A creative compliance checklist for regulated device ad copy

Most device ad reviews happen at the end. A campaign is built, the creative is approved by marketing, and a regulatory reviewer sees it two days before launch. The reviewer then rewrites the headline, the campaign slips a week, and nobody changes how the next one is written.

The fix is a checklist the copywriter runs first, built around the one document that bounds what a cleared device may be said to do: the Indications for Use in its 510(k) record.

Why does device ad copy need its own pre-flight?

Because the ad can define what the device is for. Federal regulation says a device’s intended use may be shown by advertising matter and by oral or written statements, not only by its label. Every one of the 22 FDA decision letters in this teardown also reminds the manufacturer that device labeling must be truthful and not misleading.

That is the mechanism most marketing teams miss. Under 21 CFR 801.4, the objective intent behind a device “may, for example, be shown by labeling claims, advertising matter, or oral or written statements.” A headline that promises something the device was not cleared for becomes evidence of an intended use that nobody reviewed.

FDA’s own device labeling page quotes an appellate decision: most, if not all advertising, is labeling. A LinkedIn carousel, a landing page and a trade-show banner are all in scope. The checklist below treats them the same way.

The checklist does not replace legal advice. It exists so that copy reaches regulatory review already traceable, and so the reviewer spends time on judgement calls rather than on rewriting “FDA approved” for the third time this quarter.

What does the checklist look like?

Eight sections, numbered 0 to 7 and run in order, each ending in a record someone can audit. It starts with the device’s regulatory status and its Indications for Use pasted verbatim, then maps every claim to a clause or a study, then checks format risks specific to paid social and landing pages. Nothing ships with an open item.

DEVICE AD PRE-FLIGHT: one per asset (ad, landing page, carousel, video, email)
Review aid only. Escalate every "no" or "unsure" to regulatory/legal review.

0  SOURCE OF TRUTH  (fill before writing a word)
   [ ] Product code and regulation number, from the FDA classification record
   [ ] Pathway: 510(k) / De Novo / PMA / 510(k) exempt (from the record, not memory)
   [ ] K / DEN / P number for THIS product, if any (not the product code's)
   [ ] Indications for Use pasted verbatim from the decision letter
   [ ] Rx or OTC, as labeled
   [ ] Contraindications and warnings from current labeling, pasted

1  REGULATORY STATUS LANGUAGE
   [ ] "FDA-cleared" only with a 510(k) number for this product on file
   [ ] "FDA-approved" only with a PMA number on file; never for a 510(k)
   [ ] Exempt product: no cleared/approved language at all
   [ ] No "FDA registered" or "FDA listed" used as a quality or approval signal
   [ ] Status line never sits beside an outcome claim it could appear to endorse

2  CLAIM SCOPE: one row per claim
   claim text | IFU clause it rests on | in scope? (Y/N/escalate)
   [ ] Every benefit claim maps to a named IFU clause
   [ ] Population: any audience named in copy or image (athletes, seniors,
       children, post-op) is named in the IFU, or escalated
   [ ] Setting: home / clinic / hospital / field claims match the IFU
   [ ] User: "use it yourself" vs "clinician-applied" matches Rx/OTC labeling
   [ ] Body site and condition words match the IFU, not a broader synonym
   [ ] No claim the IFU explicitly excludes ("not intended for …")
   [ ] Wellness, performance or recovery framing checked against IFU wording

3  SUBSTANTIATION: one row per claim not answered by the IFU alone
   claim | evidence (study, bench test, label) | population/setting matches? 
   [ ] Health or safety benefit claims cite competent and reliable scientific evidence
   [ ] Numbers ("in 10 minutes", "40% faster") trace to a specific result
   [ ] The study population and setting match the ad's audience
   [ ] Comparative claims ("better than", "#1") have head-to-head evidence
   [ ] "Clinically proven" or "studies show" names the study

4  ENDORSEMENTS AND PEOPLE
   [ ] Testimonials reflect typical results or carry a clear, adjacent qualifier
   [ ] Clinician endorsers: material connections disclosed
   [ ] No patient story implies an off-IFU use

5  FORMAT RISKS: paid social and landing pages
   [ ] Net impression checked with image, product name and headline together
   [ ] The claim survives truncation: qualifiers sit before the truncation point
   [ ] Image and video frames make no claim the text would not be allowed to make
   [ ] Landing page matches the ad's claim scope; no stronger claim after the click
   [ ] Targeting does not imply an out-of-scope population the copy avoids naming
   [ ] Disclosures are adjacent, legible on mobile, and do not contradict the claim
   [ ] Conversion pages about specific conditions: tracking-pixel placement reviewed

6  OUT OF SCOPE FOR MARKETING: route, do not write
   [ ] Payer, coverage or cost-to-patient statements → market access owner
   [ ] Any use outside the IFU → regulatory/medical affairs
   [ ] Comparative safety statements about a named competitor → legal

7  RECORD
   reviewer | date | asset version | IFU source (K number + letter date) | open items: 0

What does a 510(k) record actually let copy say?

What the Indications for Use says, and little else. In a teardown of 22 recent decision letters across 12 rehab, respiratory and DME product codes, every letter bounded the clearance to “the indications for use stated in the enclosure.” The statements run a median of 63 words, and most name a purpose without naming a patient.

The teardown pulled the two most recent substantially-equivalent decisions for each product code in a fixed panel that had any, fetched the public decision letter, and read the Indications for Use field. Of the 21 statements recoverable as text, 12 name no patient population. Only 8 name a use setting, and 4 name a user or supervising clinician.

