Using Research Quotes in Marketing: FTC Endorsement Guides, Material Connections, and the Reviews Rule

Answer first: a research quote becomes an endorsement under FTC law the moment it appears in any marketing message, and three obligations attach at that instant. The quote must not be reworded or trimmed in a way that distorts the speaker (16 CFR 255.1(b)). The experience it describes must be typical of what customers generally achieve, or you must disclose what they can generally expect (255.2(b)). And any material connection between you and the speaker — including the incentive you paid them for the research — must be disclosed clearly and conspicuously if the audience would not expect it (255.5). None of this is triggered while the quote sits in your repository. All of it triggers on publication.

Most research teams never learn this, because the person who runs the interview is rarely the person who drops the quote into a landing page. The quote travels: interview to repository, repository to a Slack message, Slack to a slide, slide to the website. By the time it is public, nobody remembers whether the participant was paid, whether they still use the product, or whether the sentence was three sentences originally.

This guide covers the rules that apply on the marketing side and the operational controls that keep research usable without creating advertising-law exposure. It is a companion to Customer Quotes: How to Extract, Tag, and Use the Voice of Your Customer, which covers the craft of finding good quotes; this article covers what happens legally once you use one.

Where research ends and advertising begins

The FTC Endorsement Guides define an endorsement as any advertising, marketing, or promotional message that consumers are likely to believe reflects the opinions, beliefs, findings, or experiences of a party other than the sponsoring advertiser (16 CFR 255.0(b)). The definition is deliberately broad. Verbal statements, demonstrations, depictions of a name or likeness, and the name or seal of an organization can all be endorsements.

The Guides also make an instructive point in their very first example: a film critic writing a review is not making an endorsement, but the moment the producer excerpts that review in an advertisement, the excerpt becomes one. Nothing about the words changed. The use changed.

That is exactly the research case. A participant telling your AI interviewer that onboarding took ten minutes instead of two days is research data. The same sentence in a pricing-page carousel is an endorsement, and you are now the advertiser responsible for it.

Rule 1: quotation marks are a promise of exact words

The Guides are unusually specific here. An advertisement need not present an endorser message in the exact words of the endorser unless the advertisement represents that it is presenting the exact words, such as through the use of quotation marks (255.1(b)). But in either case the endorsement may not be presented out of context or reworded so as to distort the endorser opinion or experience.

Two practical consequences follow.

Trimming is allowed; reshaping is not. Removing filler, false starts, and a tangent is normal editing. Removing the qualifier is not. "It is fast, once you get past the setup, which took us a week" cannot become "It is fast." The clause you removed was the participant assessment, not noise.

If you use quotation marks, be verbatim. The safest workflow is to publish the exact transcript text with elisions marked, and to keep the untouched transcript segment on file. AI-moderated platforms make this easy — Koji stores the full transcript for every interview with timestamps, so the source of any published sentence is one export away rather than a memory exercise.

A useful internal standard: any quote used externally must be traceable to a transcript line, and the person approving publication must have read the two turns before and after it. Most distortion happens because someone read the highlight, not the exchange.

Rule 2: the endorser has to still be a customer

Section 255.1(c) requires that when the advertisement represents that the endorser uses the product, the endorser must have been a bona fide user at the time the endorsement was given — and the advertiser may continue running the ad only so long as it has good reason to believe the endorser remains a bona fide user.

Testimonial pages rot silently. A quote captured from a delighted customer in 2024 who churned in 2025 is a live compliance problem in 2026, and nobody notices because testimonials are the least-maintained content on any website.

The fix is operational, not legal:

• Stamp every published quote with the capture date and the account it came from.
• Run a quarterly join between your testimonial inventory and your active-customer list.
• Pull or refresh anything from a churned account, and anything older than your refresh interval.
• For expert or named endorsers, re-confirm views at reasonable intervals — the Guides expect it, and what counts as reasonable turns on things like product changes and new performance information.

A quarterly quote audit takes an hour and removes an entire category of risk.

Rule 3: the typicality trap

This is the rule research teams break most often, and it is the one with real enforcement history.

Section 255.2(b) says that an endorsement relating one or more consumers experience on a central or key attribute will likely be read as representing that the experience is representative of what consumers will generally achieve. If you do not have substantiation that it is representative, the ad should clearly and conspicuously disclose the generally expected performance — and you need substantiation for that number too.

Critically, the Guides state that a disclaimer such as "these testimonials do not prove our product works; you should not expect to have similar results" is likely to be deceptive anyway unless the advertiser has adequate substantiation that new users typically experience similar results. The old "results not typical" escape hatch does not work. The disclosure has to change the net impression of the ad, and it has to be true.

