On clicks, the strongest evidence says yes, with a condition. In January 2026 Taboola published a study by researchers at Columbia, Harvard, the Technical University of Munich and Carnegie Mellon on its own ad data: more than 500 million impressions and 3 million clicks. The raw average click-through rate was 0.76% for AI ads against 0.65% for human ads, and the two performed comparably under the tightest statistical controls. AI ads that did not look AI-made did best of all groups, and raters judged 45% of the AI ads likely or definitely human-made.
The peer-reviewed version, “AI in Disguise” in Marketing Science, compares 4,633 sibling ads: AI-made and human-made images that the same advertisers ran in the same campaigns at the same time. Across more than 369 million impressions and 2.5 million clicks it finds no detectable average click-through disadvantage for AI images. The authors also find that people link sharper images and larger faces with human-made ads.
Read the scope before you rely on it. This is one research programme on Taboola native display images. It measured clicks, not purchases, and it did not test UGC video on Meta or TikTok.
The UGC-specific evidence is thin. AdTestBench, an independent one-person blog, graded 12 public reports and found none with a randomised split of AI and human UGC and a published method. The three matched comparisons it did find all favoured human UGC per ad. One, posted by the owner of an AI UGC agency without spend or sample, reported 2.1% conversion from click for AI against 3.4% for human UGC, with production at $12 against $280.
Labels change behaviour too. In survey experiments with 1,000 people each in the US, the UK and Germany, the Nuremberg Institute for Market Decisions found that people were less inclined to click on products in ads marked as AI-generated. Together, the evidence supports a narrow claim. AI UGC can match creator UGC on attention when it looks and sounds real, creators still look stronger where personal experience carries the sale, and AI’s clear advantage is the number of angles a brand can afford to test.
AI people in ads need a label in the EU, New York and California
In the EU, Article 50 of the AI Act has applied since 2 August 2026. Article 50(4) requires deployers to disclose deep fakes, which Article 3(60) defines as AI-made content that resembles real persons, objects or events and would falsely appear authentic or truthful. The Commission’s Guidelines (published 20 July 2026, re-issued 5 November 2026) name realistic AI-generated human avatars in paragraph 113, so an invented but photorealistic AI UGC presenter can count. In an ad job the deployer is usually the studio or agency, and the brand as well when it decides how AI is used. A brand that only commissions an agency is not one, according to paragraph 14 of the same guidelines.
The Code of Practice on Transparency of AI-generated Content (10 June 2026) specifies “AI” as the visual label, at the start of a video and, where possible, at regular intervals. Article 50(5) requires disclosure at the latest at first exposure. Fines reach EUR 15 million or 3% of worldwide turnover, whichever is higher, and the lower of the two for SMEs. Separately, EU consumer law as amended by Directive (EU) 2019/2161 bans fake consumer reviews, so an AI presenter who claims to have used a product is likely to be treated as a misleading testimonial, label or not. Who labels an AI ad under Article 50 goes through the provider and deployer split in detail.
The US has no federal law like the AI Act. Executive Order 14365 of 11 December 2025 seeks a “minimally burdensome national standard” and set up a task force to challenge state AI laws. The FTC’s rule on consumer reviews and testimonials, in force since 21 October 2024, bans testimonials from people who do not exist, such as AI-generated fake reviews, and from people without real experience of the product. Civil penalties reach $53,088 per violation, according to Benesch.
New York’s law took effect on 9 June 2026. California’s SB 1050 was signed on 16 September 2026 and applies from 1 January 2027, and Mayer Brown sets out its model wording. What the New York and California synthetic performer laws require covers both, line by line. AI Vidia labels every photorealistic AI person, agrees the label wording per platform and market before delivery, keeps the AI person from claiming an experience it cannot have, and records how each file was made. Check your own campaign with counsel.
When to use AI UGC, creator UGC or both
Each format does a different job. AI UGC is strongest where you are still looking for the angle: which hook, persona or format earns attention. It lets you test many concepts for the price of a few creator videos, refresh every week without waiting for the post, and try a risky idea before anyone films it.
Creator UGC earns its price where a real experience carries the sale: skincare and supplements where results are personal, products that must be handled on camera, and any ad built on a testimonial. A creator also brings an audience and a voice for whitelisted and partnership ads.
The combination is the efficient route. AI UGC finds the angle quickly, and a creator then films the proven script, so the expensive round only happens once a concept has earned it. Answer four questions to see where your next test fits.
If the tree points you to AI UGC or to the combined route, AI UGC ads made for your product shows the formats, labels and review steps AI Vidia uses.
Price one creator round against one AI round this week
Open your last creator invoice and write down four numbers: the fee, the usage rights, the shipping, and the days from brief to the file you ran. Add them up as cost per usable video and days per angle. Put the same two numbers for an AI round next to them, using the prices on the AI Vidia pricing page or your own studio’s quote. The lower cost per tested angle tells you which format should run your next 30 days of testing.