How generative AI exposed the creative intelligence gap

Despite all the advancements in media measurement, creative continues to be largely governed by instinct, post-mortem debates and the loudest opinion in the room. When generative AI arrived on the scene, it made this issue even more pressing. Because the technology dramatically expanded the volume of creative that teams could produce, the cost of not measuring advertising became that much higher. 

Consider what’s actually at stake: according to Circana’s research across nearly 450 campaigns, creative accounts for 49% of incremental sales lift — making it the single largest driver of advertising effectiveness. The problem is that most brands today still have no way to know which specific creative decisions inside a winning ad drove that return, let alone how to replicate that success in the future.

Knowing which ad won is not the same as knowing why

When platforms report on ad performance, they say which ad won, but they don’t say why. Was it the way the product was framed that caught consumers’ attention? Was it the beautifully shot lifestyle imagery? Or maybe the messaging hierarchy was particularly effective? Without analysis at that level, all of these insights die with the campaign. Instead of being built on evidence, the next brief starts once again from intuition and guesswork. 

With generative AI, teams that once produced five creatives per campaign are now producing 50. While this can open up more opportunities to learn, it only works if the team has a system in place that extracts what all those ad variants actually reveal. Without structured creative analysis, more output just means more noise. The creative library grows; the understanding of what within it works or doesn’t, does not.  

From gut feel to a data-driven brief

One common mistake teams make is confusing creative performance with media performance. For example, a video that ran in premium inventory to a high-intent audience will naturally outperform the same video served broadly. But if marketers attribute that success to the creative itself, they’ll draw the wrong conclusion and send the next brief in the wrong direction. 

Knowing how an ad performed and knowing what drove that result require two different types of analyses. Most marketing teams are only running the first one, measuring CTR, ROAS and completion rates that describe outcomes at the campaign level. Creative intelligence analysis, on the other hand, identifies which creative elements in the winning ad moved the needle and whether that pattern holds across audiences and markets. That means marketers need to deconstruct each ad into granular variables — such as visuals, colors, layouts, messaging and CTAs — so they can isolate them from factors such as placement, audience and timing. Once marketers figure out which creative decision drives audience engagement and performance, the findings can become truly accurate and actionable. 

The good news is that marketing teams don’t need to figure all of this out on their own, as technology can help them close that gap. Platforms like Celtra use creative intelligence to ingest all past campaigns, automatically identify and structure visual and messaging elements across every ad, then rank those factors by how strongly they influence performance on specific KPIs. For example, the analysis might show that including a lifestyle visual in an ad contributes to an average CTR increase of 12.5%, while highlighting a limited-time offer in the headline might boost it by as much as 25%. The output is a library of structured insights marketers can use as an evidence base when writing their next brief or building their next ad.

The compounding advantage of closing the loop

Most marketing teams today have access to capable generative AI technology, and many have mistaken that access for a competitive advantage. The ability to produce creative assets at lightning speed is no longer a differentiator. What separates effective AI-generated creative from generic output is the context layer underneath it: brand presets, prompt structures and a feedback loop that connects past results to new briefs. 

That infrastructure takes time to build, and the brands investing in it now are compounding an advantage that will be difficult to close later. A brand with two years of structured creative performance analysis behind it holds an advantage that grows with every campaign cycle: pattern confidence increases, briefs get sharper. The AI-generated output becomes more targeted because the inputs are more informed. 

So where do marketers begin? For most organizations, the honest starting point is a question about what creative data currently exists and what is actually being done with it. If campaign results live in a media dashboard that the creative team never sees, the loop is already broken. The brands that build the infrastructure to turn creative performance into creative intelligence now will have an advantage that scales with every dollar they spend.

Connecting creative decisions to business outcomes has always been the hardest part of advertising to operationalize. The volume pressure that generative AI has introduced is the strongest argument the industry has ever had for doing it properly. The question is whether teams use that pressure to build something durable or to simply produce more AI slop.

Partner insights from Celtra