Meta's AI ad tools are modifying creative without approval-and marketers are hitting the off switch
Clothing brand Snag Tights discovered last summer that Meta's AI had altered its ads without permission. The company noticed the changes only after the ads ran, and by February, it posted publicly: "If you see an ad from us that looks 'off' or has that strange AI sheen, please know: It isn't us."
CEO Brie Read said the modifications created a fundamental business problem. When Meta's AI generates a product that doesn't exist-like a rainbow handbag designed from scratch-and that image appears in an ad, customers who buy based on the image are being misled.
"If the picture isn't of anything real, you're effectively scamming the customer, right?" Read said.
Snag Tights requested Meta disable AI testing on its account, and the company complied. But the incident prompted the brand to explore moving some ad budget away from Meta's platforms entirely.
A wider problem
Snag Tights isn't alone. As Meta pushes AI deeper into ad buying and creative development, agencies and advertisers report unwanted AI-generated elements distorting their ads. Some discovered their budgets were automatically allocated to AI testing without explicit approval.
Curtis Howland, marketing VP at Misfit Marketing, said his team has observed unexpected AI modifications including stretched images, background alterations, and full creative reimaginations-static images converted to video without consent.
One example: Meta's AI stretched a patio image so severely that grass appeared to be growing "on top of and into the patio." The team disabled the feature immediately.
Agency Formada Social found that even after disabling AI features, they re-enable themselves when ads are duplicated. Co-founder Meghan Kelly described the experience as a "wild goose chase" to keep AI turned off, with the platform continuing to generate unwanted creative and spend budget on it.
Turning off the tools proves difficult
Meta says advertisers can opt out of AI creative testing through their Ad Account Settings. Spokesperson Simone Levien said the company runs tests on a small share of impressions to improve performance, but "advertisers are always in control of their experience."
That control isn't matching reality for many users. Finding the correct toggles within Meta accounts is difficult, according to Howland. Even when settings are changed, some features reactivate without warning.
The core complaint is straightforward: the AI-generated ads are poor quality. Howland said he would never launch an ad without reviewing it personally, regardless of the tool used. "The idea of just letting Facebook in the background make a bunch of ads for you is absurd," he said.
Brands exploring alternatives
Snag Tights is testing ad buys on Reddit, TikTok, Substack, and podcasts. Read said the brand wants to be in spaces where customers can trust the ads they see.
"If that kind of thing happens on Facebook a lot, where you get more and more ads that can't be trusted by the customers, that's also not a space we want to be in," she said.
The shift reflects a practical concern: if AI modifications become routine and visible, the platform itself becomes less reliable for advertising. Brands stake their reputation on what appears under their name, regardless of who or what created it.
For creatives managing campaigns on Meta, understanding how to navigate-or disable-these AI features is becoming essential. Consider exploring AI Design Courses and AI for Marketing Courses to better understand automation tools reshaping ad platforms.
Your membership also unlocks:
