Prompt Pilotby Write.law
All prompts

Draft

Substantiate an ad claim before it runs

Maps every message a consumer takes from your copy (implied ones first) against the evidence you actually hold, and returns an approve, modify, hold, or decline call with replacement wording.

About 25 minadvancedRegulatory, In-house

Your prompt4,962 characters

Still to fill in: The claim, exactly as it will appear, Evidence you actually hold, Markets and state-law overlays

RoleYou are an advertising regulatory lawyer who has answered FTC Section 5 inquiries, defended NAD challenges, and been on both sides of a Lanham Act § 43(a) case. You write down the implied messages before the express ones, because the express claim is almost never the one that gets challenged. You will not clear copy you could not defend with the documents already in the file.What I needSubstantiate the claim below for Website hero or landing page under FTC general: consumer goods and services, running in Markets and state-law overlays, and tell me whether it can go live.InputsThe claim as it will appear: The claim, exactly as it will appear Evidence held: Evidence you actually holdHow to work this1. List every message a reasonable consumer takes from this copy, numbered, implied messages first. Where the wording is ambiguous, list both readings. A regulator takes the worse one. 2. Tag each message before looking at the evidence: establishment or not, comparative or monadic, quantified or qualitative, health-related or not. The tags select the standard. 3. For each message, name the standard, name the item in Evidence you actually hold relied on, and say whether that evidence sits at the level of the claim. If single-arm baseline data is carrying a comparative claim, say so in those words. 4. For every comparative message report three checks separately: head-to-head design, current comparator version, and whether the measured endpoint matches what the copy promises. 5. Decide for each failure whether the fix is disclosure or copy. Where disclosure works, give placement, proximity, trigger, and exact wording. Where it does not, write the replacement claim. 6. Close with one disposition: Approve, Approve with modifications, Hold pending substantiation, or Decline, plus, short of approve, the replacement copy and the specific evidence that would unlock the original claim.Close with these four sections, every time, without being askedAssumptions I made. What I assumed about study design, sample, endpoint, comparator version, and how the copy renders on Website hero or landing page. Mark each [verify] or [safe]. Where this is weakest. The two or three conclusions most likely to be wrong: a message I may have read too generously or too harshly, or an evidence call that turns on a protocol I have not seen. Name the message number. What only you can decide. Present each as options with tradeoffs. At minimum: running the comparative as drafted and accepting that the named competitor has standing to sue and a self-regulatory challenge is faster and cheaper than litigation, versus narrowing to the within-product claim now and commissioning head-to-head testing that costs money and several months but buys a claim you can defend. Also yours: whether the conversion value of the strongest phrasing is worth the enforcement magnet it creates, and whether to pull live placements or let the flight run out. What would make this materially better. Rank by impact: the study protocol and statistical analysis plan, the comparator's current formulation date, the distribution of individual results behind the average, and the copy as it actually renders on the page.Output formatA privileged memo: header, a three-sentence executive summary carrying the disposition, the numbered message list with tags, an evidence map table of Message | Type | Standard | Evidence relied on | Adequate | Gap, the endorsement and disclosure section, the regime overlay, the competitor and self-regulatory risk read, state-law exposure, the disposition, paste-ready replacement copy, and the four closing sections.Never do this- If the memo would apply to any ad for any product, it is too generic. Quote this copy and this evidence in every finding. - No hedging filler. Cut "arguably," "it should be noted," and "this may raise concerns" used in place of a call. Do not tell me to consult advertising counsel. I am advertising counsel. - Never invent a study result, a sample size, an NAD decision, a consent order, or a guide section. Anything not in my inputs is marked [UNVERIFIED - confirm before the claim runs]. - Where you cannot tell whether the evidence supports a message because the protocol is not in my inputs, say you do not know and ask for it. Do not clear a claim on the strength of fluent reasoning. - Do not pad. If three messages fail, the memo is about three messages. Length is not value.Before you answer- Did I list implied messages before express ones, and both readings where copy is ambiguous? - Is every message tagged and matched to a named standard? - For each comparative, did I report head-to-head design, comparator currency, and endpoint match separately? - Did I say whether the fix is copy or disclosure, and give the exact words? - Would this memo fit a different ad? It should not.

Adds driver's-seat tunes: options instead of answers, questions before work, every citation flagged. Your values come with it.

2

Pressure-test it

Makes the AI switch hats and attack its own answer.

You will be attacked from two directions here, and only one of them is the government. First read the memo as the FTC staff attorney drafting a warning letter: name the two messages you build the letter around and the sentence of copy you quote. Then read it as counsel for the named competitor preparing a self-regulatory challenge: name the single message you attack, the evidence you demand in the challenge, and why you file there instead of in court. Then rewrite the disposition to survive both, and list the substantiation steps in the order they should be commissioned.
3

Go deeper

Pushes the work further once the basics are right.

If the challenge lands, the response goes out on the challenger's schedule, not yours. Draft the substantiation response the company would file: a page and a half, leading with the strongest evidence rather than the procedural objections, conceding the one message that is not defensible, describing the corrective action already taken, and attaching the substantiation file index. Then list what in the current file would need to be produced and what is privileged.

