Not Every AI-Assisted Ad Needs a Label. Every Claim Needs a Review.

Dear Suzannah

We use AI in our marketing. Do we need to put an AI label on everything now?

No. The better question is whether AI changed something a reasonable customer would care about when deciding what to believe. On August 18, the IAB announced Version 2 of its AI Transparency and Disclosure Framework, emphasizing practical, contextual disclosure instead of labeling every use of AI. Meanwhile, the FTC’s standing rule is wonderfully less glamorous: advertising claims still need to be truthful, not misleading, and supported by evidence.

Here’s the deal

AI disclosure is becoming a marketing operations question, not just a creative question. The IAB’s updated framework says disclosure should focus on consumer relevance. The FTC says advertising claims must be truthful and evidence-based, and its endorsement guidance requires material relationships to be disclosed clearly when they could affect how people evaluate an endorsement.

There is another clue in where marketing technology is going. On August 18, Pega announced new customer-engagement capabilities with built-in governance and transparency controls. The direction is pretty clear: AI can move faster, but the business still owns what reaches the customer.

What this actually helps you do

Build a simple pre-publish decision rule so your team can use AI without turning every ad into a tiny legal essay or, at the other extreme, publishing synthetic claims nobody stopped to inspect.

Niche use case

Consider a small marketing agency producing paid social ads for local service businesses. AI helps draft headlines, create image variations, summarize reviews, and generate campaign concepts. The agency needs a fast way to distinguish ordinary production assistance from content that could materially change what a prospect believes.

Exactly three benefits

• Faster approvals. The team knows which AI-assisted changes are routine and which ones deserve a closer look.
• Stronger trust. Material synthetic elements and paid relationships are less likely to surprise the buyer.
• Cleaner claims. Every performance statement, testimonial, price promise, and result gets checked for evidence before publication.

Step-by-step instructions

1. List the AI touchpoints. For one campaign, note where AI helped with copy, images, video, voice, reviews, audience selection, or personalization.
2. Separate assistance from representation. Grammar cleanup is different from creating a person who appears to be a real customer. Resizing a photo is different from generating a before-and-after result.
3. Circle every factual claim. Mark numbers, rankings, savings, outcomes, guarantees, customer results, credentials, and comparisons.
4. Attach evidence. Keep the source for each material claim where the reviewer can actually find it. If nobody can find the proof, the claim is not ready.
5. Check endorsements and relationships. If a creator, employee, customer, affiliate, or other endorser has a material connection to the brand, make sure the disclosure is clear and hard to miss where required.
6. Review synthetic people and experiences. Ask whether a reasonable viewer could mistake generated content for a real person, real event, real testimonial, or real product result.
7. Choose the smallest clear disclosure. When disclosure is warranted, make it understandable and close to the relevant content. Do not hide the important part behind vague language.
8. Approve the final ad, not the draft. Check the version that will actually run, including captions, creative, landing page, and offer details.

Tips and tricks

• Keep a one-page claim sheet for recurring offers so the team does not research the same proof every Tuesday.
• Use plain words in disclosures. If a customer needs a glossary to understand the disclosure, congratulations, you made a second problem.
• Save the source date with the claim. Evidence can age.
• Review the landing page with the ad. A careful ad can still lead to a misleading page.

Common mistakes

• Labeling all AI use while failing to inspect the actual claim.
• Using a generated testimonial that sounds like a real customer’s experience.
• Assuming a platform’s built-in disclosure automatically covers every material relationship or claim.
• Letting a polished synthetic image imply a product feature or result the business cannot deliver.

Human review checklist

• Does every factual claim have evidence?
• Could any generated person be mistaken for a real customer, employee, or expert?
• Does any image imply a result or feature that is not real?
• Are endorsements honest and are material relationships disclosed when needed?
• Is any required disclosure easy to notice and understand?
• Does the landing page match the promise in the ad?
• Would the overall message still feel fair if the customer knew exactly how it was made?

How to measure success

Track approval turnaround time, percentage of ads returned for unsupported claims, disclosure-related corrections after launch, complaint rate, qualified lead rate, and conversion quality. The goal is not merely fewer mistakes. A good review rule should help the team publish faster because routine AI assistance stops getting confused with high-consequence representations.

FAQ

Does every AI-generated image need an AI label?

Not as a universal rule. Context matters. Review whether the synthetic element could materially affect what the customer believes and check the rules that apply to the platform, industry, location, and specific advertising claim.

Does disclosure make a false claim acceptable?

No. A label does not rescue a deceptive or unsupported claim. Truthfulness comes first.

Can we use AI to summarize customer reviews?

You can use AI as an internal aid, but customer-facing summaries need careful review. Do not invent details, change the overall meaning of genuine feedback, or turn generated language into a fake customer endorsement.

What if our platform already has an AI disclosure tool?

Use required platform tools, but do not assume the button replaces your responsibility to evaluate the whole message. The FTC has specifically warned in endorsement guidance that built-in disclosure tools may not always be sufficient by themselves.

Glossary

AI-assisted content: Marketing material where AI helped create, revise, analyze, or produce part of the work.

Material information: Information that could affect how a reasonable customer evaluates a marketing message or buying decision.

Endorsement: An advertising message consumers are likely to understand as reflecting someone else’s opinion, belief, finding, or experience.

Clear and conspicuous: A disclosure presented so ordinary consumers can notice, read or hear, and understand it.

Sources and further reading

• Interactive Advertising Bureau: AI Transparency and Disclosure Framework Version 2 announcement, August 18, 2026
• Federal Trade Commission: Advertising and Marketing Basics
• Federal Trade Commission: Advertisement Endorsements
• Pega: Responsible AI for Customer Engagement, August 18, 2026

Related Workflow Edit: Connected AI Apps Need a Permission Review, Not Blind Trust and Stop Measuring AI by Seats. Measure Successful Work.

Practical closing note

AI disclosure should not become a sticker your team slaps on everything so everyone can stop thinking. The useful habit is simpler: inspect what the customer is being asked to believe, prove the claims, disclose material context when it matters, and keep a human accountable for the final message.

Next action: Take one live ad and circle every sentence, image, testimonial, or generated element that could change what a buyer believes. Verify those first.