Content

The Content Audit That Produces Your Next Ideas

Run a content audit as a demand map. See which gap, near miss or reader question earns the next brief, and how to rank all four signals.

By David Jubé · Jul 4, 2026 · 14 min read

TDM Insights website homepage with audit forward message and growth focus.

Table of Contents

Every content audit produces two lists, and the more valuable one is not the list of pages to fix.

Rank audit findings by demand already proven, not by the size of the hole.

Biggest gap wins is the instinct, and it is the instinct that burns a quarter. A topic nobody has searched for is still a gap on a coverage map, while a live page sitting at position eleven with real impressions is demand your library already reached and failed to close.

Signal strength decides the order. Sorted that way, the same export stops producing a repair list and starts producing a queue of briefs, each one ranked by how much proof already sits behind it.

Key takeaways

Sort an audit by how much proof sits behind each finding, not by how large the coverage hole looks.
Near misses carry measured proof: a live page ranking at positions eight to fifteen for a query it never resolves.
Repeated reader questions carry observed proof, and a query report will never surface them.
One URL ranking for two different intents is two briefs compressed into one page, not a page that needs tightening.
Coverage gaps carry inferred proof only, which is why they go last in the queue rather than first.

Sort the Content Audit by Demand, Not by Defect

An audit is a demand map. Read the export for what people already ask your library and fail to get, and it returns the next set of briefs instead of the next set of repairs.

That is the answer, and it comes out of a sort rather than a search, which is why the same export can produce either list.

The maintenance pass is real work. Thin pages get merged, dead pages get retired, and a library over a year old earns that afternoon. Moz’s walkthrough of a content audit and Animalz on auditing a library strategically both lay that pass out cleanly.

The maintenance sort asks which pages are broken. The demand sort asks which questions the library gets close to answering and never finishes.

That second sort is where the next quarter’s briefs come from.

A finding is any place the audit shows demand the library does not meet. Four kinds come out of the exports:

A near miss, where a live page ranks for a query it raises and never resolves.
A repeated reader question, asked somewhere your analytics cannot see it.
A split-intent URL, where one page is pulling two different query clusters.
A coverage gap, where the library has no page attempting the query at all.

Each one carries a different weight of proof, and the weight is what decides the order.

This is the point in the AI content quality set where the Purpose and Evidence Tests turn outward. The pillar applies both tests to a single draft before it publishes, and here they run across a whole library, where the output is a list of pages that do not exist yet.

Coverage Is the Wrong Axis for Deciding What to Write Next

Coverage is the natural axis, and it is worth respecting why. A topic map with a visible hole in it reads like an instruction, and filling the largest hole feels like the highest-value move on the board.

A hole in a coverage map is a hypothesis.

A weak ranking is a measurement.

Both can be true at once. Only one of them has already been tested against real searchers, and that difference decides which deserves the first writing day of the quarter.

Publishing more is not the lever: in Orbit Media’s 2025 blogger survey, 21 percent of 808 respondents reported strong results from their content, a share the same survey reports has been trending down for seven years.
Publishing and reaching anyone are separate events: Backlinko’s 2019 study of 912 million posts with BuzzSumo found 94 percent of blog posts have zero external links.

Picking better beats publishing more, and an audit is the only place the picking gets evidence attached to it.

So the question the audit should answer is not where the map looks thin. It is which unmet demand already has proof behind it, and how much.

Pull Three Exports Before Judging a Single Page

Three exports carry every signal this method uses, and none of them costs anything. Pull all three before reading a single page, because reading pages first biases the sort toward whatever you happened to open.

Search Console Performance, dimensioned by query and by page, over three to six months, filtered to the content section.
Every live URL with its title, its H1 and its target query, sorted by publish date.
A question log drawn from comments, support tickets and notes from calls.

Search Console, Query by Page

This one file carries both measured signals: pages ranking weakly on a query they never resolve, and single URLs ranking for more than one intent.

Impressions prove the demand exists, and position measures the distance still left to travel.

Every Live URL, Sorted by Publish Date

Sort the list chronologically rather than by topic. Missing rungs show up faster that way, because you can see where the library climbed and where it skipped a step.

