Skip to content
Stetwright

Claims checker for healthcare marketing

Find the claims a regulator would likely quote, before you publish.

Paste your ad, landing page, email or trial recruitment copy. 565 cited rules check the wording, then you resolve each claim and record the approval against the exact text.

  • The rule check runs in your browser
  • No accounts, no analytics
  • Not legal advice
Sample ad text, checked by the real engine at build time8 blocking

Our stem cell shot cures arthritis. 100% guaranteed. Book this week.

  • Blocks approvalCG-DIS-018“Our stem cell shot cures arthritis.”8 rules match this phrase

    Stem cell products and protocols marketed to treat or reverse disease, or to stop a condition, are unapproved drug and biologic claims.

    Source. FDCA 201(g)(1); Public Health Service Act 351; FDA consumer alert on regenerative medicine products; FTC Act section 5

    Safer wording. Remove disease and cure claims. Describe a stem cell service only as an investigational study, and name the approved uses if any.

  • Blocks approvalCG-ABS-002“100%”

    A 100 percent claim is an absolute claim that almost never has support.

    Source. FTC Act section 5; FTC Health Products Compliance Guidance

    Safer wording. Replace with the measured figure and its source, or remove the number.

565

rules

296 block approval, 269 warn

33

states with rules

AL AZ CA CO CT FL GA HI IL IN KY LA MA MD MI MN MO NC NJ NM NV NY OH OK OR PA SC TN TX UT VA WA WI

324 of 334

sources stored as text

Every citation to them is checked at build

0.8.0

rule-set version

Counsel review pending

Built for the person who has to say yes.

Every finding cites its source

Each hit shows the rule, the reason, the regulation or guidance behind it, and a safer way to say it. You see why, not just a red mark.

One claim, one card

Overlapping rule matches fold into a single claim per phrase, so a paragraph doesn't bury you in duplicates.

Fix it, prove it, or override it

Edit the wording, attach a proof such as a study or a consented quote, or override a finding with a written reason. Nothing is waved through silently.

An approval tied to the exact text

Approval records who signed off on which words. Change one character and the approval is void. Exports are signed in your browser.

Three readers, each doing a different job.

Regular-expression rules can't read meaning. So the checker pairs them with two readers that can, and keeps the rules as the part you can audit.

  1. 01

    Rules

    Deterministic

    Cited, versioned, re-checkable. The same text gives the same result, and a report can be re-verified later.

  2. 02

    On-device model

    Advice only

    A small trained model reads each sentence in your browser and flags wording the rules miss. It never blocks approval.

  3. 03

    AI second read

    Optional, with explicit consent

    Claude judges implication and context, and whether each rule hit is real. It sends your text to Anthropic, only after you consent.

See what each layer catches, and what it misses

What it will not do.

  • It does not certify anything.

    No findings means no rule matched. It is not legal advice, and counsel review of the rule-set is still pending.

  • It does not check truth.

    It reads wording. Whether a claim is backed by evidence is your job, and a lawyer's.

  • It does not cover everything.

    Text only. A limited set of federal sources, 33 states, and a few unverified platform rules.

Read the full list of what is not covered

Next step

Paste one real ad and see what it flags.

It takes a minute and needs no account. If it fits your workflow, join the waitlist. We are not charging yet.