On July 13, 2026, a pair of investigative reports laid bare a growing advertiser revolt: Meta’s AI-powered ad tools are silently altering live campaigns, turning pajama dresses into separates, swapping male models for elderly women, and generating copy that belongs to a different product entirely. The tools, part of Meta’s Advantage+ suite, are supposed to optimize performance—but for dozens of brands and agencies interviewed by Business Insider and Crypto Briefing, they introduced impossible anatomy, garbled text, and reputational hazards instead.
When the Tool You Trusted Rearranges Your Product
The reports surfaced first in Crypto Briefing and Business Insider, each citing multiple advertisers and agency executives who described a pattern of unauthorized, often surreal modifications. In one widely circulated example, a women’s networking-group ad suddenly featured a man. A pajama dress was transformed into a shirt-and-pants set without the retailer’s knowledge. An e-bike company called Lectric found its AI-tweaked ad copy talking about car trunks. True Classic, a clothing brand, saw its young male model replaced by an elderly woman—a change the company flagged only after the ad had been running.
REI faced public backlash when an Instagram ad displayed a bicycle with two handlebars, a distortion so obvious it went viral. The outdoor retailer told Business Insider that Meta had auto-enrolled the brand in an AI feature that produced an “inaccurate and inappropriate image.” Another advertiser said a holiday campaign was altered after approval, leaving product labels garbled and making items look like cheap imitations.
These aren’t edge cases. Eight ad professionals interviewed by Business Insider described recurring issues. The core complaint is not that generative AI can make mistakes—it’s that the modifications often happen after approval, without clear consent, and in some cases because settings became enabled through bugs or opaque default changes.
A Tangle of Settings, Defaults, and Bugs
At the center of the chaos are Meta’s Advantage+ creative tools, which include an enhancement layer that can adjust visuals, overlay music, shift colors, or rewrite copy to boost predicted performance. Meta says that its AI image-generation feature—the one that creates variations from an advertiser-supplied image—is off by default. But other creative enhancements are not, and advertisers told Business Insider that settings had been activated unexpectedly, sometimes without any logged change.
Meta’s official response, given to Business Insider, points to the company’s terms: AI can make mistakes, and advertisers are responsible for reviewing outputs. A spokesperson also emphasized that millions of advertisers are seeing value and better performance. Last month, Meta began applying an “AI info” label to qualifying ads—those made or substantially edited with Meta’s tools or certain third-party AI systems. But the label appears only after the fact, and it doesn’t prevent the kind of unauthorized overhauls that alarmed REI and others.
The timing matters. According to Crypto Briefing, Meta rolled out major AI infrastructure upgrades in early to mid-2026, and the company has publicly targeted full AI automation of ad creation by the end of this year. That context suggests a platform in a hurry—one where speed of deployment may be overwhelming quality controls.
What This Means If You Run Ads on Meta
The practical fallout depends on who you are, but three groups should pay immediate attention:
- Brand marketers and small-business owners: An altered ad can distort your messaging, confuse customers, and—as REI’s example shows—trigger public mockery. Even a garbled product label can make your brand look sloppy. If you’re not reviewing live placements routinely, you may be serving ads that damage trust without realizing it.
- Agency teams managing dozens or hundreds of accounts: If an AI setting gets toggled accidentally or via a bug, a manual review of every active campaign becomes an enormous operational task. Agencies told Business Insider that campaign-level controls are insufficient when a single misconfiguration can ripple across clients.
- IT and Windows administrators in mid-size companies: While the ad tools run in the browser, they are part of the broader Microsoft-centric workflow (Meta Ads Manager is often accessed from Windows-based machines, and ad data feeds into internal reporting tools). Security-conscious IT teams may need to remind marketing departments that these AI features are not isolated toys—they are production tools that can introduce public-facing errors with compliance implications.
In short, the risk is not just a creative embarrassment. For regulated industries—healthcare, financial services, alcohol—an AI-generated copy change could run afoul of advertising rules, with legal and financial consequences.
How Meta’s Automation Ambitions Brought Us Here
Meta’s drive toward AI-driven advertising has been building for years, but the current episode crystallized after January 2026, when the company began surfacing more of its Andromeda ad retrieval system and accelerating Advantage+ rollouts. The goal, as Crypto Briefing reported, is fully autonomous ad creation by late 2026: generative creatives built from scratch, combined with automated bidding.
Past incidents show a pattern of overreach. In June 2026, Meta’s Muse Image model included a feature that let people generate images from public Instagram posts. After criticism, Meta acknowledged it “missed the mark” and removed that capability. The REI debacle and the True Classic model swap happened within the same timeframe, suggesting that as the platform piles on new AI features, existing safeguards fail to keep pace.
Worth noting: Meta has not publicly shared an opt-out mechanism at the account level for all AI creative modifications. Instead, advertisers must navigate a patchwork of settings, some off by default, some on, and some that can become enabled without clear notification. The “AI info” label is a step toward transparency, but it does nothing to stop incorrect edits from going live.
Steps to Regain Control of Your Ad Creatives Right Now
Here is what the reports and advertiser experiences indicate you should do today:
- Audit your Advantage+ creative settings immediately
In Ads Manager, open any active campaign and check under the Ad Creative section for “Advantage+ creative enhancements.” Look for options like “Adjust brightness and contrast,” “Apply music,” “Add 3D depth,” and “Optimize text.” Switch off any that do not align with your campaign’s needs. Do this for every ad set and every placement. - Review all live ads manually
Even if you believe AI enhancements are off, pull up the actual ad as shown on Facebook and Instagram. Compare the visual against your approved master file. Check for distorted products, extra limbs, nonsensical text, or wrong colors. REI’s disaster was live for an unknown period before it was noticed. - Incorporate a post-launch check into your workflow
For high-stakes campaigns—holiday pushes, product launches, brand-awareness ads—schedule a review 30 minutes after launch, then again the next morning. Meta’s systems can modify creative even after approval, so a one-time pre-flight check is no longer enough. - Consider opting out of all automatic creative enhancements at the account level
While Meta does not offer a universal kill switch, you can reduce risk by choosing “Don’t use any enhancements” in each campaign’s settings. If you manage many accounts through Business Manager, standardize this as a policy. - Train your team on the “AI info” label
The new label, appearing on ads after the fact, is a signal that Meta’s AI may have altered the creative. If you see it on a piece you didn’t expect, investigate quickly. - Document all current settings with screenshots
Given reports of settings becoming enabled without a user action, keep a dated record. This can be critical if you need to dispute a charge or demonstrate that your brand did not authorize a change.
What’s Next for AI-Generated Advertising
Meta is unlikely to reverse course on automation. The revenue opportunity is too large, and the company has telegraphed that full AI ad creation is the destination. What may change is the severity of the backlash: if high-profile brands continue to suffer public embarrassment, Meta may be forced to introduce more granular, account-level controls—or at minimum, clearer consent dialogs.
For advertisers, the lesson is clear: AI enhancements are not a harmless preview layer. They are production tools that can overwrite human decisions after the campaign is live. Until Meta delivers better safeguards, the only reliable defense is old-fashioned manual review—and a healthy skepticism toward any feature labeled “smart.”