AI Labels Flag the Tool, Not the Lie: Why Ad Disclosures Fall Short

By James Calloway

In October 2025, three AI-generated clips of police detaining and pepper-spraying protesters racked up millions of views across X, Instagram and TikTok. The videos carried visible Sora watermarks. Users cited them anyway as proof the events happened, according to NewsGuard. That's the problem in one sentence: labels tell you AI was used, not what's fake.

The concern has mostly centered on politics. But advertising has the same exposure, and synthetic content there can do something worse — make a product claim look proven.

How a fake doctor exposed the labeling gap

Full Fact recently investigated a campaign showing a supposed doctor, claiming affiliation with Guy's and St Thomas' NHS Foundation Trust, testing a weight-loss patch and appearing to grow thinner on camera. The trust called the posts "completely fraudulent." Google's Gemini detected SynthID watermarks confirming AI involvement.

But the watermark only confirmed AI was in the mix. It didn't say where the fiction ended. Was it just the doctor's face? Or were the medical authority, the hospital connection, the personal experience and the physical transformation invented too?

That's the blind spot. A generic label flags the production tool, not the illusion. Consumers still have to guess where reality stops.

When does a simulation become evidence?

This isn't an isolated case, and it isn't new.

• In 2025, the Australian Medical Association warned that deepfakes impersonating prominent doctors promoted unproven products. Clinicians found out only when patients asked whether to swap prescribed treatments for advertised miracle cures.
• In 2026, Chinese regulators launched an enforcement campaign against businesses mass-producing fabricated consumer-experience posts.
• The legal principle predates generative AI. In FTC v. Colgate-Palmolive Co., a shaving cream ad swapped a plexiglass prop for real sandpaper. Viewers were misled because the simulation was presented as a genuine product test.

The Colgate test still works: does an ad merely illustrate a claim, or does it ask viewers to treat a synthetic scene as real-world proof?

What do current labels actually cover?

Platforms and regulators have built real machinery. Google displays "How this ad was made" panels, Meta attaches "AI info" labels, and TikTok offers "AIGC" self-disclosure controls. The EU's Article 50 rules require transparency for synthetic media. China mandates explicit AI markers. New York requires ads to identify synthetic performers.

Yet these systems mostly tell consumers that AI was used — not what it changed or why it matters. The same label can sit on an ad with a decorative AI background and one with a wholly fabricated testimonial. Those uses aren't equally misleading, but a generic label can't tell them apart.

Some policies handle the clear cases. The Advertising Standards Council of India's 2026 draft guidance names fabricated experts, endorsements and exaggerated results as high-risk uses that labeling alone can't fix. TikTok Shop bans AI-generated professional identities selling health products and alterations that misrepresent performance. The harder question: what about lawful simulations that could still be mistaken for real evidence?

Why scale makes this urgent

The gap is closing in on the industry fast. In 2025, advertisers created three times as many Gemini-generated assets on Google as in 2024 — nearly 70 million in the fourth quarter alone. Meta has reportedly aimed to automate ad creation and targeting with AI by the end of 2026. These tools sit inside ad-creation systems, so safeguards can't depend on after-the-fact complaints. They have to work across millions of ads, systematically.

The fix isn't labeling every synthetic pixel. Building on the IAB's January 2026 disclosure framework, regulators could take a risk-based approach: require specific disclosures whenever AI material appears to provide evidence for a product claim. Three outcomes follow:

• Routine editing. Basic edits and decorative backgrounds get no extra warning.
• Lawful simulations. Synthetic media that could pass as real evidence must state what it represents — say, "AI reenactment of a verified customer testimonial" or "simulated demonstration; performance claim based on laboratory testing."
• Outright rejection. Fabricated endorsements, invented expert opinions and unsupported results are blocked. That mirrors FTC guidance, which permits AI avatars but prohibits fake testimonials attributed to them.

Platforms could make this work through questions they already ask. Advertisers would disclose the role of any AI-created person or scene, whether the material depicts product performance or results, and what real-world evidence backs the claim. Meta, Google and TikTok already accept advertiser declarations, so these fit existing workflows. Platforms should also keep a record of the answers and the notice shown — an auditable trail that one industry proposal calls a served disclosure receipt.

Run the weight-loss patch ad through that system. The "AI was used" label becomes the starting point, not the finish line. The advertiser would have to identify the doctor as synthetic, classify the transformation as a performance claim, and substantiate the weight-loss claim. A falsely real doctor, a fabricated testimonial or an unsupported result means the ad doesn't run.

The point, in Sourin Chakrabarti's framing, is labels that answer the question consumers actually face: not just whether AI made the ad, but whether AI made up the reason to buy.