Build the inventory
Pull a full list of indexed URLs (from the CMS, a crawler like Screaming Frog, or Search Console's coverage report) along with basic performance data: organic traffic, average position, and last-updated date for each.
For sites past a certain size, group pages by topic or template first — auditing a blog category or a product-page template as a batch surfaces patterns a page-by-page review misses, like an entire section that's stale or an entire template missing schema.
Triage into four buckets
Keep as-is: pages performing well with no obvious gaps. Update: pages with the right topic but outdated facts, thin content, or missing structure — usually the highest-ROI bucket, since the page already has some authority and history.
Merge: near-duplicate pages competing with each other for the same query, better combined into one stronger page with a redirect from the weaker one. Remove or noindex: pages with no traffic, no backlinks, and no strategic reason to keep them — indexed clutter with no chance to help, dragging down the average quality signal.
Don't skip the AI-visibility check
Alongside classic metrics, check whether each important page could plausibly be the one an AI assistant cites — is there a direct, extractable answer near the top, backed by verifiable specifics and matching schema markup.
A page can rank fine in classic search and still be invisible to AI answer engines simply because the actual answer is buried in paragraph four; a content audit is the natural place to catch that, since you're already reading the page closely anyway.