Imagine spending weeks developing a campaign, getting the creative approved, and launching it on Meta. Then you discover the platform has made a few creative decisions of its own. For some advertisers, this is apparently becoming a reality.
According to new reporting from Adweek, Meta's AI advertising tools have been modifying ads without brands knowingly requesting the changes. We're talking AI-generated illustrations, translated copy, additional sitelinks, and entirely new creative variations.
Some advertisers have gone as far as pulling live campaigns to remove the unwanted changes. And there's another wrinkle. Meta is updating its commercial terms to make advertisers responsible for content its AI tools generate or modify, including features the platform enables automatically.
So the platform can change your ad, and you're still responsible for what goes out. Somewhere between automated optimization and creative approval, the lines of responsibility are getting awfully blurry.
Creative Changes Without Approval
Online bookseller Bookshop.org recently discovered that Meta had added an AI-generated illustration to an advertisement originally created by a human designer.
According to the company, an automated account update had enrolled it in a generative AI feature without its knowledge. Bookshop.org wasn't particularly impressed with the result, describing the supposedly enhanced ad as “ruined.” The company has since disabled Meta's AI advertising features.
And Bookshop.org isn't alone. Digital agency Exverus told Adweek that some of Meta's AI-generated changes appeared even when advertisers hadn't knowingly enabled the features. The agency has had to pull live ads after unwanted modifications made it through its quality checks.
The changes involve Meta's Advantage+ creative features, which can automatically adjust advertising assets, including images, copy, and formats. Advertisers can disable individual enhancements in Ads Manager, although agencies say some settings can be easy to overlook.
Now, I can understand why Meta is investing so heavily in automated creative optimization. Testing variations, adjusting formats, and identifying opportunities to improve performance are all useful applications of AI.
But there's a difference between helping a marketer improve an advertisement and making changes the marketer never approved.
A campaign might use a particular illustration style for a reason. A line of copy might have gone through multiple rounds of review. Even a seemingly minor change could alter the meaning of an ad or introduce something inconsistent with the brand. And the people responsible for those decisions might not even know the changes happened.
Who Is Responsible for Meta's AI-Generated Ads?
The situation gets even more complicated when you look at Meta's updated commercial terms. Under the revised provisions, advertisers are solely responsible for what the platform's AI features create or modify.
Meta also acknowledges that its AI tools can produce inaccurate or unsuitable results. The revised terms were originally scheduled to take effect October 30, but Meta has postponed implementation until November 20 following objections from publishers and digital advertisers over separate provisions involving website content.
Meta has described the changes as a consolidation of its existing generative AI advertising terms. But the company didn't directly answer Adweek's question about whether advertisers would be responsible for changes made by features they never activated themselves.
Marketers have always been responsible for what their brands publish. They're expected to review creative, substantiate claims, and make sure campaigns meet legal and brand requirements.
Under the revised terms, advertisers could find themselves responsible for creative changes they didn't even know were happening. And with the holiday advertising season approaching, brands have limited appetite for disrupting campaigns or walking away from one of their biggest advertising channels.
It's an uncomfortable position to be in. The technology is supposed to reduce the work involved in managing campaigns, yet advertisers may have to spend more time checking what the platform has changed.
The Thinking Behind Better Creative
Meta's AI tools can generate variations, adjust formats, and optimize advertisements against campaign objectives. Those capabilities can save marketers considerable time. But an ad's performance metrics don't necessarily capture everything that makes the creative effective.
Say a brand develops a campaign for an audience with very particular preferences. The creative team spends time getting the imagery, language, and cultural references right, down to details that might seem insignificant to someone outside that audience. Then an AI tool swaps out an image or rewrites a headline because it predicts better performance, without knowing why those choices were made in the first place.
Those changes might even generate more clicks. But whether they produce a better outcome for the brand is another question.
This is central to how we think about AI at RAD Intel. We're interested in using intelligence to improve decisions before brands commit their budgets, rather than relying exclusively on automation to optimize campaigns once they're running. Understanding an audience, evaluating creative choices, and anticipating how people might respond should inform what gets produced and distributed.
It's also why visibility into automated changes matters. If a platform alters creative without explaining what changed or why, marketers lose some of the context they need to evaluate the decision.
What Meta's AI Ad Changes Mean for Marketers
I don't think the takeaway from Meta's situation is that marketers should stop using AI to optimize advertising. There's too much potential value in faster testing, better audience targeting, and more informed creative decisions to dismiss the technology altogether.
But more automation also creates a need for clearer boundaries. Marketers need to know which decisions they've delegated to a platform, which changes require approval, and how they're evaluating the results. They also need to check what actually runs, especially when creative can change after a campaign launches. And those expectations shouldn't disappear because a platform makes automation the default.
Meta's situation is a useful reminder of how quickly the relationship between marketers and their technology providers is changing. Advertising platforms are increasingly participating in decisions that used to sit entirely with brand and agency teams.
The technology may be capable of making those decisions faster. Whether it has enough context to make the right ones is another matter.
For now, advertisers might want to take another look at what's actually running in their Meta campaigns. The creative they approved and the creative their audience sees may not be quite the same thing.
