AI Ads vs Human Creators: What the 2026 Research Says

Human-made ads still beat AI-generated ads, even though most viewers can no longer tell them apart. Here is what the research shows:

• In a 2026 Ipsos test, human ads scored 14% higher on predicted short-term sales impact and 17% higher on long-term brand equity.
• Only 25% of people who saw an AI ad were at least somewhat confident it was AI-made.
• European research shows that an "AI-generated" label on its own lowers purchase intent.
• Since 2 August 2026, realistic AI people in EU ads can count as deepfakes that need a label.

The practical answer: use AI to produce faster, and real creators for the faces.

Why brands are trying AI ads

AI video is cheap, quick and getting better. Paris-based Arcads, which sells AI actors for UGC-style ads in more than 35 languages, reports more than 6,000 clients and about 100,000 video assets a month (Net Influencer).

The platforms are heading the same way. Meta wants brands to be able to create and target a complete ad from just a product image and a budget by the end of 2026, as the Wall Street Journal reported in June 2025 (Campaign Asia). Kantar found that 70% of marketers already use generative AI (Kantar).

Are AI ads as effective as human ads?

Not yet, according to the most rigorous head-to-head test published so far. In May 2026, Ipsos and Syracuse University's Newhouse School took 10 human-made ads and recreated each one with Google's Gemini and OpenAI's Sora 2. They then showed all 20 ads to 3,000 US consumers (Ipsos).

The human ads were 14% stronger on short-term effectiveness, which Ipsos uses to predict short-term sales impact, and 17% stronger on long-term brand equity. Yet viewers struggled to spot the difference: only 25% of people who saw an AI ad were at least somewhat confident it was AI-made, and another 40% of all viewers couldn't tell either way.

"AI may be closing the perceptual gap, but it still has work to do on the effectiveness gap." (Ipsos U.S. and Syracuse University's Newhouse School)

In other words, the AI ads didn't look fake, but they moved people less. A test of 20 US ads isn't the final word, and the gap may close as the tools improve. Still, it is the clearest controlled comparison published so far.

Does an AI label change how people react?

Yes, and that matters now that disclosure is becoming mandatory. The Nuremberg Institute for Market Decisions (NIM) surveyed 1,000 people in each of three countries, including Germany. Only 20% said they trust AI itself. In controlled experiments, people judged an ad labelled "AI-generated" more harshly than the identical ad labelled human-made, and were less willing to research or buy the product (NIM). NIM calls the effect "small but meaningful", so it is a headwind rather than a wipe-out.

Worry runs high on both sides. 57% of consumers worry generative AI will be used to create fake ads (Kantar). 95% of advertisers have concerns about AI in creator marketing, with "the loss of human connection" as their top concern (IAB).

Backlash can be public, too. In December 2025, McDonald's Netherlands withdrew a 45-second Christmas ad generated entirely by AI after criticism on social media (NBC News). In fairness, some critics disliked the concept as much as the AI.

What the EU AI Act means for AI avatar ads

Since 2 August 2026, Article 50 of the EU AI Act has required anyone deploying AI to disclose deepfakes (Paul, Weiss). Under the Commission's guidelines, "a realistic synthetic influencer testing out a sponsored real product" counts as a deepfake, and the label has to appear at the latest at first exposure (Davis+Gilbert). Fines go up to €15 million or 3% of worldwide turnover, whichever is higher.

Put that together with NIM's results. An AI person presented as a happy customer now needs a label, and the label itself costs trust. A real creator filming with consent and agreed usage rights needs no deepfake label. Our EU AI Act guide for advertisers covers the details. This is not legal advice.

Where AI helps and where real creators win

You don't need to avoid AI. Just keep it away from the jobs that depend on trust.

AI works well behind the scenes: script drafts, hook ideas, copy variants, cut-downs and new aspect ratios from existing footage, and translating or dubbing real creator videos for new markets. That matches how advertisers already use it, mostly for content editing (49%), creator briefs (46%) and personalisation (45%), according to the IAB.

Real creators should handle the face on screen, the product in hand, the testimonial and any claim that implies "someone like you tried this". They can also run your ad from their own handle as a partnership ad, which a synthetic avatar can't do because it has no account or audience (why the creator's handle matters).

That's how Noovid works. Vetted creators across the eurozone film your product, and you can build more variants from that footage. Partnership ads are available as a paid add-on. For a side-by-side of workflows, see UGC vs AI video tools.

FAQ

Can people tell if an ad is AI-generated?

Mostly not. In the Ipsos study, only 25% of people who saw an AI ad were at least somewhat confident it was AI-made.

Do AI avatar ads need a label in the EU?

It can. Since 2 August 2026, a realistic AI-generated person in an ad can count as a deepfake under Article 50 of the EU AI Act, and deepfakes must be disclosed. Check your case with a lawyer.

Should I use AI or real creators for UGC ads?

Both, for different jobs: AI for scripts, variants and dubbing, real creators for faces, demos and testimonials.

Sources

• AI Ads Are Good Enough, and That's the Problem, Ipsos, May 2026: 14% and 17% effectiveness gap, 25% detection confidence, 40% unsure, study design.
• Transparency without trust, NIM (Nuremberg Institute for Market Decisions), 2024: AI labels lower evaluations and willingness to research or buy; 20% trust AI; 1,000 respondents per country.
• Creator Economy Ad Spend to Reach $37 Billion in 2025, IAB, November 2025: 95% of advertisers have AI concerns; loss of human connection; AI uses (49%, 46%, 45%).
• Media Reactions 2025, Kantar, October 2025: 57% of consumers worry about GenAI fake ads; 70% of marketers use GenAI.
• Arcads.ai Secures $16M Seed Funding, Net Influencer, December 2025: 6,000+ clients, ~100,000 assets a month, 35+ languages.
• Meta aims to fully automate ad creation with AI by 2026, Campaign Asia, June 2025: Meta's plan for complete AI ads from a product image and budget.
• McDonald's pulls AI-generated Christmas ad after social media backlash, NBC News, December 2025: McDonald's Netherlands withdrawal.
• EU Finalises Transparency Rules for AI-Generated Content, Paul, Weiss, August 2026: Article 50 dates and fines.
• EU AI Act Guidance Expands AI Disclosure Rules for Advertisers and PR Teams, Davis+Gilbert, 2026: synthetic influencer deepfake example, disclosure at first exposure.