In Q4 2025, Alphabet's SVP of Product Finance and Operations reported that Google advertisers used Gemini to generate nearly 70 million creative assets via text customization in AI Max and Performance Max. That number belongs in the same sentence as industrial manufacturing statistics — not marketing ones. It represents a production infrastructure operating at a scale that no human workforce, at any price, could replicate.
The efficiency case is real. Aritzia saw an 80% incremental uplift in conversion value using AI Max. L'Oréal ran AI Max across 800 unique campaigns in 23 countries and 30 brands, increasing DTC brand revenue by 23%. Video ad creation costs dropped 85–95% versus traditional production. Time-to-launch compressed from two to three weeks to under two days.
These numbers are not projections. They are Q4 2025 results. The rendering floor is running.
The IAB's January 2026 study, conducted with Sonata Insights across 505 Gen Z and Millennial consumers and 104 ad industry executives, found something that should concern every brand spending on AI creative at scale. Executives believe 82% of Gen Z consumers feel positive about AI-generated ads. The actual figure: 45%. That 37-point gap widened from 32 points in 2024.
The specific negative sentiment is where it gets structurally interesting. Gen Z negative sentiment toward AI ads nearly doubled year-over-year — from 21% in 2024 to 39% in early 2026. Millennials tracked at 20%. The generation that grew up with algorithmic everything is not warming to it. They are withdrawing from it.
David Cohen, IAB CEO: "The digital industry is embracing AI in all of its splendor at breathtaking speed… We must get transparency and disclosure right, or we risk losing the trust that underpins the entire value exchange."
Note what Cohen did not say: that the creative itself needs to be better. The industry has identified the problem as a transparency issue. The research suggests it is something deeper — a structural quality problem that cannot be disclosed away.
Svedka premiered the first primarily AI-generated Super Bowl ad in February 2026. The spot featured the brand's Fembot mascot and a new companion, "Brobot," dancing to a techno remix of Rick James' "Superfreak." The agency was Silverside AI — the same company behind Coca-Cola's controversial AI work. A human choreographer was hired via TikTok competition. The tagline claimed human connection. The comments called it "soulless" and "nightmare fuel."
McDonald's Netherlands released an AI-generated holiday ad on December 6, 2025. By December 9, it was gone. The 45-second spot depicted holiday mishaps — exploding Christmas trees, falling ice skaters, a cargo bike with a sliding tree — suggesting McDonald's as relief from seasonal chaos. The production took seven weeks, according to The Sweetshop CEO, who noted the team "hardly slept" and created thousands of takes. Social media described it as "the most god-awful ad I've seen this year," "creepy," and "poorly edited." The McDonald's statement: "This moment serves as an important learning as we explore the effective use of AI."
Coca-Cola has run AI holiday campaigns for two consecutive years. The 2024 "CreateRealMagic" campaign required 70,000+ prompts and 30 days of AI-generated footage. Marvel co-director Joe Russo called the result an "uncanny valley nightmare." Alex Hirsch, creator of Gravity Falls, described it as "Coca-Cola is 'red' because it's made from the blood of out-of-work artists." The company's own slogan is "It's Always the Real Thing."
In 2025, Coca-Cola reduced the human production crew from 50 to 20. Pratik Thakar, VP of Generative AI at Coca-Cola, stated that "craftsmanship is 10 times better this year." The public response from the same audience was: "Your boss firing you on Christmas."
The Stanford NeurIPS 2025 Best Paper, "Artificial Hivemind," provides the theoretical underpinning that explains why every brand is arriving at the same destination independently. Testing 70+ language models across 26,000 real-world queries, researchers found that models don't just repeat themselves — they repeat each other. Shared training data and optimization for statistically likely outputs create what the paper calls a "generative monoculture." All major models gravitate toward identical outputs because they are drawing from the same finite internet corpus and optimizing toward the same probability distributions.
This is not a quality defect. It is a structural feature. The efficiency that makes these tools useful is the same property that eliminates differentiation.
The Varadarajan et al. 2025 study found AI-generated text exhibits remarkably limited variation in inferrable psychological traits compared to human-authored text. AI models the "average human" with so little variation that human traits can distinguish AI from human text using unsupervised methods.
The Doshi and Hauser experiment in Science Advances (2024) captured the paradox precisely: AI-enhanced stories are rated more creative individually but are more similar to each other than human-only stories. The individual writer benefits; the collective output narrows. Each brand, acting rationally in its own interest, is contributing to an ecosystem-level distinctiveness deficit.
The attention economy is beginning to sort into two tiers. The commodity tier: AI-generated, optimized, scalable, forgettable. The premium tier: human-authored, distinctive, expensive, irreplaceable. Harvard Kennedy School research from 2025 found 79% of readers actively prefer human-written content in blind tests. Nielsen Norman Group found readers spend 40% more time on first-person narrative passages — the mode AI content defaults away from, because AI is structurally better at informational third-person synthesis than experiential first-person narration.
The market is registering this. Sixty-one percent of marketers plan to increase investment in creator content in 2026. The creator economy — individual humans with identifiable perspective, taste, and voice — is gaining relative value precisely as AI makes generic content abundant. The mechanism is simple: when supply of something increases faster than demand, unit value falls. AI creative is flooding the market. Distinctive human creative is not.
McKinsey's analysis of creative homogenization frames it as a structural risk when widely distributed templates saturate categories. The firm notes that AI produces executional variety easily but cannot produce strategic variety — that still requires human judgment about category positioning, cultural moment, and brand-specific differentiation. The brands gaining advantage are those building differentiation systems around AI tools, not those relying on AI to provide differentiation itself.
Here is what the rendering floor does not produce, by design: measurable audience withdrawal. The CTV bot fraud study from DoubleVerify found something analogous — the ShadowBot operation was detected because the fake users forgot to update their display settings from 800×600 resolution, a carryover from 1990s CRT monitors. The bots had everything except the right answer to one question about their own hardware. The rendering floor's audience withdrawal is similarly visible once you know to look for it.
Willingness to engage with AI content dropped from 60% in 2023 to 26% in 2025. Fifty-two percent of consumers actively reduce engagement when they suspect AI content. The gap between what marketers believe AI content delivers — 77% say AI produces emotionally resonant content — and what consumers experience — 33% agree — is not a measurement problem. It is a product problem.
The rendering floor produces content at industrial velocity. It is efficient by every internal metric. The cost it does not account for is the slow withdrawal of attention from content that feels like it was rendered rather than made — a category so large and so pervasive that no single impression-level metric captures the aggregate effect. The brand that figures out how to stand out from the render will find an audience that has been waiting.