AI vs Human Video Ads: What the Ipsos-Syracuse Study Actually Shows About Performance, Creativity and Sales Impact

The Ipsos-Syracuse study on AI vs human video ads is a controlled 2026 comparison of 10 human-made 30-second brand ads and 10 fully AI-generated counterparts shown to 3,000 U.S. consumers. The research tested whether AI could execute the same strategic briefs as human creative teams, then measured perception, short-term sales potential, long-term brand effects, emotional response, creativity, and storytelling. The central result is clear but narrower than many headlines suggest: viewers often struggled to identify AI-made ads, yet the human-made ads performed better on the study's validated effectiveness measures. The study matters to advertisers, agencies, creative teams, and marketing leaders because it separates visual realism from advertising effectiveness.

What the Study Actually Tested

The research tested creative execution under a controlled strategy, not the full advertising process from audience research through media buying. Ipsos worked with Syracuse University's S.I. Newhouse School of Public Communications to compare existing human-made commercials with AI-generated versions built from the same underlying strategic direction.

The researchers selected 10 existing 30-second ads that had been produced before 2021, which reduced the chance that generative AI had played a role in the original production. The brands covered consumer packaged goods, retail, automotive, technology, financial services, and related categories. The set included Cheerios, Chewy, Febreze, Fiat, H&M, Old Navy, Herbal Essences, Ray-Ban Meta, TurboTax, and Visa.

Professor Adam Peruta handled the AI production workflow. Google Gemini was used to deconstruct each original commercial and reverse-engineer a creative brief that could have produced it. Gemini then created a new concept and shot list from that brief. The shot list was passed to OpenAI's Sora 2, which generated the AI counterpart. Ipsos states that the final AI ads had no human creative intervention. Some generations needed several attempts before they reached a reasonable quality level for consumer testing.

That design gives the study a useful control. Human and AI routes worked from the same strategic basis. The research therefore focuses heavily on execution quality. It does not test whether AI can independently discover an audience insight, choose a positioning strategy, define an offer, create a media plan, or learn from live campaign data.

How the 20 Ads Were Evaluated

Ipsos used its Creative|Spark ad assessment system to evaluate all 20 ads with 3,000 U.S. respondents. The testing approach is designed to expose ads in a distracted, high-stimulus setting rather than relying only on quiet, direct evaluation.

Two core measures are central to the performance result. Creative Effect Index, or CEI, is designed to capture an ad's ability to drive short-term sales impact. Equity Effect Index, or EEI, measures contribution to longer-term brand health.

These are validated predictive measures. They are not the same as observed purchases from a live media campaign. The study did not report a matched field experiment in which equal media budgets were spent on the human and AI versions and actual incremental sales were tracked.

That distinction matters because headlines that say one type of ad drives more sales can sound like direct transaction data was collected. The research supports a more precise statement: human-made ads scored higher on validated measures linked to short-term sales impact and long-term brand health.

Ipsos also applied its MISFITS Mindset framework, which groups creative performance around Creative Experiences, Empathy and Fitting In, and Creative Ideas.

Across the 10 pairs, human and AI ads were broadly comparable on these high-level dimensions, though human work had modest advantages. Ipsos says its broader MISFITS framework is associated with a 20% lift in creative effectiveness when all three experiences are delivered. That 20% figure is a framework benchmark, not the measured human-versus-AI gap in this experiment.

The more visible differences appeared in entertainment, uniqueness, talkability, eye-catching quality, imagination, emotional engagement, and storytelling.

The Main Performance Result: Human Ads Won on Average

The study found an average effectiveness advantage for the human-made commercials. Ipsos reports that the human ads were 14% stronger on short-term effectiveness and 17% stronger on long-term effectiveness.

The Syracuse University summary adds another way to read the short-term result. It says the human-made ads averaged 11 points above the benchmark, while the AI-generated ads averaged five points below it.

That difference describes the study's predictive short-term effectiveness measure rather than actual recorded sales.

