Meta and AI: An Advertising Nightmare for Brands

Meta's Advertising Ambitions and AI: A Risky Bet for Brands

Meta, the social media giant, is actively encouraging advertisers to adopt its artificial intelligence (AI) tools to optimize their advertising campaigns. However, this strategy faces often unpredictable and problematic results. Users report anomalies such as images with distorted human limbs, incomprehensible text, and products altered to the point of being unrecognizable. In the face of these malfunctions, Meta absolves itself of any responsibility, passing the buck back to the advertisers.

For several months, Meta has integrated various AI features into its advertising offerings, promising to enhance ad effectiveness by increasing click-through rates. Nevertheless, industry professionals express frustration with these tools, deeming them unreliable and generators of absurd content. Business Insider gathered testimonies from eight advertisers and agency heads who reported recurring difficulties with Meta's AI. Jessica Gleim, an advertising consultant specializing in women's brands, noted strange and inappropriate creative recommendations for the campaigns she manages.

For one of her clients, a pajama brand, Meta suggested visual elements that altered the original product. While the brand wanted to promote a nightgown, Meta's AI proposed an image featuring a shirt and pants. In another case, for a women's network in Montana, the AI recommended adding men to the visuals, which contradicted the campaign's identity. "This doesn't help me grow my clients' businesses," Gleim lamented.

AI Bugs: A Headache for Advertisers

Although some of Meta's AI features are disabled by default, recurring bugs sometimes activate them without warning. Karissa Tuccio, executive director of social media at Mediassociates, reported that this type of bug regularly affects the majority of the 15 clients whose campaigns she manages on Meta. She recently alerted her contact at Meta about this issue.

A striking example of these malfunctions occurred with outdoor retailer REI, which ran an Instagram ad showing a bike with two handlebars, causing consumer confusion. REI explained that Meta had "automatically enrolled" it in an AI feature, resulting in an inaccurate image. A Meta spokesperson reminded that the terms of service state that AI can make mistakes and that it is the advertiser's responsibility to verify the results.

Advertisers criticize the need to check every AI feature for each campaign to ensure no errors slip through. For agencies managing hundreds, if not thousands, of ads simultaneously, these additional checks represent a considerable workload. Rok Hladnik, CEO of the agency Flat Circle, which oversees about $200 million in annual advertising spending on Meta, stated, "We have integrated this as a new operational standard."

AI errors, while amusing to some internet users, can have serious consequences for a brand's image. Robert Webster, CEO of TAU Marketing, which manages around $500 million in annual advertising spending, emphasizes that strange creations or unapproved modifications can harm brand perception, especially for those who value consistency.

The Impact of AI Errors on Campaigns

Around Valentine's Day, Abigail Hogue, a photographer and marketer, launched an advertising campaign on Meta for Quite Literally Books, a small business. She had carefully designed the campaign visuals, including chocolates, macarons, candles, and books. However, shortly after launch, she began receiving messages from her contacts, accompanied by screenshots of the ads, mocking the "AI errors."

The product texts had become unreadable, and the images appeared to depict counterfeit versions of the actual products. Faced with this situation, Hogue quickly disabled all AI creative enhancements in Meta's Ads Manager and republished the ads. She then spent hours communicating with Meta's customer service, which described the incident as "sporadic" and "unique," while acknowledging a "bug." Although she requested a refund, Quite Literally Books had not yet received compensation by the end of the week.

Other advertisers shared their experiences with strange AI ads on Meta, ranging from a grandmother in casual attire to a model with an improbably bent leg. Luke Jonas, growth director at Nest Commerce, stressed the importance of maintaining human oversight when testing AI-generated ads. "A machine optimizing for 6 million advertisers can sometimes produce absurd results," he noted, referring to the REI ad.

While some advertisers have seen improvements after Meta fixed a bug activating AI settings, Karissa Tuccio reported that similar issues persisted. A Meta representative informed her that an internal dashboard checks to ensure that all enhancements have been completely disabled for large advertisers.

Meta and AI: A Complex Relationship

Since last month, Meta has introduced an "AI information label" for ads using its AI tools or those from third parties like Midjourney or Dall-E. This label is accessible by clicking on the three dots above an ad, then selecting "about this ad" and "AI information." Google has also added similar labels to indicate whether ads were created or modified by AI.

Meta continues to improve its AI image generation models. Last week, it launched Muse Image, a model developed by its Superintelligence Labs, aimed at helping advertisers enhance their creations. However, after negative feedback, Meta removed a feature from Muse Image that allowed for generating AI images from public Instagram posts, stating it "did not meet expectations."

Advertisers criticize Meta's basic design, which encourages relinquishing control to the system, often at the expense of results. "The default settings are aggressive, and the easy options are easy to miss, which promotes a flow of money without manual intervention," explains Webster from TAU.

Meta claims that "millions of advertisers find value and improved performance using our Advantage+ creative tools to support ad creation." A Meta spokesperson clarified that the AI image generation tool, which creates variations from an initial image provided by the advertiser, is disabled by default.

Meta is not alone in automating ad creations. Google's Performance Max and AI Max products also use AI to extract text from brand websites and automatically adjust videos for formats like YouTube Shorts. Some of these features are enabled by default, although Google has avoided the high-profile issues faced by Meta.

Danny Weisman, co-founder of Obsessed Media, notes that criticisms of Google mainly concern the aesthetics of AI-generated ads, deemed "ugly" but without glaring errors like missing limbs.

Meta's advertising business, which generated about $196 billion in revenue last year, remains crucial for many brands' customer acquisition strategies. With a reach of 3.5 billion daily active users and a sophisticated targeting platform, it is difficult for advertisers to turn away, even in the face of problems.

"This means that Meta can make unpopular decisions that increase its profits with little consequence, as most advertisers cannot reasonably do without the platform," emphasizes Webster from TAU.

Despite the challenges, Meta's advertising results remain compelling. "Meta is still the best platform," asserts Gleim. "It offers the most robust options and has the most comprehensive data."