Brand voice audit: score what you’ve already published

A brand voice guide tells you how you intend to sound. A brand voice audit tells you how you actually sound. This auditor takes copy you have already shipped — a page, a deck, a product catalog — scores it against the voice you defined, and shows which lines pull away from it and how far.

What a brand voice audit is

Traditionally it is a manual exercise: someone senior reads a sample of your content, forms an opinion, and writes a deck. It is slow, it covers whatever fraction of your content fits in a week, and the finding — “the voice feels inconsistent” — is hard to argue with and harder to act on.

The automated version answers the same question with numbers attached: for each of the ten language dimensions that define a voice, how far is this content from where you said you wanted to be? That turns “feels off” into “you set 2nd person and this page is written in 3rd” — a sentence a writer can act on in a minute.

What gets measured

Ten language dimensions — perspective, language type, everyday vs scientific, technicality, emotion, fluency, figures of speech, sales argument, tense and readability. The model measures where the content sits on each; the distance from your target is arithmetic, computed server-side rather than asked for.
Readability, calculated not estimated — Flesch Reading Ease from your own sentence and syllable counts. English only, and skipped entirely on very short passages rather than guessed at.
An overall score and an A–E band — weighted across the dimensions, so a miss on emotion or sales argument counts for more than a miss on tense.
Line-by-line flags — the specific passages pulling the score down, each with what is wrong and a suggested fix.
House rules, if you keep any — phrases you have banned outright are matched literally against the copy and reported, no model judgement involved.

Band	Score	Reading
A	90–100	On voice. Publish it.
B	75–89	Recognisably yours, one or two dials off.
C	50–74	Generic. Could be any brand in your category.
D	25–49	Actively off-voice — usually written to a different brief.
E	0–24	Reads as another brand entirely.

The compliance checker uses the same A–E scale, so a page can be read on both axes at once.

What you can audit

Paste textAn email, a landing page draft, an ad, a product description. 1 credit.
Give a URLThe page is fetched and its copy extracted for you. 1 credit.
Upload a .docxBriefs, decks-as-documents, long-form articles. 1 credit.
Upload a CSVYour whole catalog or content export, scored row by row, worst first — up to 500 rows per run at 1 credit per 25 rows. Pick which columns hold the actual prose; ids and URLs are excluded automatically.

Auditing a store? Catalog audit covers the export path for Shopify, WooCommerce, Magento and Wix — no plugin, no developer.

What you audit against

There are two honest baselines, and they answer different questions.

Your brand voice guide — the voice you chose. The audit then answers “is this on brief?” Start with the generator if you don’t have one yet.
Your own best content — load a handful of pieces you already consider on-voice into the content library and audit against their median instead. That answers “is this consistent with how we already write?”, which is the more useful question for a brand that never wrote a guide.

The audit says which baseline it used in the report, so a score is never ambiguous about what it was measured against. If what you want first is to read a brand’s voice rather than grade it, that is the analyzer’s descriptive mode — free, and it works on any public site.

From finding to fixed

An audit that only grades is half a tool. Under every flagged result there is a Refresh content panel that rewrites the passage towards the same voice description the score was computed from — so the writer and the scorer cannot be aiming at different targets.

The rewrite is shown beside the original. Nothing is applied, stored or published for you.
Every rewrite is screened for risky claims the model may have reached for, and anything it introduced that wasn’t in your original is flagged on the rewrite itself.
On a CSV, it offers the rows the audit marked down, worst first, or the whole table if you tick the box.

Auditing for clients

Agencies run this per client, not per company. A workspace keeps each client’s brand, guide, register and library separate, seats spend from one shared credit pool, and per-client budgets stop one account quietly eating another’s retainer. Seats can be scoped so a freelancer assigned to one client never sees the rest of the roster.

See agency vs tool costs for the comparison.

What it doesn’t do

It does not read images, video or design. Copy only.
Readability is English-only; the rest of the scoring works in any language your brand writes in.
CSV rows are scored holistically rather than dimension-by-dimension — the full deterministic breakdown applies to single pieces.
It scores against your stated voice. It has no opinion about whether that voice is the right one for your market.

Run a brand voice audit Build the guide first Pricing

Credit costs and row caps quoted above are current as of this page’s last update; the rate card on the pricing page is authoritative. Evaluation is an automated drafting aid — flagged lines are candidates for a writer’s judgement, not verdicts.