The result does not mean every human ad beat every AI ad. Some AI ads performed effectively, and some human ads were weaker than others. The average gap is the meaningful finding. It shows that AI-generated video could produce credible commercial work in this controlled test, but credible execution did not produce equal average business-effect scores.

The distinction between average performance and universal performance is important. Saying AI video ads do not work goes beyond the research. Saying the fully AI-generated ads in this 10-pair test underperformed the human-made originals on average matches the design much more closely.

Viewers Often Could Not Reliably Identify the AI Ads

The study found that visual realism had advanced far enough that many respondents could not confidently identify which commercials were AI-generated. The research therefore separates detection from effectiveness.

The public research materials contain a reporting inconsistency that should be preserved rather than hidden. The official Ipsos page says 25% of viewers of AI ads were at least somewhat confident that the spot was AI-made.

Syracuse University's release says 13% of viewers who saw an AI-generated ad were at least somewhat confident it was made by AI, and that the same 13% share suspected human-made ads were AI-generated.

Both sources report that 40% of all viewers were uncertain either way.

Without the underlying questionnaire wording and full cross-tabulation, the difference between 25% and 13% cannot be resolved safely from the public summaries alone. The stable conclusion is that detection was weak and uncertainty was high.

That finding matters because the lower effectiveness of the AI ads cannot be explained simply by viewers recognizing the technology and rejecting it. Many people did not confidently identify the production method, yet human-made work still received stronger ratings on several creative qualities and higher average effectiveness scores.

The Perceptual Gap Closed Faster Than the Effectiveness Gap

The research shows that looking like a professionally produced advertisement is not the same as producing the same advertising response. AI video reached a level of surface credibility that often passed casual detection, but the human work retained an advantage in dimensions connected with attention, imagination, emotional response, and memorability.

Ipsos reports that human-produced ads were rated as more entertaining, more unique, and more likely to generate conversation. Viewers also rated the human work as more eye-catching and imaginative.

In direct comparison, Ipsos reports that consumers saw AI advertising as less creative, less emotionally engaging, and less informative.

This is one of the strongest ideas in the research. AI video quality can become technically convincing before creative quality becomes equally persuasive.

A realistic face, polished lighting, stable motion, correct product depiction, and acceptable editing can remove obvious production defects. Those improvements do not automatically create an original idea, emotional tension, a memorable point of view, or a strong brand story.

For marketing teams, production quality and creative effectiveness should therefore be treated as separate evaluation layers. A video can pass a visual-quality review and still fail an effectiveness review.

AI Performed Better on Direct, Product-Led Briefs

The strongest AI-generated ads in the study tended to use direct, functional advertising structures. Ipsos highlights product-driven work with a clear problem, a clear solution, and a clear reason to believe.

The AI versions for Febreze and Herbal Essences are cited as examples of this pattern.

These ads relied on familiar structures with strong product logic. The model had a defined task and a large body of conventional advertising patterns from which to construct the execution.

This result suggests a practical distinction between creative tasks. Generative video systems can be useful when the advertising job is explicit.

Product demonstrations, feature explanations, simple before-and-after structures, offer-led variations, and other direct formats give the model a narrower set of decisions.

The result does not establish that AI will always win in lower-funnel advertising. The study did not run a broad lower-funnel media experiment. It shows that within this 10-pair video set, the stronger AI outcomes were concentrated in direct, product-centered creative.

AI Struggled More With Storytelling, Emotion and Point of View

The weaker AI results appeared when the brief depended on narrative, emotional nuance, humor, or a distinct creative point of view. These tasks require more than producing plausible scenes in the correct order.

Ipsos uses the Fiat pairing to illustrate the problem. The creative direction asked for energy, confidence, humor, contrast, and attitude. The generated execution did not create the same emotional strength expected from the brief.

The wider Ipsos storytelling research cited in the study found storytelling in 49% of ads. In the AI set used for this experiment, storytelling appeared only 30% of the time.

The comparison should be read carefully because the 49% figure comes from a broader ad research base, not from the 10 human ads in the matched experiment.

