Content quality audit: six checks, one decision, no readability grade.

An SEO audit asks whether a page can be crawled, indexed and loaded. A content quality audit asks the harder question: whether the page deserves to exist. Below are the six checks we score, the five verdicts an audit ends in, and how to run the whole thing on one draft.

Run it on a draft
1 free analysis a day · no signup
your draft isn't stored
Inventory→Six checks→One of five verdicts→Fix or cut→Re-audit
What a content quality audit answers

An SEO audit measures access. This measures substance.

A technical audit tells you the page is crawlable, indexed and fast. All three can be true of a page that says nothing a reader can use. The quality audit works on the other side of that line.

Did it do its job? Purpose fulfillment

The page answers the question its title set up, stays inside the scope its format promises, and leaves a reader able to act without a follow-up question.

Can someone use it? Reader experience

Specifics rather than generalities, the conditions that make the advice hold, terms defined where they appear, one idea per sentence.

Is it carried? Claim support

Numbers a reader might doubt point to where they came from, claims carry the reason they hold, and the piece says where its conclusions stop.

Those three fold into one number. Three more reads run beside it — AI-tone phrasing, evidence needs and publishing readiness — and never fold in. A style flag cannot drag a solid draft down, and a high score cannot hide a claim that has no source.
Where an audit has to end

Every audited row ends in one of five verdicts.

A scoring sheet without a decision column is a report nobody acts on. The content audit toolkit published in New Brunswick's provincial digital standard assigns exactly five outcomes per row, which is a good schema because each one is a different amount of work rather than a different opinion.

Keep Clear job, still accurate, doing it

Protect the page and check it again next cycle. Most audit templates stop here and call the row done.

Improve Right job, decayed execution

Rewrite the part that fails, not the whole page. Outdated figures and a stale opening usually carry the damage.

Merge Two pages, one question

Competing pages split whatever attention the topic had. Combine the strongest parts and redirect the rest.

Remove No job, no audience, nothing worth merging

Take it out of the index. Pages that carry links or traffic get a redirect instead of a removal.

Create A gap the inventory cannot cover

The only verdict that adds work. It should be the rarest one on the sheet, not the first reflex.

On a single draft the same five collapse into four things you can do in an editor: ship it, fix one section, cut it, or add the evidence it is missing.
One draft or the whole site

The same rubric works at two scales.

A full audit is spreadsheet work: one row per indexable URL, with clicks, impressions and position from Search Console, then a verdict per row. It answers what to do with three hundred pages. A draft audit runs the identical checks on one piece before it ships, which is the only point where a fix is still cheap.

Full-site audit Inventory, data, verdict per row

Finds decay, cannibalisation and pages that never had a job. Quarterly at most, or the sheet becomes the work.

Draft audit One piece, six checks, once

Catches the problem before the page has a history to decay. Cheapest moment in the entire lifecycle.

Most audits we see run on pages that already shipped, because that is when the numbers arrive. The draft in front of you has no numbers yet, and it still has a verdict.
Running it on one draft

Paste, read three answers, decide.

The checks above are the ones the tool scores, in the same order. Nothing is rounded into a general impression, and every point lost is attached to a sentence you can edit.

1. Paste the draft No account, no card

One free analysis a day, and the draft is not kept after the run.

2. Read the three answers Publish, cite, read-as-AI

Can I publish this, will AI search cite it, and will it read as AI-written. Three decisions, three plain answers.

3. Fix the two worst sentences Article, paragraph, sentence

Each problem comes with a fix at the size it belongs to, so a paragraph problem is not answered with a rewrite of the piece.

4. Re-score Did the edit move the number

If the score does not move, the edit did not address the check that was failing.

Run a content quality audit on your draft →  ·  See a real scored report (PDF)

Already using an optimization tool? We wrote the same comparison for Surfer, Clearscope and Frase. Each one says plainly who should stay where they are.

What this is not

Three claims we deliberately don't make.

The category is full of tools that promise more than they can show. These three are the ones we hold back on, and why.

Not an AI detector Flags are phrasing-level

The tone pass names specific habits and suggests an action for each. It does not claim to identify what wrote the text, and the flags never enter the score.

Not a readability grade Flesch, passive voice, sentence length

Those measure how a sentence is built. A clean grade sits happily on a draft that says nothing a reader can use.

Not a ranking promise Different question entirely

Optimisation scores correlate loosely with position — 26% for Surfer and 17.5% for Clearscope in independent testing. We answer whether the piece is ready, not where it will land.

A note on earlier scores

Runs recorded before November 2026 used a different model.

Those runs were measured on an earlier four-dimension model with fixed weights — information gain, specificity, source compliance and argument depth. Those weights were internal and are no longer used. The 29-draft dataset published in September 2026 was measured on that model, so its numbers describe the old rubric, not the three named dimensions above. Runs filed from November 2026 onward use the current model, and every record on /receipts carries its date so the two are never mixed.

Questions about content quality audits

What is a content quality audit?
A page-by-page review of what a piece of content is supposed to do, whether it actually does it, and what to do about it next. It ends in a decision per page rather than a report: keep, improve, merge, remove, or create. It is a judgement about substance, not about whether a page can be crawled.
How is it different from an SEO audit?
An SEO audit measures access: can the page be crawled, indexed, and loaded quickly. A content quality audit measures substance: does the piece answer the question its title set up, can a reader act on it, and do its claims carry their evidence. A page can pass every technical check and still say nothing.
How long does one take?
Across a whole site it is spreadsheet work: one row per indexable URL, with clicks, impressions and position exported from Search Console, then a decision per row. That runs over weeks, not hours. On a single draft it is much shorter: our own recorded run scored a 1,754-word draft in 59 seconds.
Should an audit happen before or after publishing?
Both, for different reasons. Before publishing is the last moment a fix is cheap and the first moment anyone can still change the outcome. After publishing the audit becomes decay detection: which pages lost their traffic, and whether that is worth rewriting, merging or removing.
Does a high score guarantee rankings?
No, and we do not sell it that way. A score tells you whether the piece is ready to go out. Ranking also depends on competition, links and demand, none of which a draft-level check can see.
Does this replace an AI detector?
No. The tone pass flags specific phrasing habits and suggests a fix for each. It does not claim to identify what wrote the text, and those flags never enter the main number.
Featured on Whatsthebigdata