Skip to content
Stetwright

FAQ

Questions before your first review.

How to get started, what the checker covers, and how your copy is handled.

Is this legal advice?

No. It is a wording checker. It compares your text with a published set of cited rules and shows where the wording matches. It does not create an attorney-client relationship, and it does not replace review by qualified counsel.

If the checker finds nothing, is my copy fine?

No. It means no rule matched. It cannot tell whether a claim is true, it does not read images or video, and it only knows the laws and policies listed on the coverage page. You are still responsible for what you publish.

What is not covered

Does my text leave my browser?

The rule check, the on-device model read, the proofs, the approval and the exports all run in your browser. The one exception is the AI second read, which sends the text to Anthropic's API using your own key, and only after you tick a consent box and press Run. Clerk handles accounts, sessions, workspaces and authentication security data. The copy you check is not sent to Clerk or our account server.

Security and privacy

Can I paste patient information?

No. The product is not built to receive it. There is no business associate agreement with Anthropic, so patient information must not go through the AI second read, and you should not paste it anywhere in the checker.

What does it check?

565 rules in 13 categories, covering FTC, FDA, HIPAA, OIG and IRB guidance, state law for Alabama, Arizona, California, Colorado, Connecticut, Florida, Georgia, Hawaii, Illinois, Indiana, Kentucky, Louisiana, Massachusetts, Maryland, Michigan, Minnesota, Missouri, North Carolina, New Jersey, New Mexico, Nevada, New York, Ohio, Oklahoma, Oregon, Pennsylvania, South Carolina, Tennessee, Texas, Utah, Virginia, Washington and Wisconsin, and a few ad platform policies. Every rule carries its reason, its source and a safer wording.

Rules and coverage

How accurate is it?

On 150 real FDA and FTC warning-letter claims, the rules alone block 56 percent, measured on rule-set 0.6.0. The first version of our research classifier flags 81 percent, and rules together with it catch 95 percent. That classifier is a research prototype and not the model in the checker. Those are the real-text numbers. Scores on AI-written text are higher and flatter the system. Precision on real customer copy has not been measured.

Research results

Who writes the rules, and has a lawyer reviewed them?

The rules are written from public sources and each citation is checked against stored source text when the rule-set builds. Counsel review is pending. Until a qualified reviewer signs off, treat every rule as unreviewed.

Methodology and sources

What does the approval record prove?

It binds a named approver and a timestamp to a hash of the exact text, the findings and any included optional AI second read. Any later edit voids it. Exports are signed in your browser, so later edits to the file can be detected. It does not prove who you are, and it does not prove the content is lawful. The timestamp comes from your own device.

Can I review a CSV file?

Yes. Choose CSV batch, upload a file with a header row, and select the copy column. The checker processes up to 5,000 rows and 5 MB locally. Filter the results and download a summary that preserves your original columns. Open individual items to resolve findings and approve them; batch screening is not bulk approval.

CSV batch workflow

Do I need to run an external AI review?

No. You can check, approve and export without an external AI read or a waiver. If you choose the optional AI second read, it uses your Anthropic connection or key and requires explicit consent before sending copy.

What is the model read?

A small trained model that runs in your browser and reads each sentence. It flags wording a regulator would likely object to, including wording the rules miss. It is advice only and never blocks approval. You can turn it off.

Does it check Meta, Google or TikTok ad policies?

Only a few, and they are unverified. 6 platform rules exist. They were written from memory of the policy pages, which could not be stored. Do not rely on them for platform approval.

Limitations

Can I add my own rules?

Yes. A team can add its own policy, such as an IRB's advertising requirements or a sponsor's approved-language list. It runs in your browser and is saved in the exported report. Nobody reviews it for you.

Do you keep a copy of my text or my reports?

No. Copy and reports stay on your device. Local history is off by default and is stored in this browser only when you enable it. We cannot recover it. Clerk handles account, session and workspace data separately.

Privacy

What does it cost?

We are not charging yet. The pricing page shows the plans and prices we plan to launch with. Billing waits until counsel reviews the rule-set and the disclaimer. The checker is free to try after signing in and choosing a workspace.

Pricing

A finding is wrong, or it missed something. What now?

Tell us. The checker has links on each finding that open an email with the rule id and the matched phrase. Those reports are how the rule-set gets better.

Contact

Next step

Still have a question?

Ask it. If it's one other people would ask, it goes on this page.