The human-made Chewy holiday commercial was the strongest individual ad in the 20-ad set.

Ipsos attributes its performance to believable emotion, a real human tension, and a story closely connected to the brand's role. The AI counterpart remained around average but did not reproduce the same emotional strength.

The useful lesson is not that machines cannot produce stories. Generative video can create narrative sequences. The challenge is whether the story contains a human observation that creates meaning, emotional movement, and brand relevance at the same time.

The Cheerios Pair Shows Why the Brief Matters So Much

The highest combined effectiveness across a human and AI pair came from Cheerios. Both versions performed near the top of the Ipsos database, which makes the brief one of the most informative parts of the study.

The strategic idea connected heart health with being present for people who depend on you. That proposition gave both production routes a clear human motivation, a product reason, and an emotional purpose.

This result limits any simple human-versus-AI interpretation. The creator mattered, but the input strategy also mattered.

When the brief contained a clear human insight, the AI execution closed much of the gap. When the creative direction depended on less explicit emotional interpretation, the AI route had more difficulty.

For marketers, this means generative production does not remove the need for audience understanding. The quality of the brief still controls message focus, emotional territory, brand role, tone, and the type of creative idea the production system has to express.

What the Study Does Not Prove

The study does not establish that AI-generated advertising is generally worse than human advertising across every format, model, funnel stage, market, audience, or production workflow. Its design is much narrower.

It tested 10 matched pairs of 30-second video ads. It did not test hundreds of static ads, short social videos, creator-style ads, carousels, product images, landing pages, email creative, or every other common performance format.

The study used a specific AI pipeline based on Gemini and Sora 2. Results from one model combination should not automatically be treated as results for all current or future generative systems.

The strategy was inherited from human-made advertising. The AI system did not originate the market research, audience insight, positioning, or creative strategy from a blank starting point. That means the study is stronger as a test of AI execution than as a test of AI strategy.

The AI condition also excluded human creative intervention. Many commercial workflows use writers, art directors, editors, strategists, legal reviewers, producers, and media teams around AI generation.

The study therefore does not answer the question that many companies actually face, which is whether a human team using AI can outperform a human team without AI.

The research also relied on predictive ad-testing measures rather than matched live sales data. The CEI and EEI measures are designed around business outcomes, but the experiment was not a direct randomized media test of revenue.

Why the Model and Workflow Have a Shelf Life

The result is a snapshot of a specific 2026 production pipeline, not a permanent score for generative video. The study used Gemini for brief reconstruction and concept development, then Sora 2 for video generation, with no human creative intervention in the AI condition.

That matters because generative video systems change quickly. OpenAI's current support material says the Sora web and app experiences were discontinued on April 26, 2026, and the Sora API was discontinued on September 24, 2026.

The study therefore remains useful as a controlled test of the workflow it used, but it should not be read as a direct benchmark for every video model available after that period.

Model changes also make reproducibility important. Public reporting explains the broad pipeline, but it does not provide every prompt, generation setting, rejected output, or full iteration history.

Those missing details limit how precisely another team can reproduce the AI condition.

The Study's Biggest Methodological Strength

The strongest part of the design is the attempt to hold strategy constant while changing the producer. That makes the comparison clearer than a loose collection of unrelated human and AI advertisements.

Using the same strategic basis reduces several sources of noise. Both versions should be trying to solve a similar communication problem, express similar brand goals, and work from similar messaging priorities.

That gives the comparison more meaning than simply comparing a random AI ad with a random human ad.

The use of original ads produced before 2021 also reduces ambiguity about whether the human set had generative AI assistance.

The 3,000-person U.S. sample is much larger than a small creative-review panel, and the use of the same testing system across all ads creates a consistent measurement structure.

The design still has limits, but the controlled brief makes the central comparison useful: given comparable strategic direction, fully autonomous AI execution produced weaker average results than established human-made executions in this sample.

The Biggest Limitation Is the Missing Hybrid Condition

The most commercially relevant comparison was not included: human-only versus AI-only versus human-plus-AI. The study compares two endpoints, while many production teams operate between them.

