Two years of argument over whether generative AI can produce good creative work has been aimed at the wrong target, according to research published on 14 July by WARC and LIONS Advisory in partnership with TikTok. Across 400 marketers in the UK, United States, Australia and Brazil, 88% reported higher creative volume since adopting the technology and 45% reported a significant improvement in quality.

The report locates that shortfall upstream of the models, in the briefs marketers write. That reframing lands while marketing budgets are still argued over in terms of which system to buy. It also lands days before AI-assisted advertising meets new disclosure rules, as transparency obligations under Article 50 of the EU AI Act become applicable on 2 August.

What did the WARC and TikTok survey measure?

WARC surveyed 400 marketers in May 2026, all of them involved in decisions about how creative and content are produced, and supplemented the survey with interviews with senior marketers and a review of WARC and TikTok global data. The report, titled The new creative advantage: How community intelligence is reshaping creative success in the age of AI, was published under the LIONS Advisory banner. WARC is part of LIONS.

Adoption is no longer the variable the study is testing. According to the survey, 90% agree generative AI has become a key tool in creative development and 72% use it frequently as a regular part of content production. A further 87% believe their organisation is using it effectively.

Most of that usage sits at the production end. Visual, image or video production leads at 73%, followed by content personalisation at 64% and concept development at 61%. Ideation and brainstorming registers 57%, audience targeting and segmentation 53%, and testing and optimisation 41%.

Why does the report blame the brief rather than the model?

WARC found a direct contradiction between what marketers believe about audience data and what they actually feed an AI tool. Demographic data remains the primary input used to brief generative AI, relied on by 67% of respondents, while 59% agree that traditional demographic segmentation is no longer effective. Only 17% make it a rule to always bring community or audience insight beyond demographics into those workflows.

Marketers were also asked what is missing. Behavioural data covering how audiences act, not who they are, was named by 45%, high-quality brand and creative guidelines by 40%, and real-time cultural signals by 35%. Historical performance data detailed enough to be useful was cited by 34%. Only 10% said they saw no significant gaps in the inputs available to them.

The limitations respondents attribute to AI output track those gaps closely. Over-reliance on generic or familiar styles was the most cited at 40%, unpredictable and hard-to-control quality at 36%, and a lack of creative distinctiveness at 32%. Andy Yang, global head of creative and brand ads at TikTok, wrote in the report’s foreword that the divide opening up is “not a technology gap, it is an intelligence gap”.

That ordering points to a supply problem. Marketers named the missing inputs precisely and ranked them; what they lack is organisational access to behavioural data and usable brand guidelines, not an understanding of why demographics are thin.

How much weight should a TikTok-commissioned study carry?

TikTok commissioned the research, and the remedy the report recommends routes through exactly the participatory platform signals TikTok generates and holds. The report states this, and it leaves a question unresolved that matters commercially: whether an advantage built on community signals is portable between platforms, or accrues mainly to the environment that captures them.

One figure invites particular caution. The 87% who believe their organisation uses generative AI effectively is a self-reported measure from a vendor-commissioned survey, and it sits awkwardly beside independent measurement. A MiQ survey from November 2025 found 72% of marketers planning to grow AI adoption while only 45% felt confident. The two ask different questions of different samples, but they describe the same industry.

The input thesis itself has support from outside the TikTok orbit. Brandwatch research published in April 2026 identified social listening and consumer research as a top marketer skill gap at 55%. CreativeX estimates, cited in the WARC report, that half of media budgets still go behind advertising not built for the platforms it runs on. Kantar data in the same report finds that customising content for context can raise a campaign’s contribution to brand equity by 57%.

Does AI-made creative actually perform worse?

The report does not argue that it does. It cites System1 testing in which AI-generated advertising scored above the global advertising average, and a large-scale Columbia Business School analysis published in 2026 finding AI-generated display ads competitive in market. A Taboola study across a publisher network of roughly 600 million daily active users found AI-generated creative maintained or increased click-through rates without damaging conversion.

The penalty attaches to detection rather than origin. Research from Raptive found that content readers suspect of being AI-generated cuts reader trust by 50% and degrades brand advertisement performance by 14%. On that evidence the operative distinction is not human against machine, but creative that reads as native against creative that reads as machine-made, which returns the question to the specificity of what went into the brief.

The European Commission published its guidelines and finalised Code of Practice on AI transparency on 20 July, and the narrower disclosure carve-out for artistic and satirical work does not extend to advertising. Article 50 becomes applicable on 2 August.