AI didn't kill the creative. It killed the average.
AI makes competent work free. That makes competent work worthless, and the only work still worth paying for is the work a model cannot average out.
Every brand now has the same tireless junior creative on staff. It writes clean copy, renders clean visuals and never misses a deadline. The thesis: that is terrible news for any brand whose plan was to be good enough. AI did not replace creative work. It flooded the middle of it, and the middle is now worth close to nothing.
AI lifts the floor, not the ceiling
The best evidence comes from a controlled experiment in Science Advances. Anil Doshi of UCL and Oliver Hauser of the University of Exeter had 293 people write short stories, some with GPT-4 story ideas on tap. Access to up to five AI ideas raised the novelty of the stories by 8.1% and their usefulness by 9.0%. Most of the gain went to the least creative writers, whose stories were rated up to 26.6% better written. AI pulled them level with the naturally creative ones. It was a lab study of short stories, not ads, but the pattern is the one that matters here.
Now the part that matters for marketers. The researchers found no evidence that AI pushed the best writers any higher. And the AI assisted stories were more similar to each other than stories written alone. The authors call it a social dilemma: each writer is better off, while the collective output gets narrower.
Translate that to a category. If every competitor uses the same models with similar prompts, everyone's work improves and converges at the same moment. The average rises, and it becomes the most crowded place to stand.
Average was already expensive
Dull work was a problem long before ChatGPT. System1's Extraordinary Cost of Dull, built on its Test Your Ad database, found the most common response to a TV ad is no emotion at all. Neutral is 52% of responses to UK TV ads and 47% in the US. For B2B ads it rises to 60% and 54%.
System1 then modelled what that costs. A brand with 5% share and ten points of excess share of voice, airing its least dull ads, could expect about 1.3 points of share gain a year in the US. The same spend behind the dullest quartile buys about 0.1 points. Closing that gap would take an estimated $109bn of extra US media money. Dull does not damage a brand. It quietly burns the budget.
AI does not cause dullness. It makes dullness cheap to produce at scale, which means far more of it will be competing for the same attention.
The tell is in the branding
Kantar analysed hundreds of ads made with generative AI and found no built in penalty or bonus. Where AI was used seamlessly, the ads did well, with over 40% landing in the top tier for branded cut through. Where it was obvious, they did worse. AI ads also had weaker branding on average, mostly because the models were never taught the brand's assets or tone. Kantar puts it plainly: AI does not have your brand at heart unless you put it there.
Audiences notice more than marketers think. The same Kantar work cites its 2024 Media Reactions study: 41% of consumers say AI generated ads bother them, against 29% of marketers.
Where AI earns its keep
None of this is an argument against using it. The Doshi and Hauser data shows AI is a superb way past the blank page; stories written with more AI ideas had more plot twists and were rated less boring. Used well, it is the fastest research assistant, storyboard artist and versioning engine a team has ever had.
The trick is to point it at the volume work and keep it away from the decision. Let it generate fifty routes so a person can kill forty nine. Let it resize, translate and version the winning idea into every format the media plan needs. Do not let it pick the idea, because the idea it picks will sit closest to the middle of everything it has read.
What changes for brands
Three things follow. First, the bar for good enough just moved, and it moved for every competitor on the same day. Work that would have passed a year ago now looks like everyone else's.
Second, the money is in the tails. McKinsey's design research across 300 listed companies found top quartile design performers grew revenue 32 percentage points faster than peers over five years, while the differences between the second, third and fourth quartiles were marginal. The study shows correlation rather than proof of cause, but the shape is hard to ignore: markets pay for standing out, not for being fine.
Third, the scarce input is no longer production. It is the brief, the insight, and the nerve to approve something that does not look like the category.
So the useful question about any campaign is not whether it used AI. It is whether a competitor with the same tools could have made it. If the answer is yes, it is average, and average is now free.
FAQ
Does AI make advertising worse?
Not by itself. Kantar found ads made with AI land anywhere on the effectiveness scale. Seamless use performed well; obvious AI and weak branding performed worse.
Why does AI make creative work look the same?
In a Science Advances experiment, stories written with AI ideas were more similar to each other than stories written alone, because writers anchored on the AI's suggestions.
How much does dull advertising cost?
System1 estimates US brands airing the dullest quartile of TV ads would need about $109bn of extra media spend to match the share gains of non dull ads.
Where should marketers use AI?
On volume work: generating options, versioning, resizing and translation. Keep the choice of idea and the judgment about the brand with people.
Related reading and work
Cliff
Writes The Monthly Ramble on AI, creativity, brand and social.
