| Question about AI ads | What the evidence says | How to use it |
|---|---|---|
| Do AI ads work on clicks | Field tests show AI-made or AI-assisted ads matching or beating usual creative on click-through rate in narrow, product-led campaigns | Use AI as a production tool for direct, product-driven briefs |
| AI ads vs human ads on sales | In the largest head-to-head test, human-made ads scored higher on predicted short-term sales and long-term brand effect | Do not scale an AI winner on clicks alone |
| Emotion and brand memory | Viewers tended to spot AI ads, found them less engaging, and showed weaker memory activation, even for high-quality ones | Keep story, insight and point of view with a human director |
| Disclosing AI use | Labelling an ad as AI-made lowered clicks in one field study, yet most respondents in another study said companies should disclose | Plan disclosure into the creative and check each platform's rules |
| Who owns the idea | AI struggles with story and does best on direct briefs | A director or strategist owns the insight, story and tone |
| How to test AI ads | Studies measure one thing at a time, so clicks, recall and predicted sales can disagree | Same brief, budget and audience as the human control, then add a recall or brand-lift check |
Do AI ads work? The honest answer is: sometimes, and mostly when people stay in charge of the idea. Controlled studies show AI-made ads can match or beat human work on clicks in narrow, product-led tests, yet they still trail human craft on emotion, brand memory and predicted sales.
Quick takeaways
- Clicks: in field tests, AI-made or AI-assisted ads have matched or beaten the usual creative on click-through rate.
- Sales and brand: human-made ads still scored higher on predicted sales and long-term brand effect in the largest head-to-head test.
- Disclosure: labelling an ad as AI-made lowered clicks in one field study, so plan disclosure as part of the creative.
- Briefs: AI does best on direct, product-driven briefs and struggles with story and point of view.
What Does the Evidence on AI Ads Actually Measure?
Most studies measure one thing at a time: clicks, recall or a predicted sales score. None of them settles the whole question, so read each result against what it measured.
| Study | What it compared | What it found |
|---|---|---|
| Ipsos and Syracuse Newhouse, May 2026 | 10 human-made ads vs fully AI versions from the same brief, 3,000 US viewers | Human ads stronger on predicted short-term sales and brand effect |
| MIT IDE (Kumar and Kapoor) | Personalised AI avatar video vs image and generic video ads, 21,000+ consumers in India | AI video drew more clicks than both |
| NYU Stern and Emory, Nov 2025 | Human, AI-edited and fully AI display ads | Fully AI ads raised clicks; disclosure cut them |
| NIQ, Dec 2024 | AI ads of varying quality, 2,000+ viewers, EEG on about 150 | Weaker memory activation, even for high-quality AI ads |
Where AI Ads Have Performed Well
AI ads win most clearly when the job is narrow and measurable. In the MIT IDE experiment, personalised AI video ads sent over WhatsApp by an Indian e-commerce retailer earned a 9.4% higher click-through rate than personalised image ads (MIT IDE). The marketing team kept control of the message and used AI as a production tool, not as the creative engine.
The NYU Stern and Emory study found that fully AI-generated display ads lifted click-through rate by up to 19% over expert-made ads (NYU Stern). Ads that were human-made and then edited with AI showed no significant gain.
Where AI Ads Still Fall Short
The weak spots are emotion, distinctiveness and memory. In the Ipsos and Syracuse test, human-made ads scored 11 points above the Ipsos sales-validated benchmark on average while the AI versions scored five points below it (Syracuse University). Ipsos reports the human ads were 14% stronger on its short-term sales index and 17% stronger on long-term brand health (Ipsos).
NIQ's neuroscience work points the same way. Viewers tended to spot AI ads and found them less engaging, and even the high-quality ones produced weaker memory activation than traditional ads (NIQ). Low-quality visuals made viewers work harder and pulled attention off the message.
Does Telling Viewers an Ad Is AI-Made Change Results?
In one field study, yes. Labelling an ad as made with generative AI cut its click-through rate by 31.5% (NYU Stern). At the same time, 79% of respondents in the Ipsos study said companies using AI should have to disclose it (Ipsos).
Our read: hiding AI use is not a strategy. Disclosure rules differ by platform and market and change often, so check the current rules for each placement with your legal team. This is not legal advice.
How Should Brands Use These Findings?
Treat AI as a way to make more and better-targeted versions of a strong human idea, then let testing decide. In practice that means three habits.
- Keep the concept human: a director or strategist owns the insight, story and tone.
- Use AI for range: hooks, languages, formats and personalised variants of one approved idea.
- Measure beyond clicks: add recall or brand-lift checks before scaling a winner.
Our performance creative formula shows how to structure variants for a test, and our guide to choosing AI content helps decide where AI fits in your mix.
FAQ
AI-generated ads can be effective on clicks in narrow, product-led campaigns. Field studies from MIT IDE and NYU Stern found click-through gains, but the same research base shows weaker results on emotion, brand memory and predicted sales, so test before you scale.
Consumers don't always notice AI ads, but they often rate them lower. Ipsos found AI ads trailed human work on entertainment, uniqueness and talkability, and NIQ found viewers described many AI ads as less engaging, especially when visuals were poor.
AI ads tend to draw backlash when they look uncanny, replace a much-loved human tradition or seem made only to cut costs. Low visual quality makes it worse, because viewers notice the tool before the message, which NIQ linked to weaker engagement.
Disclosing AI use can lower clicks. The NYU Stern and Emory study measured a 31.5% drop in click-through rate when ads were labelled as AI-made. Many consumers still expect disclosure, so plan honest labelling into the creative rather than avoiding it.
Brands should test AI ads with the same brief, budget and audience as the human control, changing only who made the ad. Track clicks and conversions, then add a recall or brand-lift check, because the studies show those measures can disagree.
Brief Your First AI Ad Test
Bring us one approved idea, your hero product and the markets you sell in, and we will plan human-directed performance ads built to be tested. Schedule a call to scope it.



