What Is an SEO Audit? What to Check, and in What Order

Aug 05, 2026 · SEO Glossary

Most sites I've audited were leaking traffic through problems nobody had looked for: a robots rule blocking a template since the last redesign, 30% of the sitemap redirecting, title tags duplicated across 400 pages, one slow database query putting a full second under every page load. None of those announce themselves. That's the case for an audit — not as a deliverable with a cover page, but as a periodic, systematic answer to "what's silently broken?"

What is an SEO audit, concretely? A structured review of every layer that determines whether your pages can rank: whether search engines can reach your content, whether it gets indexed, whether the pages send the right signals, whether they're fast and usable, and how the site's links — internal and external — distribute authority. The order matters, because each layer gates the next: perfect content on a page Googlebot can't crawl is worth exactly nothing.

Layer 1: Can It Be Crawled?

Start where Googlebot starts. Fetch your robots.txt and read every rule. Then crawl the site the way a search engine would and compare the result against reality:

• URLs blocked by robots.txt that shouldn't be (and assets — CSS, JS — that must not be).
• Redirect chains and loops, broken internal links, 404s linked from live pages.
• Orphan pages — in the sitemap or database but linked from nowhere.
• Click depth: important pages more than 3–4 clicks from the homepage get crawled and weighted accordingly.
• Server health under crawl: if your own crawl at modest speed produces 5xx errors, Googlebot's does too, and Google throttles crawling in response.

On sites past a few hundred thousand URLs, add server logs to this layer: they show where Googlebot actually spends its requests, which is frequently a surprise — endless parameter variations, an old API endpoint, a calendar widget generating infinite months. No crawler simulation replaces watching the real bot's behavior at that scale.

Layer 2: Is It Indexed — and Should It Be?

Search Console's page indexing report is the ground truth. The number to interrogate isn't "indexed" — it's the ratio. A site with 2,000 real pages and 40,000 known-to-Google URLs has a URL generation problem (filters, parameters, calendar pages) diluting its crawl budget. Read the exclusion reasons like a diagnosis: "Crawled — currently not indexed" at scale suggests quality or duplication problems; "Duplicate without user-selected canonical" points at missing canonical tags; "Discovered — currently not indexed" on new content suggests crawl prioritization issues. Cross-check that the XML sitemap contains exactly your indexable URLs — no redirects, no 404s, no noindexed pages — because a dirty sitemap makes every other indexing signal noisier.

Layer 3: What Signals Do the Pages Send?

Now the classic on-page pass, done at inventory scale rather than page by page: titles missing, duplicated, or truncating; meta descriptions absent or cloned across templates; multiple H1s or none; thin pages with a few dozen words competing against full articles; structured data errors. Individually small, these compound — and because they're template-driven, they cluster: fixing one archive template often corrects hundreds of URLs. This layer is where audit tooling earns its keep, since no human is reading 3,000 title tags. It's also where you check intent alignment on your money pages against what actually ranks — does the page type match what Google rewards for the query?

Layer 4: Is It Fast and Stable?

Pull field data first — Search Console's Core Web Vitals report groups failing URLs by template, which tells you where the work is. Lab-test representative URLs from each failing group rather than the homepage five times. The usual suspects, in payoff order: slow server response time, oversized images, render-blocking assets, third-party scripts. Details and thresholds are in our Core Web Vitals breakdown.

Layer 5: How Does Authority Flow?

Internally: which pages hoard inlinks (usually the privacy policy) versus which pages earn revenue, and whether anchor text says anything useful. Externally: a backlink review for lost links pointing at 404s — reclaimable by redirect — and for toxic patterns worth knowing about, though disavowal is rarely needed now. The internal half is fully under your control and routinely the fastest win in the whole audit.

What Is an SEO Audit Supposed to Produce?

Not a 90-page PDF. A prioritized fix list, where priority = (traffic or revenue at stake) × (severity) ÷ (effort to fix). Sitewide crawl blocks and 5xx instability outrank everything; template-level title duplication outranks one page's missing alt text. Every finding should name the evidence (which report, which URLs), the fix, and the owner. If you deliver audits to clients, a sample audit report shows one workable format — agencies typically lead with a one-page summary and stack the URL-level detail behind it.

And schedule the re-run. Sites regress: deploys ship noindex tags, plugins rewrite titles, redirects pile up. A scheduled automated audit that recrawls on a cadence and diffs against last month catches regressions the week they ship — which converts the audit from an annual archaeology project into a monitoring practice. The first audit finds years of accumulated problems; the recurring one keeps the list near zero.