A hybrid workflow can use people for research, insight, concept selection, scripting, brand judgment, cultural context, and final editing, while AI handles storyboards, visual exploration, scene generation, localization, versioning, or production speed.

The Ipsos authors themselves point toward a human-plus-AI future. Their recommendations place more AI use around product work, variations, and scale, while keeping people central to emotional storytelling and brand-building work.

A future study with three matched conditions would help isolate the value of AI assistance more clearly.

It would also better reflect real agency and in-house workflows, where teams rarely publish the first autonomous generation without review.

Disclosure and Trust Add a Separate Performance Question

The Ipsos-Syracuse study also raises a transparency issue. Ipsos reports that 79% of respondents said companies using AI should have to disclose that use.

Disclosure is not the same variable as creative quality. A strong AI-generated ad can still face a different consumer response once viewers learn how it was produced.

Other 2025 and 2026 research helps explain why this deserves separate testing. A New York University and Emory University study of visual generative advertising reported that fully AI-generated display ads produced click-through rates up to 19% higher than human-expert ads in its tested settings, while disclosure that an ad was AI-generated reduced click-through rate by 31.5%. Those results concern visual display advertising and click behavior, not 30-second brand video or long-term brand health.

The lesson is that marketers should measure at least two different effects. One is the quality of the creative itself. The other is the audience response to visible or disclosed AI involvement.

Why Other AI Advertising Studies Do Not Automatically Contradict Ipsos-Syracuse

Research published around the same period can appear contradictory because different studies measure different formats, outcomes, and production conditions. The Ipsos-Syracuse work focuses on fully AI-generated 30-second video ads and validated brand-effect measures.

A large-scale 2026 field study using more than 500 million ad impressions compared AI-generated and human-made display creative. The raw click-through rate was 0.76% for AI ads and 0.65% for human-made ads, while tightly controlled comparisons found broadly similar performance.

That study examined click behavior at very large scale, not emotional storytelling in matched 30-second commercials.

A separate analysis of hundreds of ads that used generative AI found that AI use could appear across the full performance range. Ads with less obvious AI involvement performed better on branded cut-through, while visibly artificial execution tended to perform worse.

Current 2026 creative-testing data also reports that generative ads can perform near conventional work in some markets and use cases.

These results can exist together. Click-through rate, predicted sales impact, brand equity, emotional response, memorability, and brand linkage are different outcomes.

Static display creative, short-form performance ads, and 30-second brand commercials also demand different creative skills.

The broader research picture as of October 28, 2026 is conditional, not binary. AI can match or beat human work on some direct-response tasks. Human-made video still shows an advantage in the Ipsos-Syracuse test when emotional storytelling and longer-term brand effects matter.

A Better Way to Test AI vs Human Video Ads in Real Campaigns

Brands should treat the Ipsos-Syracuse findings as a starting point for controlled testing rather than a universal rule. A practical experiment should isolate the production method while keeping the commercial conditions as similar as possible.

Use the same audience, offer, landing page, campaign objective, media window, budget structure, and core strategic brief.

Create at least three routes when resources allow: human-made, fully AI-generated, and human-plus-AI.

Choose the decision metric before launch. For performance campaigns, that can include cost per qualified lead, cost per sale, conversion rate, or incremental revenue.

For brand campaigns, measurement can include brand attention, ad recall, brand linkage, consideration, and longer-term brand tracking.

Creative diagnostics should sit beside business metrics. Measure whether viewers understood the message, remembered the brand, noticed distinctive assets, felt the intended emotion, and found the execution credible.

This helps explain why one version performs differently rather than only recording the winner.

Run enough impressions and conversions to avoid making a decision from early noise. The required sample depends on baseline conversion rates, expected effect size, budget, and the statistical method used. A fixed universal minimum would be misleading.

What Marketers Should Change in Their AI Video Workflow

The study supports a selective AI workflow. AI is most defensible where the task is clear and the output can be tested quickly, while human judgment remains especially important when the advertising job depends on empathy, cultural reading, narrative tension, humor, originality, or brand meaning.

