Everything you need to get from “here’s my website” to a documented brand voice, scored content and a compliance pre-screen — and how to read every number in the reports.
The tools build on each other. Each step unlocks the next, which is why sidebar links start greyed out until the step before them has data for your brand.
Start at New analysis. Enter your website and answer a short survey. The AI reads your existing content and produces a voice profile: where your brand sits on ten language dimensions (formality, technicality, emotion, perspective and more), plus tone-of-voice traits. This becomes the baseline everything else measures against.
From the profile, generate a full brand voice guide: tone description, dimension settings, do/don’t writing examples, and ready-to-use AI prompts for ChatGPT, Claude or Gemini so any tool writes in your voice. Share it with your team or clients via its link.
Paste text, drop in a URL, upload a Word doc, or batch-check a CSV on the Evaluate page. Every piece is scored against your voice profile, dimension by dimension, with an overall match score, an A–E band and concrete rewrite suggestions.
The Compliance checker screens marketing copy for risky claims — superlatives, guarantees, health and financial claims, pricing promises and more — against a 14-category ruleset with per-market (EU / UK / US / AU) overrides. Each flag shows the exact quote, the issue and a suggested fix. See a sample report.
The headline number (0–100) is how closely the content matches your brand voice profile. It maps to a letter band; both evaluation and compliance use the same thresholds:
| Band | Meaning | Score |
|---|---|---|
| A | On-voice — publish as is | 90–100 |
| B | Close — minor polish | 75–89 |
| C | Mixed — noticeable drift, revise | 50–74 |
| D | Off-voice — substantial rewrite | 25–49 |
| E | Doesn’t sound like your brand | 0–24 |
Your voice lives on ten dimensions — perspective (1st/2nd/3rd person), expository vs persuasive vs narrative, everyday vs scientific language, technicality, emotiveness, sentence rhythm, figurative language, rational vs emotional argument, tense, and readability. The report measures where the content sits on each dimension and scores the distance to where your profile says the brand should sit. A low score on one dimension tells you exactly what to fix — e.g. “too technical” or “switched to third person” — rather than a vague “doesn’t sound right”.
The AI is only asked to measure the content on each dimension. The matching and the overall
score are then computed by fixed arithmetic on the server (you’ll see
score_method: deterministic in the report). That means re-running the same text against
the same profile gives the same result — scores are stable enough to use as a quality gate in a
review workflow. Dimensions the brand cares most about (emotional register, sales argument) carry
extra weight in the overall. If too few dimensions can be measured (very short or unusual content),
the report falls back to holistic AI scoring and says so. CSV rows are scored holistically.
The Flesch Reading Ease number is pure arithmetic computed on the server — never an AI estimate. It’s calculated for English text of 10 words or more; otherwise it shows as blank rather than a made-up number.
If you’ve built a content library (see below), you can opt to score against the median
of your approved, published content instead of the survey profile —
baseline: corpus in the report. That answers “does this match what we actually
publish?” rather than “does this match what we said we sound like?”
The CSV path was built for store exports: Shopify, WooCommerce, Magento / Adobe Commerce and Wix all export products to CSV, and one run scores every description row by row. The catalog audit guide walks through the export, the column picker and the per-25-row pricing.
The verdict is decided by the worst flag found — one critical flag makes the whole piece critical, no matter how clean the rest is. That’s deliberate: one prohibited claim is enough to create risk. The 0–100 compliance score and its A–E band are clamped so they can never contradict the verdict (a piece with a critical flag can’t score above 35, high above 65, review above 85).
the best tool on the market”cuts review time by 40%”| Level | Meaning | Action |
|---|---|---|
| CRITICAL | Language that is prohibited or highly likely to breach rules in the selected market (e.g. absolute health claims, guaranteed investment returns) | Do not publish — rewrite |
| HIGH | Claims that need substantiation you may not have (superlatives, “#1”, specific percentages, environmental claims) | Substantiate or soften |
| REVIEW | Wording worth a second look — puffery, vague comparisons, promises that read stronger than intended | Judgement call |
Severity is enforced by the ruleset’s jurisdiction resolver on the server: if the evidence matches a prohibited pattern for your market, the flag is critical regardless of how mildly the AI phrased it.
