The reflex where audiences read high-polish work as possibly fake, which quietly inverts the oldest assumption in advertising: that better production means more credibility.
Also known as authenticity backlash, AI suspicion, polish penalty
Synthetic skepticism is the default suspicion audiences now bring to content that looks expensive. For most of the industry’s history, production value functioned as proof of resources and therefore proof of legitimacy: a real company made this, so a real company stands behind it. Generative tools broke that link by making the polished version cheap, so polish stopped being evidence of anything.
The observable behaviour is a search for verification signals instead. Visible human involvement, imperfection that would be expensive to fake, named people, dated and specific claims, provenance metadata. Standards work like C2PA and Content Credentials is one institutional response, but the consumer response arrived first and is not waiting for a standard.
It attacks the thing agencies sell. If craft can read as a liability, the pricing logic of an entire department is in question.
The polish dial now has a wrong end. Some categories still reward gloss and some now punish it, and the split does not follow the usual premium versus mass line. Testimonial, founder and community content in particular can lose credibility the more it is finished. This has to be a deliberate decision per asset, not a house style.
Proof of humanity becomes a creative device. Named people, on-camera specificity, dates, locations, things a model would have no reason to invent. This is not a return to raw UGC aesthetics, because those are now trivially generated too. It is closer to evidence than to style, and it changes what you shoot.
Disclosure is a brand decision, not just a legal one. The EU AI Act’s synthetic content marking obligations applied from August 2026, and platforms have their own labelling rules. Deciding how a brand discloses AI use, ahead of being forced to, is now a positioning question that lands on the agency’s desk whether or not anyone asked for it.
A B2B software client approves a testimonial film: three customers, studio lighting, colour grade, the version the agency would put on its own reel. It tests poorly in a way nobody expects. In the open-ended responses, roughly a third of viewers question whether the customers are real people, and two mention that one speaker’s delivery feels generated. The reshoot is not cheaper, it is differently expensive: real offices with bad acoustics, full names and job titles on screen, unedited specifics about implementation problems, and a fourth customer who says something mildly unflattering. The client’s legal team spends longer on it than the original. It outperforms the polished cut, and the agency has to explain to its own creative directors why the worse-looking film is the better one.
The automations and agents module of the workshop teaches you how to build AI workflows that compress the busywork without taking the craft out of the studio.