The Guides also state plainly that consumer endorsements themselves are not competent and reliable scientific evidence (255.2(a)). One participant saying they saved fifteen hours a week is not evidence that customers save fifteen hours a week — no matter how many participants said something similar in your qualitative study.

So there are exactly two lawful routes when a quote describes a result:

1. Substantiate the general case. Run a study designed to support the claim, then use the quote as colour on top of a number you can defend. That design problem is covered in Advertising Claim Substantiation.
2. Disclose the general case. Publish the quote alongside the typical outcome, clearly and conspicuously, in the same visual field as the quote — not in a footnote, and not behind a hover.

A third option is often the best one: use quotes for experience rather than results. "The AI kept asking follow-ups until it understood what I meant" is a description of the experience. "It cut our research cycle by 80%" is a performance claim with a substantiation burden attached.

Rule 4: your research incentive is a material connection

Section 255.5(a) requires disclosure of any connection between endorser and seller that might materially affect the weight or credibility of the endorsement and that is not reasonably expected by the audience. The Guides list monetary payment, free or discounted products (including unrelated products), early access to a product, the possibility of being paid, the possibility of winning a prize, and the possibility of appearing in media promotions.

Every one of those is a standard research incentive. A gift card, beta access, an entry into a prize draw, a discount on renewal — all of them are material connections if the quote ends up in an ad.

But the Guides contain a timing nuance that matters enormously for research, and almost nobody knows it. In the material-connection examples, a scenario is described in which patrons asked for their opinions did not know and had no reason to believe, until after they answered, that they were being recorded for use in an advertisement. The Guides state that no disclosure is required in that situation — even if the patrons were told afterwards that they would be paid for allowing their opinions to be used in advertising.

The logic is that a payment which could not have influenced the opinion does not need to be disclosed to let the audience weigh the opinion. Apply that to research:

This is a strong argument for keeping recruitment honest: a study run as research, with the marketing question asked at the end, is both better research and a cleaner endorsement position than a study run as a testimonial harvest. It is also why the request for permission belongs at the end of the interview rather than in the recruitment email.

Two caveats. The nuance covers the material connection disclosure only — the verbatim rule and the typicality rule still apply in full. And it is fact-specific; where the incentive is unusually large relative to the audience expectation, disclose it anyway. Disclosure costs a line of text.

Rule 5: employees, insiders, and the Reviews Rule

Since 21 October 2024, a second instrument sits alongside the Guides: the FTC Rule on the Use of Consumer Reviews and Testimonials, 16 CFR Part 465. The Guides are administrative interpretations. Part 465 is a rule, and rule violations carry civil penalties — currently up to $53,088 per violation.

Four provisions intersect with research operations:

Never condition an incentive on sentiment (465.4). It is a violation to provide compensation or other incentives in exchange for, or conditioned expressly or by implication on, reviews expressing a particular sentiment. "Leave us a review and get a $25 card" is fine only if it is genuinely sentiment-neutral. "Leave us a five-star review and get a $25 card" is a per se violation. In research terms: never route a satisfaction study into a review request in a way that filters on the score. If you ask promoters to review and detractors to book a call, you have built exactly the machine the rule prohibits.

Generalized solicitations are protected (465.2(d)(1)). Asking all purchasers to post reviews about their experience sits outside the prohibitions in 465.2(b) and (c). Segmenting that ask by predicted sentiment does not.

Insider testimonials need disclosure (465.5). A testimonial from an officer, manager, employee, or agent — or their immediate relatives — must clearly and conspicuously disclose the relationship unless it is otherwise clear to the audience. This catches internal dogfooding quotes and employee research more often than people expect. See Anonymous Employee Research with AI Interviews for why employee feedback and marketing content should not share a pipeline.

Do not suppress by sentiment (465.7 and 255.2(d)). If you host reviews and represent that what is displayed reflects most or all reviews submitted, suppressing on rating or negative sentiment is a violation. Neutral criteria applied equally — profanity, personal information, suspected fake, off-topic — are permitted. The Guides make the parallel point at 255.2(d): advertisers should not procure, suppress, boost, organise, publish, upvote, downvote, report, or edit reviews in ways that distort what consumers think.

Consent for research is not consent for marketing

A research consent form covers participation, recording, analysis, and internal use. It does not, on its own, authorise commercial use of a person likeness or words. Under GDPR the marketing use is a different purpose from the research purpose, and purpose limitation means you cannot quietly repurpose. In the US, most states recognise a right of publicity that attaches to commercial use of a name or likeness.