Template-driven walkthroughs earn their keep at this step. Conductor’s audit template and Surfer’s eight-step audit both give a column set worth copying, and the one column to add is the signal each row eventually produces.

The Question Log a Query Report Cannot See

Search Console records what people typed into Google. It does not record what a reader typed into your comment box. It does not record what a prospect asked on a call, because by then they had given up on the library and asked a person instead.

Export comments, tickets and call notes, then filter for question patterns: how do I, what is, can I, why does. HubSpot’s audit walkthrough and Siege Media’s SEO content audit both start with inventory, and this third export is the piece a page-level inventory does not produce on its own.

Pull them in that order, because Search Console sorts fastest, the URL list is quick to scan once you know which queries the library is already fighting for, and the question log takes the longest to read.

Demand Signal Strength Is the Real Driver Behind a Next Brief

Signal strength is how much independent proof exists that somebody wants an answer, measured before you write it. It comes in three grades, and every finding lands in one of them.

Measured: Google recorded the demand and your library was already in the result set. Impressions and position are the evidence.
Observed: a person asked, repeatedly, somewhere you control. Repetition across separate people and separate months is the evidence.
Inferred: nobody has asked yet, as far as your data shows, and the evidence is a pattern in your own topic map.

Rank by grade first and by size second. A measured finding on a small query beats an inferred finding on a large one. The small one has already cleared the question that sinks a new page before it starts: does anyone want this.

Impressions Are Not Clicks, and That Changes What a Rank Gain Buys

Measured demand is proof of interest, not a promise of traffic. SparkToro’s 2024 zero-click study put open-web clicks at 360 for every 1,000 US Google searches, and the study now carries its own note that the data may be out of date.

That does two useful things to the queue: it keeps a near miss ahead of a guess, since interest is still interest, and it stops anyone from costing a rank gain as though every impression turns into a visit.

Decay belongs in the same read, pointing the other way. Ahrefs on content decay describes pages losing traffic they used to hold, which is a defect signal rather than a demand signal, so it goes on the maintenance list where it does its work.

The Matcher: Four Signals, Ranked by the Proof They Carry

Four findings come out of the three exports, and this is the order they go in.

Signal	What proves it	What it points to writing	Queue position
Near miss	Live page at positions eight to fifteen with impressions on a query it raises and never resolves	Second page built to finish the job the first one started	First
Repeated reader question	Same question from separate people across separate months, in comments, tickets or calls	Brief with demand proven by repetition rather than by rank	Second
Split-intent URL	One URL ranking for two query clusters that serve clearly different intents	Two briefs, one per intent, each with its own angle	Third
Coverage gap	Missing rung between two published pages, with no impressions attached to it yet	New page built from nothing, validated before it is committed	Fourth

Near misses go first because they are the cheapest confirmed demand in the library. The page already ranks, which means the topic sits inside your reach, and the missing part is usually one question the page raised and moved past.

Reading a Near Miss the Way Its Searcher Did

Open the page and look for the thread it drops. A term mentioned but never defined, or a method cited but never walked through, is the next brief rather than a patch on the old page.

Updating the old page instead is the reflex, and sometimes it is the right call. Ahrefs on what a real republish involves and Animalz on refreshing old content both draw the line at whether the page gains new substance. A near miss usually needs a page of its own rather than new paragraphs in the existing one.

Split intent is the finding people mis-sort more than any other. One URL pulling two query clusters looks like a page that needs tightening, and Moz on keyword cannibalization sets out why the fix is structural rather than editorial.

Coverage gaps go last, and they still go on the list. Semrush’s guide to content gap analysis covers the mechanics of finding one. The discipline is to validate that gap against a live result page before committing a brief, and reading the SERP first sets out how.

Running the Matcher on a Hypothetical Library

Take a hypothetical two-person studio with sixty published posts and one writing day a week. The three exports return nine findings:

Four coverage gaps, none of them carrying impressions yet.
Three near misses, all live and ranking between eight and fifteen.
One split-intent URL pulling both a beginner query and a pricing query.
One question three separate prospects asked in the same quarter.

Sorted by coverage, the studio writes the largest gap first, a topic with no page and no measured demand behind it.

