Most content audit spreadsheets are excellent at proving that pages exist. They collect word count, status code, title length, traffic, backlinks, and a vague quality score. Then somebody colours the rows and the site remains unchanged.

The missing field is the page's job. Without it, low traffic can condemn a necessary policy page, high traffic can protect irrelevant visits, and two pages serving the same intent can both look healthy in isolation. Audit the page as part of a search and business system.

The minimum worksheet

ColumnWhat to record
URLThe canonical URL being decided
Page jobThe reader decision or task this page should own
Demand evidenceQueries, customer questions, referrals, navigation use, or another reason the job matters
Current evidenceClicks, impressions, conversions, links, indexation, crawl state, and qualitative use
SubstanceWhat the page uniquely knows, demonstrates, calculates, or enables
OverlapPages competing for the same job or repeating the same proposition
ConstraintThe most important reason the page cannot do its job
ActionKeep, improve, merge, redirect, remove, or investigate
DestinationSurviving URL when merging or redirecting
Owner and checkWho implements it and what result will be verified

Collect evidence before assigning actions

Export the canonical URL inventory from the site and crawl. Join it to Search Console page and query data, analytics or conversion data, known backlinks, sitemap membership, and index status. Search Console's Performance report can group by page and show the queries for which Google displayed it, but its tables omit anonymised and some lower-volume rows. Treat the export as evidence, not a complete demand map.

Then inspect the page itself and its neighbours. A crawler cannot tell you whether an explanation contains original evidence, whether a product comparison is fair, or whether three URLs answer the same reader question. Those decisions require reading.

Use actions with strict meanings

Constraint rule

Write one primary constraint per improve decision. “Needs SEO” is not a constraint. “Google selects the category page as canonical,” “the article lacks evidence for its central comparison,” and “three pages divide the same query set” are constraints somebody can act on.

Decide in clusters, implement in dependencies

Review overlapping pages together so each job has one owner. Then sequence implementation: create or improve the surviving target, update internal links, add the redirect if a genuine replacement exists, remove the obsolete source, update sitemaps, and verify the resulting status and canonical.

A broad topic plan is stronger after this consolidation because its gaps are real gaps rather than spaces between duplicates. A performance fix should likewise follow the content decision; there is little value in perfecting a template for pages the audit will remove.

Verify the decision, not merely the edit

After implementation, crawl the affected set, test redirect targets, inspect representative URLs, and confirm internal links and sitemaps point at the surviving pages. Later, review query and page performance against the reason for the action. A merge intended to consolidate overlapping queries is assessed differently from an update intended to improve conversion.