Ship the model output
The grammar can be clean. The sample still pairs Black Friday, Christmas, and a named faith in one sentence, and the German button is 53 characters. Those problems show up after the page is live.
Product
Which lines can ship, and which still need a person
The score starts at 100. A cultural clash takes 24 points off, a stiff phrase takes 10, and a button that will not fit takes 8. The built-in sample lands at 18. You can send the marked draft to a native reviewer, or keep editing the lines the list already named.
- Serious
- −24
- A holiday and a faith in one sentence
- Moderate
- −10
- A stiff phrase from a fixed list
- Length
- −8
- A button label over 28 characters
Three ways the review gets expensive
Scan the draft before you publish it, and before you pay someone to read all of it.
Pay for every word first
At the spec rate, English is $0.03 a word and German, French, Spanish, and Japanese are $0.06. A 400-word English listing is $12 before anyone has seen whether the problem is five phrases.
Stop after spellcheck
“Delve into”, “game-changer”, and a label longer than 28 characters can all be spelled correctly. This list marks them. A spellcheck leaves them in.
The rates are from the product spec. This build does not charge a card. The price list separates the monthly word cap from the per-word edit.
The sample scores 18
5 moderate marks at 10 each, 1 serious at 24, and 1 length mark at 8. That is 82 points off 100. Under 60, the band is high risk.
Source
Store copy · English (US). Same text as the sample report.
In this modern era, we delve into analytics that unlock growth. Furthermore, it is important to note that the platform is a game-changer.
This Black Friday offer is the perfect Christmas gift for our Muslim customers.
Primary button (de-DE): Kostenlose Testversion jetzt ohne Kreditkarte starten
“Unlock” sits inside “that unlock growth”. The longer match is the one that counts, so the ledger has seven rows.
Each mark opens the same card
Original, why it matched, and one suggestion. In the spec, Free only shows the marks, and writing the suggestion back is Pro. This demo still has Replace this, so you can watch the line change. The payment dialog does not charge a card.
Cultural clash
Serious −24
Original
This Black Friday offer is the perfect Christmas gift for our Muslim customers.
Why
A religious holiday and a specific faith are in the same sentence. That pairing can get a listing reported. Keep the sale, drop the religion.
Suggestion
Our late-November sale is open to shoppers in this market.
AI phrasing
Moderate −10
Original
delve into
Why
“Delve” is a hallmark AI verb. Use a plain verb.
Suggestion
look at
UI overflow
Length −8
Original
Kostenlose Testversion jetzt ohne Kreditkarte starten
Why
This label is over 28 characters. German and French buttons often clip at this length.
Suggestion
Kostenlos testen
The other four moderate marks on the sample are the same kind of card: “In this modern era, we”, “that unlock growth”, “Furthermore, it is important to note that the”, and “game-changer”. Each is 10 points.
Four places the list stays narrow
A longer match wins
Phrases are tried in order, and overlapping spans are skipped. “Delve into” is one mark. The shorter pattern for “delve” does not fire inside it.
“Unlock” ignores the product
A lone “unlock” is marked even on a padlock, and the suggestion is “help”. In the sample the longer phrase wins, so the suggestion there is “to support growth”. Read the card before you replace it.
One cultural pairing
The serious mark fires when one sentence contains “Muslim” and either “Christmas” or “Black Friday”. Jokes, slang, and other faiths are invisible to this build.
Length is a count
A line labeled Button, Label, or CTA is marked when the text after the colon is longer than 28 characters. The sample starts with “Primary button (de-DE):”. “Primary” and a parenthetical tag are optional. That label is 53 characters. On a product-UI scan set to German or French, a line with no period, longer than 32 characters and shorter than 90, is also treated as a control when it is not already labeled that way.
Fetching a URL, the Figma plugin, Google sign-in, a brand glossary, and an API are named in the spec. This site cannot open them. The comparison on the pricing page marks each one as not connected.
What to do with the band
Under 60 is red, 60 through 80 is yellow, and over 80 is green. A score of 80 stays yellow. A submitted draft is sent back only when the new score is under 80.
Under 60
High risk. On this sample the serious mark is 24 of the 82 points. Clearing every moderate mark and leaving the clash and the button still scores 68, which is yellow. Edit the clash first.
60 to 80
The draft can be read, and it still sounds like a machine. The suggestion is already on the card. Use a person when the line is about to reach a customer.
Over 80
This list is done with it. A single moderate mark scores 90, which is green, and the report still offers a reviewer while that mark is open. A wrong product term, a joke, or slang can pass.
The reviewer is a button on the report
While any mark is open, the report asks whether to send this draft to a native reviewer in the next 15 minutes. That is the button copy. If the list matches nothing, the same spot offers a full read, at any score.
After a reviewer submits, the same list runs again. Under 80, the draft goes back. The spec’s target for 500 words or fewer is one hour, and it also describes assigning a job if nobody claims it in five minutes. You can walk the queue on this machine. There is no reviewer abroad to meet either clock.
The order of those three steps is on How it works. Listings and buttons are separate pages: ecommerce and software.
Count it on a paragraph you plan to publish
Free is 3,000 words a month and, in the spec, only marks problems. The sample above is the same text as Try sample text on the homepage.