A practical workflow starts with human-led audience research and a precise brief. The brief should define the audience, problem, desired response, product truth, brand role, tone, and reason to believe.

AI can then generate multiple visual routes, scene ideas, scripts, storyboards, or video drafts.

Human reviewers should reject work that is merely technically acceptable but lacks a memorable idea or emotional logic.

Testing should compare creative routes before large media spend. Feedback from real target audiences should inform the next generation.

AI's speed has more value when it increases the number of meaningful learning cycles, not simply the number of files produced.

The study's warning is therefore about approval standards. Looking finished is not a sufficient test. The creative must also communicate, persuade, build the brand, and produce the intended commercial response.

Quick Facts About the Ipsos-Syracuse AI vs Human Video Ad Study

The study gives marketers several concise facts that should stay attached to any interpretation of the results.

• The core research was published on May 18, 2026. A management-focused article by Adam Peruta followed on September 3, 2026.
• The experiment tested 20 video ads across 10 brands with 3,000 U.S. consumers.
• Ten original human-made ads were produced before 2021.
• Ten AI counterparts were generated from reverse-engineered briefs using Gemini and Sora 2, with no human creative intervention in the final production route.
• Human-made ads averaged 14% stronger short-term effectiveness and 17% stronger long-term effectiveness in Ipsos reporting.
• Syracuse University's summary says human ads averaged 11 points above the short-term benchmark, while AI ads averaged five points below it.
• Public source summaries disagree on whether 13% or 25% of AI-ad viewers were at least somewhat confident the ad was AI-made, while both report 40% overall uncertainty.
• AI performed best in direct product-led work and had more difficulty with storytelling, emotion, and distinct point of view.

What the Research Means as of October 2026

As of October 28, 2026, the most accurate reading is not that human ads beat AI in every setting. The study shows that fully autonomous AI video generation was already realistic enough to confuse many viewers, but the tested human commercials still produced stronger average creative and business-effect scores.

The research also shows why AI advertising performance cannot be reduced to one metric.

Some large-scale field work reports parity or advantages for AI on click-through rate. Other testing finds weaker brand or emotional outcomes when AI execution feels artificial, generic, or poorly connected to the brand.

The Ipsos-Syracuse study sits in the second category because it measures more than clicks.

The durable lesson is to separate production efficiency from creative effectiveness. Faster generation can reduce time and cost, but an ad only creates value if it attracts attention, communicates the brand, creates the intended response, and supports business goals.

For direct product communication, variations, and scalable production, AI video can already be useful.

For emotionally demanding brand work, the 2026 study supports keeping experienced human creative judgment close to the idea, story, and final edit.

The strongest operating model is not a blind choice between people and machines. It is a measured workflow where each production route is judged by the business result it is supposed to create.

The Ipsos-Syracuse study shows that AI-generated video ads have reached a point where many viewers cannot reliably tell them apart from human-made advertising. Technical realism, however, did not translate into equal advertising performance. Across the 10 matched ad pairs, human-made commercials produced stronger average scores for short-term effectiveness, long-term brand effects, emotional engagement, creativity, entertainment, and storytelling.

The research also shows that AI performance depends heavily on the type of brief. AI worked better with direct, product-focused advertising where the problem, solution, and product benefit were clear. Human creative work retained a stronger advantage when the advertising depended on emotion, humor, cultural understanding, narrative tension, or a distinct point of view.

The study should not be interpreted as proof that human advertising will always outperform AI advertising. It tested 20 thirty-second videos, 3,000 U.S. consumers, and a specific 2026 AI production workflow. It did not test every model, format, market, campaign objective, or human-plus-AI production process.

For marketers, the most useful lesson is to separate production efficiency from creative effectiveness. AI can reduce production effort, create variations, support product-led campaigns, and speed up experimentation. Human judgment remains especially important for strategy, audience insight, storytelling, emotional meaning, brand differentiation, and final creative decisions.

