The SEO Audit Process: A Step-by-Step Framework

The crawl is your evidence base, and bad setup poisons every downstream conclusion. Before hitting start:

1. Crawl with JavaScript rendering on (at least for a representative sample). If content or links only exist after hydration, an HTML-only crawl audits a site that isn't the one Google sees.
2. Respect the real robots environment, crawl as a search bot user-agent, honoring robots.txt, so blocked sections show up as blocked rather than silently missing.
3. Connect the data layers, plug Search Console, analytics, and log files (or crawl-stats exports) into the crawl so every URL carries impressions, clicks, sessions, and crawl frequency alongside its technical attributes.
4. Crawl the sitemaps separately, the delta between “URLs in sitemaps” and “URLs found by crawling” exposes orphan pages (linked from nowhere) and phantom entries (sitemapped but dead).

From the crawl, the findings worth extracting are architectural, not cosmetic: redirect chains and loops, canonical conflicts, crawl-depth outliers (money pages more than 3–4 clicks deep), duplicate-content clusters from parameters and facets, and template-level failures like a bad canonical shipped across a whole page type. Individual missing alt tags are noise at this stage; a template that noindexes a category is a headline.

Indexation analysis compares three sets: URLs that should be indexed (your canonical, valuable pages), URLs that are indexed, and URLs Google is spending crawl on. Misalignment runs in both directions, and both cost you:

• Wanted but not indexed, valuable pages sitting in “Discovered, currently not crawled” or “Crawled, currently not indexed” in Search Console. The first usually signals crawl-priority or internal-linking starvation; the second is typically a quality or duplication verdict on the page or its template.
• Indexed but not wanted, parameter variants, tag archives, paginated fragments, staging leaks, expired listings. These dilute site-level quality assessments and compete with the pages you actually want ranking.

Work through Search Console's page indexing report reason by reason, and reconcile each bucket against your intended index. The output of this stage is an index policy: an explicit statement of which page types should be indexed, which should be crawlable but noindexed, and which shouldn't be generated at all, then a gap list of everything violating it. On large or programmatic sites this stage is where most recoverable value hides, as we detail in the indexation section of our programmatic SEO guide.

Where is crawl actually going?

If log data is available, rank URL patterns by Googlebot hits and compare against value: it's routine to find 40% of crawl going to faceted URLs that will never rank while new content waits days for discovery. Fixes are usually structural, kill crawl paths to worthless permutations, flatten access to valuable ones.

With the technical floor established, audit whether the content deserves to rank. This stage assigns every indexable URL a verdict based on two questions: does it match a real search intent, and is it the best answer you could give? Word counts, keyword densities, and “content scores” are not the rubric, intent match is, as unpacked in our search intent guide.

1. Map each URL to its primary query and intent using Search Console query data. Pages with impressions but poor CTR or position 8–20 rankings are your improvement shortlist, demand exists, delivery is off.
2. Detect cannibalization, multiple URLs rotating for the same query. Decide a canonical page per intent and consolidate the rest into it with redirects.
3. Find coverage gaps, intents in your topic where you have no page at all. This connects the audit to your topical map: an audit tells you what's broken, the map tells you what's missing.
4. Assign verdicts, every URL leaves this stage tagged keep, improve, consolidate, or remove, with the evidence attached.

“The most valuable line in an audit is rarely “add this.” It's “these 300 pages are hurting you, merge 80, fix 40, and delete the rest.””

Include a quality pass on the pages you keep: authorship and evidence signals, accuracy, freshness. Site-level quality is judged on the whole corpus, so authorless, stale, or unverifiable pages drag down the pages you're counting on.

Audit the internal graph before the backlink profile, because it's the one you fully control and it's usually where the bigger problem lives. From the crawl data:

• Equity distribution, rank pages by internal links received and compare against pages ranked by business value. The mismatch list (money pages starved, junk pages fed) is your fix list.
• Orphans and near-orphans, indexable pages with zero or one internal link. If nothing links to it, you've told Google it doesn't matter.
• Anchor quality, “click here” and bare-URL anchors waste the relevance signal; anchors should name the target entity, per our internal linking strategy.
• Cluster integrity, do pages within a topic cluster actually link to each other and up to their hub, or is the “cluster” only a spreadsheet concept?

Then the external profile: referring-domain trend, link distribution across the site (all to the home page is a common weakness), anchor risk, and, most usefully, the gap between your link profile and the competitors who outrank you for target intents. The output isn't “get more links”; it's specific: which pages need authority, how much, and which existing assets are earning links that internal linking fails to route onward.

Now the four evidence sets collapse into one deliverable. Every finding becomes a task scored on three axes: impact (traffic/revenue at stake, estimated from the affected URLs' demand data), confidence (how sure you are the fix moves the metric), and effort (engineering and content cost). Priority = impact × confidence ÷ effort, an ICE-style score that forces every recommendation to justify its place in the queue.

Two rules make the roadmap survive contact with reality. First, every task gets an owner and an acceptance check, the specific metric or report that will prove it worked (indexed count, position, CWV pass rate, crawl distribution). Second, ship a P0 within the first two weeks: early visible wins are what buy the political capital for the P1 and P2 work.

Run this way, an audit stops being a document and becomes an operating system for the next two quarters of SEO work, every task evidence-backed, priced, and sequenced. That's the deliverable worth paying for, and the standard your own internal audits should meet. For the recurring technical checks between full audits, keep our technical SEO checklist in the rotation.