The grid at the bottom lists every rule category the engine checked — including the ones that found nothing. A clean category is a result, not an omission: “no pricing flags” means pricing claims were checked and passed, so you can show the report to a stakeholder as evidence of what was screened.
If your workspace keeps an approved-claims register (claims your legal or compliance team has signed off, with an expiry date), matching flags are annotated as approved and downgraded to review. Three rules protect you here: a flag is never removed (you always see what was matched), critical/prohibited findings can never be downgraded by the register, and register entries expire so stale approvals don’t linger.
Credits are prepaid units — 1 credit = €5 of output, VAT-inclusive. Credits are valid for 365 days from purchase and work across every tool. Current rates:
| Output | Credits |
|---|---|
| Brand analysis / voice profile (the funnel step) | Free |
| Brand voice guide (full, in-depth) | 5 credits · €25 |
| AI-tailored platform prompt (ChatGPT / Claude / Gemini) | 1 credit · €5 |
| Evaluate a URL, Word doc or pasted text | 1 credit · €5 |
| Evaluate a CSV | 1 credit per 25 rows |
| Claims & compliance review | 1 credit · €5 |
| Content library upload (per document) | 1 credit · €5 |
| Refresh content (rewrite onto your brand voice) | 1 credit per 2,000 characters |
| Register & library management (list, revoke, archive) | Free |
If a paid run fails — a fetch error, a timeout, a corrupted generation — the credits are automatically refunded. You are only ever charged for output you received. See the refund policy.
A per-workspace list of claims your reviewers have signed off (each with a reference and a mandatory expiry). Compliance reports annotate matching flags instead of making your team re-litigate the same approved phrase every week. Add entries straight from any flagged claim in a report, or curate the whole register — including what’s expiring — on the Claims register page under Assets in the sidebar.
Upload your best published content once (1 credit per document). Each document gets claim extraction and dimension scoring at upload, building a corpus baseline — the median voice of what you actually publish. Evaluations can then score against that corpus, and the tool warns you if your written guide has drifted from your published reality. Manage documents on the Content library page under Assets in the sidebar; the choice of baseline stays on the evaluation page.
The team plan (€49/month) adds five invited seats to your workspace. Seats share your credit balance, approved-claims register and content library. Each seat gets a role: member runs evaluations, legal also maintains the claims register, brand the content library, and manager both. You can group work by client, restrict each seat to its assigned clients, and set soft per-client credit budgets — useful for agencies running the tool across a book of accounts. Manage it all from the workspace page.
Run checks where the work already happens. Every app is a thin client over the same actions as the web tool: the same credits, the same rules, the same reports. Connect each one from its page under Apps in the sidebar.
Brand Voice Checker for Slack — run checks where the
“can we say this?” question actually gets asked. With the
Brand Voice Checker app in your Slack workspace, type /compliance followed by your
copy for a claims & compliance check, or /brandvoice for a brand voice check —
or use the Check compliance shortcut in any message’s ⋯ menu. A check
takes about 15–20 seconds, spends the same credits as the web app, and the result is
visible only to you.
Connecting takes one minute per person:
/brandvoice command
does not exist until it is there, so nothing below this step can work. The
connect Slack page carries the install link./brandvoice link YOUR-CODE. Slash commands must be typed —
pasting the whole line posts it as an ordinary message instead of running the command.Manage or revoke Slack connections any time on the same connect page — revoking a link cuts that person’s Slack access immediately, and disconnecting the workspace revokes everyone in it at once. The Slack app is currently in early access: it isn’t publicly installable yet, so if you’d like it in your workspace, get in touch.
Right-click any selected text in Chrome to run an evaluation or compliance check without leaving the page you’re reviewing. Uses your existing account token — results render in a panel on the page, and checks spend the same credits as the web app. Install it from the Chrome Web Store, then choose “Connect account” in the extension to pair it with your account.
