Can AI-generated food really get the tastebuds tingling?
Olaf van Gerwen, Founder & Global Creative Director at Chuck Studios, finds that AI-generated food may be everywhere, but our finely tuned appetite for the real thing means even the smallest synthetic slip-up can leave a bad taste.
)
This summer I went to Notting Hill Carnival for the first time.
Among the feathers, the sequins and bass lines I felt in my fillings, I noticed something else: AI-generated menu boards. Just, everywhere.
The chicken had suspiciously smooth skin. The steak had the texture of a vintage leather armchair. The sauces sat on the food the way no liquid would ever sit on anything. It all looked like food, and most of it made me a little less hungry. Luckily, the smell more than made up for it.
Now, writing from Japan, I’m seeing much the same thing. Apparently, AI food travels well.
A slightly off-looking bun, cheese that stretches like plastic or strangely textured meat doesn’t just make an ad worse. It plants the wrong sensory memory of the product, and therefore of the brand.
Welcome to food’s uncanny valley. The term usually describes AI-generated people: a face that’s almost convincing, until some tiny wrongness in the eyes, skin or expression makes the hair on your neck stand up. Food now has its own version. I’d argue it will be one of the hardest categories for generative AI to crack.
Above: An AI-generated menu board spotted by van Gerwen at Notting Hill Carnival, where even the promise of hot food came with some distinctly unnatural-looking textures.
We know food better than we think we do
People have an incredibly sophisticated visual understanding of food, even though most of us couldn’t describe it to save our lives. As a species, we’ve spent millions of years working out what’s edible and what’s poisonous. The ones who got it wrong didn’t get a vote. So we know how lettuce should break. We know meat has texture, and isn’t just brown. We know how a droplet trickles down a cold glass.
As a species, we’ve spent millions of years working out what’s edible and what’s poisonous.
Most people could never explain any of this. They’d simply recognise it. Show someone an AI steak and they’ll just think: “Eh. That looks weird.” Then they’ll scroll on, and your media budget goes with them.
Hunger makes raw beans taste sweet
Back to Carnival, where a day of kidney-stone-crushing bass had left us starving. We were dangling at the very bottom of the funnel, and down there, none of this matters much. There’s a Dutch expression I love: honger maakt rauwe bonen zoet. Hunger makes raw beans taste sweet.
We headed for the best-smelling food stall, and there the horrendous AI imagery did its job. It didn’t need to create desire, because the smell had already done that. It had one job left: help us choose exactly what to order.
Big food brands, though, don’t have the luxury of a starving customer with money in hand.
Above: AI-generated food imagery spotted by van Gerwen in Japan, where impossibly glossy meat, suspiciously sculpted tempura and unnervingly perfect egg yolks suggest that, as he puts it, “AI food travels well.”
The real job is creating a memory
Big food brands talk to people long before they’re three metres from a grill. Their job is to create desire early and leave something behind in the memory. We are, after all, in the memory-building industry.
Food imagery can help build that memory. We want to show a particular kind of crispness, creaminess, juiciness, freshness or indulgence, and repeat it again and again. Over time, those cues become the brand’s Culinary Identity: a visual shortcut to the brand. Then, when hunger does arrive, the brand already stands for something and is more likely to come to mind in the first place. Byron Sharp calls this mental availability.
That’s where the “almost right” of so many AI images becomes a huge problem. A slightly off-looking bun, cheese that stretches like plastic or strangely textured meat doesn’t just make an ad worse. It plants the wrong sensory memory of the product, and therefore of the brand. That memory is very hard to undo.
Appetite lives in the details
This is why food photographers, stylists and directors have spent decades obsessing over details that may look ridiculous from the outside. I’ve personally lost many hours of my life to single drips of sauce. Still, je ne regrette rien. It taught me how to design, light and curate food images like a pro, and it’s why I believe a cheaper or quicker tool is never an excuse to care less about the output.
A slightly off-looking bun, cheese that stretches like plastic or strangely textured meat doesn’t just make an ad worse. It plants the wrong sensory memory of the product, and therefore of the brand.
To be fair to the robots, bad food imagery is nothing new. We’ve all been wincing at dreadful restaurant billboards for decades. Slop isn’t new; it just looks different now. AI has given us a faster way to make it, at scale.
Don’t get me wrong: I’m not against GenAI. As a maker, I find it thrilling. The real magic will start the moment we humans can no longer tell a captured image from a generated one. At that point, “you don’t know what you don’t know” becomes profoundly true.
But when your product is shown countless times across dozens of markets, “roughly chicken-shaped” isn’t good enough. It’s a liability. One synthetic-looking drumstick, multiplied by a global media plan, is a very expensive way to teach people your food isn’t real.
Global brands can’t wait for hunger to do their work for them. They need to be remembered long before anyone is starving in Notting Hill.
