People shown food imagery labeled as AI-generated expressed lower purchase intentions than those shown imagery without that label in a set of three scenario-based experiments, researchers report.
The British Food Journal study connects the response to participants’ perceptions of image authenticity, the effort involved and how the food might taste. The difference was stronger when the food was presented at a higher price premium.
The work measured stated intentions in scenarios, not restaurant orders or sales. It does not establish how every customer responds to an AI label or whether an unlabeled image was assumed to be a photograph.
What an image tells a customer
A food image can communicate more than the appearance of a dish. Viewers may also infer whether a seller prepared the food, arranged a photograph or invested effort in presenting it.
Han Chen, Shuaifang Liu and their co-authors used three experiments to examine how a label identifying AI-generated imagery affected those judgments. The authors found lower reported authenticity and perceived effort, followed by less favorable expectations about taste and purchase intention.
Those linked perceptions are the researchers’ explanation of the pattern in their tests. They are not direct measurements of how the food tasted, nor evidence that an image-generation tool changes the quality of the product.
One experiment compared responses under different price-premium conditions. The negative association with the AI label was stronger in the higher-premium setting, suggesting that customers’ expectations of a premium offer may affect their interpretation.
Honest images and disclosure
The result should not be read as a recommendation to conceal AI use. Advertising rules, platform policies and the risk of misleading customers remain relevant. A food business should check whether its imagery accurately represents what a customer can order and whether a disclosure is required in the market where it advertises.
A generated illustration used to create an atmosphere is different from a product photograph used to show portion size, ingredients or appearance. Where a picture makes a factual claim, the seller needs a reliable way to check that claim.
Our earlier coverage of AI disclosure in emotional video advertising examined a different format and found that the reason for using the technology affected viewers’ reactions.
Testing beyond intention
A restaurant or food brand considering different visuals can compare whether customers understand the offer, trust the image and recognize the product. For a commercial test, it can also measure orders, complaints and returns where applicable, rather than relying entirely on stated purchase intention.
The study appeared in 2026. Its scenario-based design gives researchers a controlled way to compare responses, but behavior at a checkout may differ. The result is a prompt to test both accuracy and customer expectations, not a forecast of the sales effect of every AI-made food image.