The best approach is therefore not to choose AI or humans by default. Brands should test human-made, AI-generated, and human-plus-AI creative under comparable campaign conditions and judge each version using the metric that matters for the business, whether that is sales, conversions, attention, recall, brand consideration, or long-term brand health.

As of 2026, AI video has become convincing enough to compete for production work. The Ipsos-Syracuse research shows that being convincing is not the same as being persuasive. The competitive advantage will come from combining AI production speed with strong human strategy, creative judgment, and disciplined performance testing.

AI vs Human Video Ads: FAQs

What Did the Ipsos-Syracuse Study Compare?
The study compared 10 professionally human-made 30-second video ads with 10 fully AI-generated versions created from comparable strategic briefs. Around 3,000 U.S. consumers evaluated the ads. The research examined short-term effectiveness, long-term brand effects, creativity, emotional engagement, storytelling, entertainment value, and viewers' ability to identify AI-generated advertising.

Did Human-Made Ads Perform Better Than AI-Generated Ads?
Yes. Human-made ads performed better on average in the study. Ipsos reported that human-created commercials were 14% stronger on short-term effectiveness and 17% stronger on long-term effectiveness. These results came from validated advertising-effectiveness measures, not from directly recorded sales transactions in live campaigns.

Could Consumers Tell Which Video Ads Were Created by AI?
Many consumers could not confidently identify AI-generated ads. The public research summaries report high uncertainty among viewers. Ipsos reported that 40% of viewers were uncertain about whether ads were AI-generated or human-made. The published sources contain different figures for confident AI detection, so the safest interpretation is that production method was often difficult for viewers to recognize reliably.

Why Did Human Video Ads Perform Better Even When AI Ads Looked Realistic?
Visual realism and advertising effectiveness measure different things. AI-generated ads were often technically convincing, but human-made ads generally performed better on creativity, emotional engagement, entertainment, uniqueness, storytelling, and attention. A realistic-looking commercial can still lack the emotional idea, narrative structure, or brand connection needed to influence viewers.

What Types of Video Ads Did AI Handle Best?
AI performed better with direct, product-focused advertising. Ads built around a clear problem, product solution, product benefit, or reason to believe were easier for AI systems to execute effectively. The research highlighted product-led examples where AI came closer to human performance than it did with emotionally complex brand storytelling.

Where Did AI-Generated Ads Struggle the Most?
AI struggled more when the creative brief depended on emotion, humor, human tension, storytelling, cultural understanding, originality, or a distinct creative point of view. These forms of advertising require more than technically correct scenes. They depend on meaningful human observations and a strong connection between the story and the brand.

Does the Study Prove That AI Advertising Is Worse Than Human Advertising?
No. The study tested only 20 video ads, 10 human-made and 10 AI-generated, using 3,000 U.S. consumers and a specific AI production workflow. It did not test every AI model, advertising format, market, audience, campaign objective, or production method. The findings show that human-made ads performed better on average within this specific experiment.

Did the Study Measure Actual Sales Generated by the Ads?
No. The study used validated predictive measures designed to estimate short-term sales potential and long-term brand effects. It did not run equal-budget live campaigns and compare actual purchases or incremental revenue generated by each advertisement. Marketers should therefore describe the results as advertising-effectiveness findings rather than direct sales results.

What Is Missing From the AI vs Human Advertising Comparison?
One important missing condition is human-plus-AI advertising. Many real marketing teams do not choose between fully human production and fully autonomous AI production. They use AI for concepts, storyboards, visual generation, localization, variations, or editing while people manage strategy, audience insight, storytelling, brand judgment, and final approval. Testing all three approaches would provide a more practical comparison.

What Is the Main Lesson for Marketers From the Ipsos-Syracuse Study?
The main lesson is that production efficiency should not be confused with advertising effectiveness. AI can produce realistic video quickly and can work well for product-led creative, variations, and scalable production. Human judgment remains especially valuable for strategy, emotional storytelling, originality, brand meaning, and final creative decisions. Marketers should test human-made, AI-generated, and human-plus-AI ads using business metrics that match the campaign objective.