Before you run it

What to gather first

  • The exact copy as it will appear, including subhead, footnote, and any voiceover
  • Every study, survey, and internal dataset behind the claim, with design and sample size
  • For comparatives, which competitor and which version of their product was tested
  • Whether endorsers, influencers, or testimonials appear anywhere in the campaign
  • Where the ad will run and how long it stays up

Watch for

  • The model reads only the copy you paste. Voiceover, on-screen supers, packaging panels, and the landing page the ad clicks through to all carry messages. Omit them and the analysis is incomplete.
  • Verify every study result, sample size, and citation. The model will produce plausible-sounding NAD decisions and consent orders that do not exist.
  • Comparative claims built on a competitor's older formulation create exposure in two forums at once, and the competitor moves faster than the regulator.
  • Disclosures cannot contradict or cure a false main claim. If the memo proposes a footnote to fix a headline, that is the finding to push back on.
  • Health, earnings, environmental, and AI-capability claims all sit under heightened scrutiny and carry regime-specific disclaimer requirements. Confirm the current standard for your regime before clearing.

What comes back

A privileged memo: a three-sentence executive summary carrying the disposition, a numbered list of express and implied messages with tags, an evidence map table showing message, type, standard, evidence relied on, adequacy, and gap, an endorsement and disclosure section with placement and wording, the regime overlay, the competitor and self-regulatory risk read, state-law exposure, the disposition, paste-ready replacement copy, and the four collaboration sections.

See an example of what you’ll get
MEMORANDUM - PRIVILEGED · To: Marketing and Compliance · Re: "Clinically proven to reduce stress 47% faster than the leading brand" · April 27, 2026 Executive summary. Hold pending substantiation. The headline carries an establishment claim and a comparative superiority claim, and the file supports neither: the 47% figure comes from a single-arm study measuring participants against their own baseline, and the only comparator data tests Brand A's 2018 formulation on 42 people. A within-product version of the claim can run this week with the modifications in Section 11. The comparative cannot run until a head-to-head study exists. Messages consumers take. 1. *(implied)* The 2025 study compared this product to the leading brand head-to-head. Comparative, establishment 2. *(implied)* A typical user gets a 47% improvement. Quantified, monadic 3. *(implied)* "The leading brand" is a specific identifiable product, and in this category that reads as Brand A. Comparative 4. *(express)* Clinically proven. Establishment, health-related 5. *(express)* Reduces stress. Monadic, structure-function Evidence map (selected). | # | Standard | Evidence relied on | Adequate | Gap | |---|---|---|---|---| | 1 | Head-to-head testing of current versions | None. The 2025 study has no comparator arm | No | The subhead attributes a comparison to a study that made none; this is the finding, not a technicality | | 2 | Typical-results substantiation | Mean only; no distribution supplied | No | Need median and range; a 47% mean with an unreported spread invites the "up to" problem | | 4 | Competent and reliable scientific evidence | n=240, single arm, PSS-10 self-report | Partial | Uncontrolled self-report will not carry "proven"; "studied" is defensible | Comparative checks (message 1 and 3). Head-to-head: no. Comparator currency: no; the n=42 crossover tested a 2018 formulation against a 2026 ad. Endpoint match: partial; PSS-10 measures perceived stress, not speed of onset, and the copy promises "faster." Disclosure versus copy. The existing "*Results may vary" footnote does not cure any of this. A disclosure cannot convert a comparison the study never made into a true statement. Messages 1 and 3 require copy change. Message 2 is curable by disclosure adjacent to the claim, not in the footer. 11. Replacement copy. Headline: "Clinically studied to reduce stress." Body: "In an 8-week study, 240 participants reported an average 47% reduction in perceived stress compared with their own starting scores. Individual results varied." Adjacent disclosure, same type size family: "Based on an 8-week single-arm study of 240 users measuring self-reported stress. Not a comparison to other products." Plus the DSHEA structure-function disclaimer. --- Assumptions I made. "The leading brand" reads as Brand A to consumers in this category [verify - a message survey would settle it and would also settle message 1]. The 2025 study's primary endpoint was PSS-10 change from baseline [verify against the protocol]. The footnote renders below the fold on the hero [verify with the live page]. Where this is weakest. Message 3. I have treated "the leading brand" as identifying Brand A, which drives the standing analysis. If the category has no clear leader, the claim is still unsupported but the competitor risk drops sharply and the disposition could soften. What only you can decide. Run a narrowed within-product claim now and commission head-to-head testing for a Q4 relaunch: that costs the study budget and roughly five months but produces a claim that survives a challenge. Or hold the whole hero and keep the current creative dark until testing completes, which protects the record and costs you the launch window. Separately: whether the conversion lift from "proven" is worth being the company that used the word on uncontrolled self-report data, and whether to pull the placements already live or let the flight expire. What would make this materially better. (1) The study protocol and statistical analysis plan. (2) The distribution behind the 47% mean. (3) The date of Brand A's current formulation. (4) A consumer message survey on "the leading brand": it decides messages 1 and 3 together.
Why this prompt is built the way it is
## Framework 1. **Implied messages first.** The express claim is rarely the one that draws the challenge. Where copy is ambiguous, both readings are live and the regulator takes the worse one. 2. **Tag before you evaluate.** Establishment or not, comparative or monadic, quantified or qualitative, health-related or not. The tags select the standard. 3. **Match evidence to the level of the claim.** Single-arm baseline data does not support a comparative superiority claim. Say that plainly rather than calling it insufficient. 4. **Comparatives need three checks.** Head-to-head design, current comparator version, and an endpoint that matches what the copy promises. 5. **Disclosure versus copy.** A footnote asterisk does not cure a false headline. When disclosure works, specify placement, proximity, trigger, and wording. 6. **One disposition.** Approve, approve with modifications, hold pending substantiation, or decline, with the replacement copy attached. 7. **Name the parallel forums.** FTC is not the only risk. The competitor with standing and the self-regulatory body move faster.