Check copy in the document you’re drafting: claims & compliance or brand voice, on your selection or the whole document, with results in a sidebar. It never edits your document — suggested rewrites are shown for you to apply, and it reads the document only when you press a button.
The add-on isn’t published to the Google Workspace Marketplace yet, so installing it is a one-time copy into your document’s script project. It takes about two minutes:
Code.gs with the contents of Code.gs from
the gdocs-addon folder of our repository.Sidebar, and
paste in Sidebar.html.ReportCommon, and paste in ReportCommon.html. Both names have to
match exactly — the sidebar looks them up by name, so a typo shows up as a panel that
never draws rather than as an error.The connection key is stored against your Google account, not the document, so people you share the document with don’t inherit your credits — each person connects their own account. A script added this way lives in that one document; for several documents, repeat the steps or ask us about a domain-wide install.
Check posts and pages from inside the editor — and, if you want it, stop a post publishing until it passes. This is the only place a check can gate anything; everywhere else it advises and you decide.
A Brand Voice Checker box appears in the editor sidebar with two buttons: Check compliance (works on any content) and Check brand voice (needs a brand voice domain set). Results appear in the box, worst flag first, with the phrase and a suggested rewrite. Checks read what’s in the editor including unsaved changes, so what you see is what gets checked. Each costs 1 credit.
The gate has three modes. Content fails it when it has never been checked, when it changed since the last check, when the last compliance verdict was hold, or when the last score is under the minimum you set.
| Mode | What happens |
|---|---|
| Off | Checks are available; nothing is ever blocked. |
| Warn (default) | A notice appears when content would fail. Publishing goes ahead. |
| Block | Publishing reverts the post to draft, with the reason shown. |
Start on Warn. The gate reads the stored result of your last check, never the network — so it can’t block you because our service was slow, and publishing never costs a credit. Blocking is enforced on your server, not in the browser, and administrators are exempt so a site can never lock itself out.
The connection key is site-wide: everyone who edits that site spends the connected account’s credits, which suits a publishing gate but means you should only connect a site you control. It never reaches anyone’s browser — WordPress calls us from the server. Revoke it any time on the connect WordPress page.
If your site cannot reach the plugin directory, the manual route still works:
download the wordpress-plugin folder from our repository, zip it, and use
Plugins → Add New → Upload Plugin.
Longer background pages that sit outside the tool itself. Most are available in English only.
Why are some sidebar links greyed out?
They need data that doesn’t exist yet for the selected brand. “Voice guide” unlocks after you generate a guide; “Evaluate” unlocks after an analysis has produced a voice profile. Start at New analysis — see how the tool works.
Do credits expire?
Credits are valid for 365 days from purchase. Within that window they work across all tools and are shared with your team seats, if you run a workspace. See Credits & costs for the full rate card.
What happens if a run fails?
The credits are refunded automatically and the response tells you so. You’re only charged for completed output — see the refund policy.
Why is the readability number blank?
Flesch Reading Ease is computed only for English text of at least 10 words. Below that (or for non-English content) the formula isn’t meaningful, so the tool shows nothing rather than guessing. More in Reading an evaluation report.
I re-ran the same text — why is the score identical?
By design. Tone matching is deterministic arithmetic against your profile, so identical input produces identical output. If the score changes, the content (or your baseline) changed. More in Reading an evaluation report.
Can I dismiss a compliance flag?
Flags are never deleted. If a claim has genuinely been signed off, add it to the approved-claims register — future reports will annotate it and downgrade it to review, while keeping it visible. Critical/prohibited findings can’t be downgraded at all. More in Reading a compliance report.
Which markets does the compliance checker cover?
The ruleset carries market overrides for the EU, UK, US and AU regimes. When you select several markets, the strictest applicable rule wins. More in Reading a compliance report.
Is the compliance checker legal advice?
No. It’s an automated pre-screen to catch risky language before it reaches your reviewers. Final sign-off on high-risk claims belongs with a qualified person for your market.
How does this compare to hiring an agency?
We wrote up an honest comparison — including where an agency is still the right call — with typical market rates side by side: Why use this tool?