That’s not an accident, and it’s not a lowering of standards. It’s a response to how viewers actually process what they’re looking at in the first half second of a video.

A viewer scrolling through a feed isn’t evaluating production quality. They’re running a much faster, much cruder classification: is this something someone I know made, or is this something a company made? That classification happens before any conscious thought about the content, and it’s based almost entirely on visual texture rather than anything being said.

Polish is one of the strongest signals in that classification. Even lighting, a locked frame, clean color grading: all of it reads as effort, and effort is exactly what gives an ad away as an ad. Once an ad has been classified as an ad, the viewer’s guard goes up before a single word of the pitch has landed, and everything that follows has to work against that initial resistance instead of with it.

Imperfection reads the other way. A slightly wobbly frame, a moment where the subject isn’t perfectly centered, lighting that looks like a bedroom rather than a studio - these signal something a phone captured rather than something a production team built. That signal gets the viewer’s guard down before the pitch has even started, which means the actual message is being received by someone who hasn’t already decided to tune out.

AI generation tools default toward polish. Ask for a person talking to camera and the model’s instinct is even lighting, centered framing, and a clean background, because that’s what most of the training data considered a well-made shot. Left alone, AI UGC drifts toward exactly the visual signature that reads as produced, which is the opposite of what the format needs to work.

This means the imperfection has to be deliberate. It has to be specified in the prompt the same way any other production decision would be, because the model won’t introduce it on its own. Off-center framing, inconsistent lighting, a slight motion blur - these need to be requested directly rather than hoped for, and that’s a meaningful shift from how briefs used to get written when the presenter was a real person who would naturally produce some of that imperfection just by being a person holding a phone.

Deliberate imperfection can be overdone, and when it is, it reads as try-hard rather than authentic, which defeats the entire purpose. A frame that’s slightly off-center feels real. Tilt it hard on purpose, and it starts to feel like someone was told to make it look amateur, and viewers pick up on that distinction even if they can’t articulate why.

The imperfections that work tend to be the ones a real person wouldn’t have noticed or bothered to fix: a background that’s a little cluttered, lighting that’s slightly uneven across the face, framing that’s close but not perfectly centered. The imperfections that don’t work tend to be the ones that call attention to themselves: exaggerated shake, artificially degraded audio, a lens flare placed for effect rather than because it happened. One kind disappears into the format without drawing attention to itself. The other becomes the thing the viewer notices instead of the message.

We treat the imperfection level as its own production variable, specified in the brief the same way angle and entry point are, rather than left to chance or overcorrected in the other direction. Different categories call for different degrees of it. A skincare testimonial can carry more visual roughness than a financial product, where too much imperfection starts to undercut trust rather than build it.

We also test it deliberately rather than assuming more roughness always wins. A batch will sometimes include a version with heavier imperfection and a version with a cleaner look, specifically to check whether the category and audience are actually responding to the signal the way the broader trend suggests they should. The trend is real, but it isn’t universal, and the only way to know where a specific product sits on that spectrum is to test it rather than assume it.

If you want to look at whether your current creative is over-polished for the platforms it’s running on, or overcorrected in the other direction, book a call with us and we’ll get into it.

Until next time,

Miro and the Inceptly Team