Meta is pushing advertisers to use its AI tools — and results are proving chaotic: strangely twisted limbs, gibberish writing, or entirely changed products.
Meta’s response to brands: That’s on you, not us.
The tech giant has inserted a slew of AI features into its ad products in recent months. Working as designed, they can help make tweaks to ads that improve their likelihood of being clicked. But advertisers say the tools are clunky and generate misrepresentations and absurdities.
Business Insider spoke with eight advertisers and agency execs who said dealing with Meta AI problems had become routine.
Jessica Gleim, an ads consultant who works with female-founded brands, told Business Insider she regularly sees odd outcomes in Meta’s AI creative recommendations for ads she’s working on.
For one of her clients, a pajama brand, Meta recommended new assets that altered the actual product. The brand was promoting a pajama dress, and Meta suggested a new image with a shirt and pants. For another client, a networking group for women in Montana, Meta had a new vision for those ads: adding men.
BI
“It’s not usable to help my clients grow their business,” Gleim said.
While some of Meta’s AI ad features are turned off by default, advertisers say they have been prone to bugs that accidentally turn them on. Karissa Tuccio, executive director of social and influencer at Mediassociates, said a bug that toggled AI settings on had regularly affected most of the 15 clients for whom she handles Meta advertising. She said she had flagged the bug to her Meta rep as recently as Thursday.
BI
Outdoor retailer REI drew consumer backlash last month for running an Instagram ad depicting a nonsensical bike with two handlebars. REI said Meta had “auto-enrolled” it in an AI feature that spat out an “inaccurate” and “inappropriate” image.
A Meta spokesperson said that the company’s terms of service state that “AI can make mistakes and that it is the advertiser’s responsibility to review the AI outputs.”
Advertisers’ chief complaint about Meta’s AI ad tools is simple: They feel they have to double-check all the AI features for each campaign to make sure nothing is inadvertently switched on or has gone haywire. With some advertisers and agencies running hundreds or thousands of ads at any given time, the extra steps required to wrangle the AI tools create more work.
BI
“We somehow accepted that as a new standard operating procedure,” said Rok Hladnik, CEO of the marketing agency Flat Circle, which manages around $200 million in annual Meta ad spending for numerous direct-to-consumer brands.
Brands and advertisers say that while AI failures can be an amusing talking point on the internet, they can pose real problems for a brand.
“When the AI starts generating weird creative or making unapproved changes, it can quietly damage brand perception — especially for anyone who cares about consistency,” said Robert Webster, CEO of TAU Marketing, which manages around $500 million in annual ad spending across various platforms.
A Valentine’s Day surprise
Around Valentine’s Day, photographer and marketer Abigail Hogue was uploading an ad campaign to Meta. She works with a small business, Quite Literally Books, and had shot the creative assets for the holiday campaign with chocolates, macarons, candles, and books. Hogue was proud of the work she’d done.
“About 12 hours later, when everything was approved and started to run, I started getting some messages from friends and people that I knew and screenshots of some of these ads that were running, cheekily accusing me of AI slop,” Hogue said.
The text on the products in the images was “garbled,” and the “actual products look like knockoff iterations of themselves,” she said.
Quite Literally Books; BI
When Hogue saw the AI ad, she went into a panic and edited the campaign in Meta’s Ads Manager, turned off all AI creative enhancements, and then republished the ads.
She then spent hours in a back-and-forth with Meta customer service. Representatives told her it was a “sporadic” and “one-off occurrence,” and also said it was a “glitch,” according to screenshots of their exchange viewed by Business Insider. She requested a refund, and Meta acknowledged her request. Quite Literally Books said it hadn’t received a refund as of Friday afternoon.
Other advertisers have told Business Insider about their strange Meta AI ads, ranging from an unrealistic granny in loungewear to a model whose leg appeared to be completely bent the wrong way.
Luke Jonas, chief growth officer of the marketing agency Nest Commerce, emphasized the importance of keeping a human in the loop when testing AI-generated ads.
“A machine optimizing for 6 million advertisers will occasionally give you two handlebars,” Jonas said, referencing the REI ad.
While two advertisers said Meta appeared to have fixed a bug that was toggling AI settings on for their clients, Mediassociates’ Tuccio said a similar issue persisted for her as of last week.
Tuccio said a Meta rep told her last week that Meta had developed a quality-control dashboard for big advertisers to ensure their ads don’t go live with unwanted AI enhancements.
“She mentioned, ‘If you guys have a big launch coming up, you can send me all the ad IDs, and we have an internal dash that will check to make sure all of the enhancements have been fully turned off,'” Tuccio said. “So that leads me to believe it has not been resolved.”
‘Meta’s still the best platform’
Starting last month, Meta began automatically applying an “AI info label” to ads when they use its AI tools — or third-party tools like Midjourney or Dall-E — to create or significantly edit their ads. To see it, users must click the three dots above an ad, select “about this ad,” and then tap on “AI info.” Google added labels last week to indicate whether ads were created or edited using AI.
Meta is also improving its AI image generation models. Last week, it began rolling out Muse Image, a model developed by its Superintelligence Labs, which can help advertisers develop their creative assets. (Following backlash, Meta on Friday removed a feature in Muse Image that let users generate AI images from other people’s public Instagram posts, saying it “missed the mark.”)
Still, advertisers say Meta’s basic design encourages relinquishing control to the system, which can lead to disastrous results.
“The defaults are aggressive, the toggles are easy to miss, and the system is clearly designed to reduce friction so more money flows through the platform with less manual intervention,” TAU’s Webster said.
Meta says “millions of advertisers are finding value and improved performance using our Advantage+ creative tools to support ad creation.” The Meta spokesperson added that the company’s AI image generation tool, which creates variations based on a seed image provided by the advertiser, is turned off by default.
Meta isn’t alone in automatically modifying advertisers’ creative. Google’s Performance Max and AI Max products also use AI to scrape ad copy from brand websites and automatically crop or shorten videos for placements such as YouTube Shorts. Some of these AI automation features are enabled by default, though Google has largely avoided the kind of high-profile issues Meta has seen.
Danny Weisman, cofounder of Obsessed Media, said the main complaint he’s heard about Google from brands is that ads made with its AI tools could turn out looking “ugly.”
“It’s not like someone’s hand is missing,” he said.
Meta’s ad business, which pulled in around $196 billion in revenue last year, remains essential to most brands’ customer acquisition strategies. Its reach of 3.5 billion daily active users and highly sophisticated ad targeting platform make it difficult to quit, even if problems arise.
“That means it can make unpopular decisions that boost its own profits with near impunity, because most advertisers cannot realistically walk away,” TAU’s Webster said.
Then there’s the simple truth: Meta ads generally get results.
“Meta’s still the best platform,” Gleim said. “It has the most robust options. It has the most data.”
