We use AI daily. We charge for it. We also know exactly where it breaks. Founders who ignore the breaks ship bad work and rarely know why.
Six things AI imagery still gets wrong in beauty
Tactile finish is the first failure point. AI does not understand the difference between soft-touch matte and rubberised matte. The first looks premium, the second looks Halloween. Production designers know this difference intimately, and AI tools have to be told it every single brief.
Typography in render is the second failure point. Logos on bottles consistently break. Lettering hallucinates. Trademark integrity is non-negotiable, so every AI packshot needs a manual paste-in of the wordmark, by a human who knows the brand.
Colour accuracy at point of sale is the third failure point. The peach you brief is rarely the peach you get. Then add the printer's bias and the shelf lighting. Three layers of drift between brief and shelf, and a human still has to sign off the Pantone.
Claim accuracy is the fourth failure point. Generated ingredient lists are usually wrong. Generated efficacy visuals can imply claims you do not legally own. This is a compliance landmine, and it gets worse as the imagery gets more photoreal.
Restraint is the fifth failure point. AI tools are trained to produce more. Beauty design wins on less. Reducing the AI output is its own skill, and the studios that master it are the ones whose work feels effortless rather than overdone.
Soul is the sixth failure point. Soul is what happens when a creative director makes a choice that the data says is wrong. AI does not make that choice, yet. The brands that still hire humans for the choice that matters end up with work that lasts.
What this means for buyers
When you brief a studio that uses AI, ask what they do not use it for. If the answer is everything is AI, walk away. If the answer includes typography, claim copy, finish specification and final art direction, you have found a studio that knows the medium.