Copy in your voice, with claims you can prove.
The copy desk keeps your voice, the things you never say, and the claims you can back up. Agents write from it, and a claim without a source can’t be used.
Every figure in an email has a source.
The running-low email says a bag lasts about five weeks. That figure comes from an approved claim, so its check passes. A claim without a source stays a draft and can’t be used.
Brand review rewrote two health claims.
On October 26, your banned phrases stopped two Holiday Blend ads that made health claims before they ran, and brand review rewrote both. Rewriting is within its permission. It went ahead and told you.
- Holiday Blend ad, headline
- BeforeThe healthiest coffee you can drinkAfterRoasted Monday, at your door by Thursday
- Holiday Blend ad 2, headline
- BeforeGuilt-free coffee for better morningsAfterThis week’s roast, in your mug by the weekend
Holiday Blend ads · Brand review agent
From a brief to approved copy
- 1
Your voice, written as rules
“Warm and unfussy, like a good barista on a slow morning.” Then the rules: always name the roast date and where the coffee comes from; never make health claims, use urgency tricks, or name other roasters.
- 2
Every claim carries its source
A claim is a figure or promise plus its source. “9 in 10 pour-over kit buyers order coffee again within two months” rests on kit buyers’ order data. “Subscribers save 20% a year compared with grocery-store coffee” has none, so it stays a draft.
- 3
Reusable copy, with its results
Headlines, body lines, descriptions, calls to action and subject lines live in one library, tagged by use. Once they run, each shows the ads and messages that use it, with impressions, clicks, conversions and cost per conversion.
Reusable copy, with its results - 4
Emails written from the library
The Reorders agent wrote its running-low note from library copy: the subject line, and the claim that a bag lasts about five weeks. The email shows which library pieces it uses.
Emails written from the library - 5
Checked before it goes out
Banned phrases are blocked in copy from agents and people alike, and every figure needs a source. Brand review reads every ad and email against your voice and your approved claims. You see each result on the proposal before you decide.
One claim approved, one refusedOne example, in order
Aug 29 · Maya Chen
Adds “Every bag ships within 4 days of roasting”.
Sep 18 · Maya Chen
Rewords it to “Every bag ships within 3 days of roasting” and updates its source to 2025 order and fulfillment data. The claim is now on version 2.
Oct 25 · Copy desk
Adds “Subscribers save 20% a year compared with grocery-store coffee”. It has no source, so it stays a draft: “Needs a source before it can be used.”
Oct 26 · Copy desk
Adds “9 in 10 pour-over kit buyers order coffee again within two months”, sourced to order data for kit buyers from January to September. It’s approved, and the creative desk used it as the text of the gift-wrapped kit ad, now waiting for your approval.
Oct 28 · Reorders agent
Proposes the running-low email with a library subject line and the five-week claim. Its checks find no banned phrase, and every figure has a source.
What the copy desk enforces
The content rules run on every ad and message an agent proposes. Some you set; some are always on.
A claim needs a source before it can be approvedFixed
Until someone adds the evidence, a claim stays a draft and can’t go into an ad or email.
Brand review reads every ad and emailCompany setting
On by default. Before an ad or email is proposed, the brand review agent checks it against your voice, your banned phrases and your approved claims library, and passes it, rewrites it or holds it. A figure that no sourced claim backs is held. Its verdict is on the proposal: a rewrite reads Adjusted, a hold stops it. Turned off, the check reads Off.
Every figure needs a sourceCompany setting
A figure in an ad or email without a source stops the proposal. If no check has run, the proposal says “Not checked yet” instead of showing a pass.
Banned phrases stop a proposalCompany setting
They’re matched word for word, in copy written by agents and by people. This roaster bans “healthiest”, “guilt-free” and “best coffee in the world”.
No discounts, if you say soCompany setting
Set discounts to none, the default, and an ad or email that offers one, such as a percentage off or a free month, is stopped.
Every email carries an unsubscribe link and your addressAlways on
Checked in each email’s own words, against the postal address in your settings. An email missing either is stopped. The law and the platforms require it, so it can’t be turned off.
Only owners and admins change the voiceBy role
The voice, the always and never rules, and the audience description define the brand. Everyone else can read them.
When a line isn’t right
What happens when an agent writes a figure you can’t prove?
A new claim stays a draft until someone adds a source, and a draft can’t go into an ad or email. A figure without a source stops the proposal.
What happens when you change your voice?
Agents read the voice before they write, so new drafts follow the change. Copy you already approved stays as it is until you edit or retire it.
What happens when an approved line stops being true?
Edit it or retire it. A retired line can’t go into anything new, and a launch that still uses it is refused.
Where does the voice come from at the start?
When you set up, it’s drafted from your website and you review it. After that, you edit it like any other copy.
Can I write copy myself?
Yes. Write it yourself or brief the copy desk. Everything goes into the same library, marked with who wrote it.
What counts as a source?
Whatever the claim rests on, named so a person can check it: order data for a date range, or help desk response times for a quarter. Each claim shows its source beneath it.
Say it your way. Prove what you claim.
Your work email is all we need to start.