Sorted by signal strength, it writes the three near misses first, then the repeated question, then the split, and the largest gap lands in month three with a live result page checked against it.

Same nine findings, same writing day. The first eight weeks now go to demand that is already proven.

Nine Findings, No Agreement on Which Goes First Sorting is where an audit’s value concentrates, and it is also where a queue quietly reverts to whatever felt urgent that week. Send me your export and I will hand back the ranked queue with the evidence attached to every row. Book a free diagnosis

Filling the Biggest Gap First Is the Most Expensive Mistake

The biggest coverage gap is the finding with the least proof behind it and the most work attached to it. It needs a page built from nothing, with no existing rank, no internal links pointing at it, and no measured evidence that the question gets asked.

It is also the finding that feels the most like progress, which is why it keeps winning the argument.

Meanwhile the near-miss page sits at position eleven, already indexed, already earning impressions, one brief away from finishing what it started. The cost of the wrong sort is not the page you wrote. It is the quarter the confirmed demand spent waiting.

The rewrite trap: a near miss gets read as a rewrite request, the old page gets more words, and the question it raised still has no page of its own.
The template trap: sorting by demand does not license a page per gap on a template, which is the line scaled content abuse draws in this set: volume without page-specific value fails on purpose, whatever produced it.

Both come from the same habit, which is judging a finding by how it feels to fix rather than by what evidence sits behind it.

Naming the evidence inside the brief is the fix. It is why a content brief that carries its own evidence is worth the extra ten minutes at the top.

What Good Looks Like When an Audit Runs Forward

Good looks like a quarter’s queue where every row names three things:

The signal it came from, in the matcher’s own language.
The evidence behind it, as a position, an impression count or a repetition count.
The page that surfaced it, so the new brief knows what it links back to.

Nothing in that queue is there because a map looked thin.

Cadence follows. Run the full pull quarterly across the whole library, since a missing rung often sits between a page you checked and one you skipped, and keep the question log running as a standing filter between passes.

By the time the next quarterly pull runs, the question log has done half the sorting for you.

The library then compounds in one direction. Ordering an archive before adding to it is the companion move, and an old library read correctly often beats the same effort spent on a brand new topic.

Where an audit returns only coverage gaps, the library is too young to have measured demand, and the case calls for a cold start cluster rather than a bigger audit.

The inward-facing half of the same export still deserves its own list. Winning back rankings you lost is real work, and it belongs on a separate queue sorted by decay rather than by demand.

Two lists, one afternoon, and only one of them tells you what to write.

Frequently Asked Questions

What is a content audit, in one working sentence?

A content audit is a structured read of everything a library has published, scored against what readers want. Run for maintenance it produces thin pages to merge and stale pages to retire. Run for demand it produces the queries and questions the library has not answered yet.

How do I tell a content gap from a near miss?

Check whether a page already ranks. A near miss is a live page sitting around positions eight to fifteen for a query it raises and never resolves, so the demand is measured. A gap is a query with no page attempting it, so the demand is inferred from your topic map rather than recorded.

Which audit finding should I write first?

Write the near misses first. They carry measured proof that the query is inside your reach, the page is already indexed, and the missing piece is usually one unanswered question. Repeated reader questions come second, split-intent URLs third, and coverage gaps last, because their demand is still unproven.

Do repeated reader questions really count as demand?

Yes, when the repetition is real. One person asking is a data point. The same question from separate people across separate months, in comments, tickets or sales calls, is demand a query report cannot record, because the person asking had already given up on finding the answer and asked a human.

When should the audit run again, and on how much of the library?

Run the full pull quarterly across the whole library rather than a sample, since a missing rung often sits between a page you checked and one you skipped. Keep the question log running as a standing filter between passes, so observed demand never waits a full quarter to be noticed.

Can I run this without buying a tool?

Yes. Search Console’s Performance report, a spreadsheet of your live URLs and titles, and an export of comments or tickets cover every signal the method uses. Paid tools speed up the gap analysis and the crawl, and none of them is required to produce the ranked queue.

Continue Reading

More From This Series

More from TDM Insights

Explore TDM Insights Topics