Brand value, not compliance, is the biggest AI risk.
Most of the talk about AI in marketing is about cost. The blind spot is what it does to the top line. When we tested real content with a made with AI label, engagement and conversion fell, and premium brands were hit hardest.
The blind spot
Agencies and brands have been quick to count the savings AI offers. Faster production, cheaper versions, fewer shoots. What gets far less attention is the effect on revenue and on the value of the brand itself.
That's where we think the real risk sits. Regulation matters, and it's changing quickly, but compliance is a process problem with known answers. A drop in conversion across your best selling range is a much bigger problem, and it doesn't announce itself.
I set out this thinking in an interview with LBB in August 2026. This is the longer version.
What our impact assessments found
We ran impact assessments for brands across categories, adding a made with AI label to real content and measuring the response. Engagement and conversion fell whatever the product's price point, and the size of the fall was strongly linked to price.
| Measure | Change with a made with AI label |
|---|---|
| Social media engagement rate | -15 - 20% |
| Ecommerce conversion rate | -13 - 16% |
Premium brands, and premium products within a category, saw the biggest falls. Shoppers expect premium treatment, and low cost AI can read as a signal of lower quality.
The wider research points the same way. The IAB's 2026 study found 39% of Gen Z feel negative about AI generated ads, up from 21% two years earlier, against 20% of Millennials. It also found that advertisers think consumers like AI ads far more than consumers say they do.
The rules are hardening
Three forces are reshaping what AI costs brands. Legislation, platform rules and consumer behaviour.
Legislation
From 2nd Aug 2026, Article 50 of the EU AI Act requires deepfakes to be disclosed clearly, at the latest at first exposure. Many global brands are now applying EU standards in every market, which creates contradictions. AI usage targets on one side, labelling rules that reach everyday tools like generative fill on the other.
In the US, New York's synthetic performer law took effect on 9th Jun 2026. Ads that include a digitally created person must say so conspicuously, with fines of $1,000 for a first violation and $5,000 after that. The fines are small. The labels are not.
Platform rules
Every major platform is building AI compliance ahead of the law. Tags and self certification are already appearing in Amazon's Vendor Central, and undisclosed AI changes risk a suppressed listing. Meta and TikTok label realistic synthetic media, even if enforcement is light for now.
Music streaming hit AI saturation first. Supply of AI tracks kept climbing while demand stayed flat, and platforms responded by labelling AI content and pushing it down. There's no reason to think social platforms won't follow, and hiding labels is unlikely to last.
Likeness and scams
The issues go beyond copyright. Our clients now get messages claiming someone's likeness was used, and scam claims that blame a brand's AI. Both need a clear process, handled with empathy and documented properly.
Six steps to use AI without hurting the brand
- Run an impact assessment before you scale. Weigh the cost saving against the possible loss of revenue. Ideally model substitute purchasing, price elasticity and reputation too.
- Build governance into the workflow you already have. Quality control is the natural home. A standard checklist lets creatives work freely inside clear guardrails.
- Share what you learn. Case studies and test results, passed between teams, spread good practice faster than policy documents.
- Start in safe spaces. Martech, research, ideation and creative development build skills at low risk.
- Agree policy with clients for high visibility work. Brand campaigns, social and ecommerce need a shared view of where AI is used and how it's labelled.
- Prepare for likeness claims and scams. Know who responds, how they respond and what gets recorded.
- The biggest AI risk is to revenue and brand value, not compliance
- Labelled AI content cut engagement and conversion in our tests
- EU and US rules on disclosure are now in force
- Assess the impact first, then build checks into quality control
An advantage for agile teams
We've been here before. When GDPR arrived, the fear was that it would end interactive tech and lead generation. Both grew, because new practices like zero party data took their place. AI regulation will reward the same agility, and smaller, faster studios are well placed to lead.
The winners won't be the brands using AI the most. They'll be the ones who know where to use it and where not to, and who replace guesswork with evidence. That's how we build and use Freak Lab, and how our content team decides where AI belongs on each brief.
This article is a general overview and isn't legal advice.
Sources
Questions clients ask us.
Does labelling content as made with AI hurt performance?
In Freak's impact assessments it did. When real content carried a made with AI label, social engagement rates fell by 15 - 20% and ecommerce conversion rates by 13 - 16%. The biggest falls were for premium brands and premium products within a category.
Does the EU AI Act require labels on AI generated ads?
Article 50 of the EU AI Act requires deployers to disclose deepfakes, meaning AI generated or manipulated images, audio or video that resemble real people, objects, places or events and would falsely appear authentic. It applies from 2nd Aug 2026. Evidently creative or satirical work has a lighter duty, and it's worth taking legal advice on each use.
What is New York's synthetic performer law?
Signed in December 2025 and in force from 9th Jun 2026, it requires advertisers to conspicuously disclose when an ad includes a synthetic performer, a digitally created person who isn't a recognisable real individual. Penalties are $1,000 for a first violation and $5,000 for each one after.
Where is it safe to use AI in marketing?
Start in lower risk places such as martech, research, ideation, creative development and quality checks. Use it more carefully in high visibility work like brand campaigns, social and ecommerce content, and test the impact before you scale.
More from the Resource Centre.
How we score Amazon images before shoppers ever see them.
The five measures behind our neural marketing framework, what each one tests and how we use them to fix Amazon images before they go live.
UGC that works on Amazon and social, from brief to final cut.
How we brief, cast and cut UGC for TikTok, Reels and Amazon, with rights agreed first and versions for organic and paid.
HFSS advertising rules in 2026, and what food brands can still do.
What's banned on TV and online since 5th Jan 2026, what's still open and how food and drink brands are adapting their media and creative.
Got a brief like this?
Tell us where you want to grow, and we'll show you how we'd get there.