms that are longer than they need to be. Read three of your own key pages on a phone, on a normal connection, as a stranger would. It is a more honest test than any score.

Local SEO audit

If you have physical locations or defined service areas, this layer is separate and often more valuable than everything above it. The short version:

• Google Business Profile claimed, verified, complete, with the right primary category
• NAP identical across the profile, your site footer, your schema and the top directories
• Duplicate listings identified and removed
• Review count, average and recency benchmarked against your three nearest competitors
• Every review responded to
• Location pages with genuinely local content, not templated city swaps
• LocalBusiness schema per location with correct coordinates
• Grid-based rank tracking rather than single-point rank checks

The full detail sits in our local SEO guide, including how to handle multi-location structure and service area businesses.

AI search readiness audit

This layer barely existed two years ago and now belongs in every audit. The checks are straightforward:

• Robots.txt. Are retrieval crawlers such as OAI-SearchBot and PerplexityBot allowed? Many sites blocked them accidentally while trying to block training crawlers, and quietly removed themselves from citation eligibility.
• Extractability. Do your key pages answer their question in the first 40 to 60 words under a heading phrased as that question?
• Entity consistency. Is your company described the same way on your site, your schema, your social profiles and your directory listings? Inconsistency is how models get you wrong.
• Freshness signals. Visible last-updated dates, and current figures rather than three-year-old ones.
• Prompt testing. Ask ChatGPT, Perplexity, Gemini and Claude to recommend a provider in your category. Record whether you appear, what they say, and which sources they cite. Repeat monthly. This is the closest thing to rank tracking these surfaces offer.
• Referral check. Look for traffic from assistant hosts in analytics. Most sites have some and have never noticed.

Our AI search optimization guide covers what to do about each of these.

Step-by-step audit process

1. Day one: data. Search Console 16 months, page and query comparisons, analytics by conversion, change history. Form a hypothesis.
2. Day one: indexation. Page Indexing report end to end. Anything blocking indexation goes straight to the top of the findings.
3. Day two: crawl. Full crawl with Search Console and analytics connected. Response codes, canonicals, redirects, depth, rendering.
4. Day two: on-page. Top 30 commercial pages, individually. Cannibalisation check across the query set.
5. Day three: content. Full inventory sorted into improve, refresh, consolidate, remove.
6. Day three: internal links. Orphans, distribution against importance, cluster integrity.
7. Day four: backlinks and competitive context. Strength gap, risk, recovery opportunities.
8. Day four: UX, local and AI readiness. Field Core Web Vitals, profile checks if applicable, crawler access and prompt tests.
9. Day five: prioritise and write. Score by impact and effort. Write the report backwards, conclusions first.

Audit checklist

Blocking issues, fix regardless of effort

• No unintended noindex tags in production
• Robots.txt not blocking indexable content or render-critical assets
• Commercial pages indexed and returning 200
• No accidental canonical pointing away from money pages
• Main content present in server-rendered HTML

High impact, low effort

• Titles rewritten on pages with below-average CTR for their position
• Internal links added to under-linked commercial pages
• Internal links repointed from redirects to final URLs
• Broken backlinks redirected to relevant live pages
• Cannibalising pages consolidated or differentiated
• Direct answers added near the top of key pages
• AI retrieval crawlers unblocked

Higher effort, still worth it

• Improve bucket pages rewritten and expanded
• Thin and dead pages removed or consolidated
• Redirect chains resolved
• Core Web Vitals moved from poor to good on mobile
• Structured data implemented and validated
• Local pages given genuine local content

Ongoing, not one-off

• Monthly indexation and crawl error check
• Quarterly content decay review
• Monthly prompt testing across assistants
• Monthly competitor review and referring domain tracking

Common issues and their actual fixes

"Crawled, currently not indexed" on good pages. Not a technical fix. Add internal links from relevant pages, deepen the content, and make sure the page is not near-duplicating something else you published.

Traffic dropped after a redesign. Check redirects from old URLs first, then compare rendered HTML on old and new templates, then check whether content was trimmed in the name of cleaner design. It is usually the third one.

Rankings stuck at positions 5 to 10 across the board. Almost always authority, not on-page. Check referring domains against the sites above you before touching the pages again.

Impressions up, clicks flat. Either SERP features taking the click, or title tags that do not earn one. Segment by query to tell which.

New content not getting indexed. Check internal links to it, check the sitemap includes it, and check overall site quality signals. New pages on sites with large volumes of thin content get indexed slowly.

Reporting so the work gets done

Write the report backwards. Nobody reads to page 40 to find the recommendation.

Page one: the three things. The three changes that will matter most, what each is expected to do, and who needs to do it. If you cannot narrow it to three, you have not finished analysing.

Page two: the ranked list. Every finding, scored by impact and effort, with an owner and a rough estimate. A table, not prose.

Then the evidence. Detail per finding, with the screenshot or export that proves it. This section exists so someone can verify your reasoning, not so they can read it front to back.

Finally, what you checked and found healthy. Short, but include it. It stops the same questions being asked again in three months, and it tells the reader you looked.

Two rules that matter more than format. Never present an issue without a fix and an owner, because unowned findings are just complaints. And never pad the list to look thorough. An audit with nine real findings and a clear order of operations is worth more than one with ninety, and it is far more likely to result in anything changing.