The gap is widest where sports medicine and rehab marketing spends most. In the nerve stimulation, muscle stimulation and therapeutic-heat codes, 8 of 9 statements name neither a population nor a setting. They list purposes: symptom relief, muscle re-education, local blood circulation. An ad that says “for athletes” or “for faster return to play” is adding a population and an outcome the statement does not contain.

Silence in an Indications for Use is a question for regulatory review, and the checklist forces it to be asked at the claim-scope row rather than after the creative is built.

Some statements go further and exclude uses outright. One muscle-conditioning stimulator in the sample states that its programs are not suitable for rehabilitation and physiotherapy. Copy that pitches that device to physical therapists as a rehab tool contradicts its own clearance, however natural the audience looks from a targeting point of view. In the file, 7 of the 21 statements carry an explicit limit of this kind.

Why is “FDA-cleared” the riskiest phrase in the ad?

Because it is a regulatory claim with a precise meaning, and the most common errors are all upgrades. A 510(k) clearance is not an approval. Establishment registration is not an approval. An exempt product has no clearance to cite. Each upgrade implies an FDA review that did not happen, and federal regulation calls that implication misbranding.

The regulation is unusually direct. Under 21 CFR 807.97, a substantial equivalence determination “does not in any way denote official approval,” and any representation that creates an impression of official approval “is misleading and constitutes misbranding.” Section 807.39 says the same of registration: a registration number does not denote approval of the establishment or its products.

The FTC’s Health Products Compliance Guidance tells advertisers not to mischaracterize the extent to which a product or claim has been reviewed by the FDA. Its Example 49 describes a Class II device cleared “for the intended purpose of stimulating and strengthening healthy muscle,” then advertised as FDA approved next to weight-loss and waist claims. It was the status line and the outcome claim together that produced the deceptive impression. That is why the checklist keeps them apart.

Exempt products need their own line. In openFDA’s classification data, 76.4% of physical medicine product codes, 168 of 220, are 510(k) exempt. For those products there is often no clearance to cite. Exemption belongs to the product code, though, and individual products can differ: in this panel, 2 exempt codes still had 510(k) decisions in the window, for devices that needed one. The checklist asks for this product’s own number.

How does it apply to LinkedIn creative and landing pages?

Through net impression. The FTC assesses an ad as a whole, including the text, product name and any charts, graphs and other images. On paid social that whole is assembled from a truncated headline, a thumbnail, a product name and a landing page, and each can make a claim the others were careful to avoid.

Truncation is the paid-social specific risk. A qualifier placed after the fold may never be read, so the checklist puts it in the first line or removes the claim. The FTC’s guidance is plain that a disclosure cannot directly contradict a claim, so a qualifier is never a repair for a headline that overreaches.

Images claim things too. A stimulator shown on a sprinter’s hamstring makes a population and a use claim without a word of copy. So does a respiratory product photographed at a bedside when its statement names the home.

Targeting can do the same. An audience built on titles or skills implies who the product is for. If the copy carefully avoids naming a population the statement does not include, and the targeting selects exactly that population, the campaign as a whole is still saying it. Row five of the checklist exists for that reason.

The landing page is the last place claim scope drifts. Ads get reviewed; pages get edited. A page that ran the checklist at launch and has been changed three times since needs a fresh pass. So does any page tuned for AI search, a concern for the AI search visibility hub, because answer engines lift sentences out of context and the qualifier rarely travels with them.

Where does the checklist stop?

At the edge of marketing’s authority. It does not interpret regulations, decide whether a claim is within scope, or replace review by qualified counsel. It does not cover payer or coverage statements at all. What it does is make every claim traceable before review, so the questions reach the right person early.

There are four limits. The checklist is federal and device-focused, and says nothing about state law or other countries. It assumes the Indications for Use is available; for some records the statement is not recoverable from the public PDF as text, and in this teardown 1 of 22 was not. It treats the Indications for Use as the boundary for claim scope, but evidence supporting a claim is a separate question, answered in section three. And it cannot govern what a sales rep says in a room.

Evidence raises its own questions. Much of the published research on a device category is produced outside the manufacturer, as covered in who generates device clinical evidence, and a study’s population must match the ad’s before it substantiates anything.

How was the teardown run?

Against openFDA and the public decision-letter PDFs, on 2026-09-30. A fixed panel of 20 product codes common in rehab, sports medicine, home respiratory and DME was classified by pathway; for each code with decisions, the two most recent substantially-equivalent decisions from 2021 to 2025 were fetched and their Indications for Use extracted and scored.

The panel was chosen by hand to represent the categories a device, DME or sports medicine marketer sells into. It is not a random sample, and nothing here estimates a rate for all clearances. Two decisions per code is a structured read. Codes with no decision in the window contribute no rows, which removes most of the exempt codes and one 510(k)-pathway code.

Extraction used the Form 3881 field, the newer electronic-submission form, or the 510(k) summary’s own section, in that order. Scoring is keyword-based: the patterns are recorded in the teardown file, so any individual call can be checked or disputed. A keyword flag says a statement names a population; it does not judge whether the naming is adequate. The prescription or over-the-counter checkbox is not recoverable from the PDF text and is not scored.

What the teardown cannot tell you is whether any specific ad complies. It shows how narrowly the records are written, which is the case for checking every claim against them. The same public register is the basis for tracking where new device competitors are entering.

What is the smallest first step?

Fill in section zero for one product. Pull its record, paste the Indications for Use verbatim into the brief template, and require every future brief to carry it. Copy written with the statement on the page drifts far less than copy checked against it afterwards.

Then run section two on the live campaign with the most spend, one claim per row. The rows marked “escalate” are the review agenda, and they usually take an hour. The paid acquisition hub collects the rest of the paid-media pieces.

A claim that arrives with its clause attached takes a reviewer minutes. A claim that arrives without one takes the campaign.