Run a two-permission model:

1. Research consent at intake, covering the session itself. See Research Consent Form Templates and Intake Forms and Consent.
2. Marketing release asked separately, after the interview, specifying: which quote, attributed how (full name and company, first name and role, or anonymised), on which surfaces, for how long, and how to revoke.

Attribution level matters more to participants than teams assume. Offering three tiers — named, role-only, anonymous — routinely converts people who would refuse a blanket release.

Koji makes this a structured step rather than an email thread. Add the release as structured questions at the end of the guide: a yes_no question for permission, a single_choice question for attribution level, and a multiple_choice question for approved surfaces. Because these are typed fields rather than free text, permission state is queryable — you can filter your repository to quotes cleared for external use instead of asking a lawyer to read transcripts. The AI interviewer asks them conversationally at the end, so nothing about the research portion is contaminated by the marketing ask.

The pre-publication checklist

Run this before any research quote goes public. It takes about four minutes per quote.

Store the answers with the quote. If a challenge ever arrives — from a regulator, a competitor, or the participant — the file is the defence, and it has to have existed before publication, not after.

How Koji supports this end to end

Traditional survey tools give you a spreadsheet cell containing a sentence, with no context and no chain of custody. That is the worst possible input for an endorsement decision. Koji is built the other way round:

• Full transcripts with timestamps mean every published sentence is traceable, so the verbatim rule is verifiable rather than assumed.
• AI follow-up probing captures the qualifier in the participant own words. A survey text box gives you "It is fast." An AI interview gives you "It is fast once the data is imported, which took our ops team a day" — which is both better research and a safer quote.
• Six structured question types (open_ended, scale, single_choice, multiple_choice, ranking, yes_no) let permission, attribution, and surface approvals live as typed data next to the quote.
• Real-time reports show which quotes cluster around a theme, so you pick a representative quote rather than the most flattering outlier — which is exactly what the typicality rule is asking you to do.
• Voice and text modalities mean you can capture the release in the participant own voice, on the record, in the same session.

Start with 10 free credits and run a study whose last three questions are the release. You will never chase a quote permission by email again.

Frequently asked questions

Does a quote in a sales deck count as advertising?

Yes. The Guides cover any advertising, marketing, or promotional message, and a deck shown to a prospect is promotional. The practical difference is exposure, not category — a one-to-one deck is less likely to draw attention than a homepage, but the same rules apply to both.

Can I fix a participant grammar in a quote?

Light cleanup of filler and false starts is normal and acceptable, provided the substance and tone are unchanged and you are not misrepresenting the words as exact. If you use quotation marks, you are representing the words as exact, so keep it verbatim and mark elisions. Never fix grammar in a way that makes a hesitant speaker sound emphatic.

Do I need to disclose that participants were paid for the research?

It depends on whether the payment could have influenced the opinion. Where a person took part in genuine research, was paid a standard incentive, and was asked for quote permission only afterwards, the Guides material-connection examples indicate no disclosure is required. Where people were recruited knowing marketing content might result, or were given anything of value in exchange for a testimonial, disclose it.

What is the difference between the Endorsement Guides and the Reviews Rule?

The Guides (16 CFR Part 255) are administrative interpretations of Section 5 of the FTC Act and mainly govern how advertisers use endorsements. The Reviews Rule (16 CFR Part 465, effective October 2024) is a trade regulation rule aimed at fake, purchased, insider, and suppressed reviews. The Rule carries civil penalties — currently up to $53,088 per violation — which the Guides do not.

Can I ask satisfied customers to leave a public review?

Yes, if the ask is sentiment-neutral and generalized. Asking all customers to review their experience is protected under 465.2(d)(1). Offering compensation conditioned on positive sentiment violates 465.4, and routing only high scorers to the review page is the same thing implemented in software.

How long can we keep using a testimonial?

Only as long as you have good reason to believe the endorser remains a bona fide user (255.1(c)) and the claim remains substantiated. There is no fixed expiry, which is why a quarterly reconciliation between your testimonial inventory and your active-customer list is the standard control. Set a review date on every published quote.

Related resources

• Structured Questions Guide — the six question types that turn permission and attribution into queryable data
• Customer Quotes: How to Extract, Tag, and Use the Voice of Your Customer — finding quotes worth publishing in the first place
• Advertising Claim Substantiation — designing research that can support a marketing claim
• AI Customer Testimonial Interviews — capturing case studies and win stories at scale
• Research Consent Form Templates — the research half of the two-permission model
• Interview Recording Consent Laws — one-party, all-party, and biometric rules
• Research Ethics and Informed Consent — the ethical floor beneath the legal one