AI images in food and drink stores: serving suggestions, food styling, labels

In a food store, the image sells the taste: the chocolate cut open, the muesli with berries, the juice in a misted glass. Such images are made as photos or with AI, for example as an AI serving suggestion, as an AI scene around the real pack or as a complete AI packshot. Whether such an image needs an AI label is decided by Article 50(4) of the EU AI Act (Regulation (EU) 2024/1689); whether it misleads about the food is decided, independently, by the EU Food Information Regulation (FIC Regulation, Regulation (EU) No 1169/2011). This guide classifies typical food images along the European Commission’s guidelines of 20 July 2026 (C(2026) 5054 final) and flags its own interpretations. Cut-outs and retouching in general are covered in the guide AI background, cut-out, retouching. The transparency obligations under Article 50 in general are covered in a separate guide.

What do the AI Act and the FIC Regulation check?

The AI Act asks whether an image must be disclosed. Article 50(4) addresses deployers of an AI system that generates or manipulates deepfakes; the machine-readable marking under Article 50(2) is owed by the provider of the tool. Under Article 3(60), a deepfake is AI-generated or manipulated image content that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful. The guidelines count consumer goods among objects, animals among entities, and the actual appearance of objects and the actual behaviour of humans and animals as part of authenticity (para. 113); the assessment is objective (para. 114). The Digital Omnibus (Regulation (EU) 2026/1744) only changed Article 50(7).

The FIC Regulation asks whether the information is correct. Under Article 7(1), food information must not be misleading, particularly as to the nature, identity, properties, composition, quantity, durability, country of origin or place of provenance, method of manufacture or production (point (a)), and under Article 7(4) this applies to advertising and presentation too, including the way foods are arranged and the setting in which they are displayed. In the interpretation of this guide, a product image in a store is food information (Article 2(2)(a) mentions “modern technology tools”), and in any case advertising, however it was made.

An AI label only answers the first question: Article 50(6) AI Act leaves other transparency obligations unaffected, and the disclosure does not permit misleading advertising (para. 129). The guidelines, the Code of Practice and the icon page do not deal with food specifically (as of 1 October 2026).

Does the AI show the food or only the scene?

For food images, one cross-check helps (interpretation of this guide): what did the AI draw, the food or what surrounds it?

• The scene: table, light, board, cutlery, background. For advertising and packaging, para. 116 names backgrounds for clearly aesthetic purposes as usually of minor impact, and the guidelines’ negative example shows a real product against an AI environment, as long as the ad does not mislead about the product’s representation, characteristics and use.
• The food: contents, cross-section, filling, portion, the prepared dish. If the AI generates or changes these, it determines how the product looks. The guidelines’ positive example: an AI-generated image of a product in advertising or on packaging that makes it appear, for example, more appealing or of higher quality than in real life.

In a serving suggestion, both are often in the same image: the real pack in front, the finished dish next to it. In the interpretation of this guide, the dish is not decoration but the product after preparation.

Which food images need a label?

You label photorealistic borderline cases, as the German Wettbewerbszentrale also recommends (guide version 2.0, as of 29 July 2026).

Where does food styling tip over?

For food, the typical tipping points match the characteristics in Article 7(1)(a) of the FIC Regulation (mapping is the interpretation of this guide):

• Quantity and portion: the AI fills the glass higher, puts more pieces in the bowl or makes the pack look fuller than it is.
• Composition: more fruit, nuts or meat in the cross-section than the recipe provides. For the case where a usual ingredient has been substituted, Article 7(1)(d) expressly names pictorial representations.
• Appearance and condition: crispier, juicier, fresher than the product turns out when prepared as instructed. Para. 116 covers colour correction, not a freshness the product does not have.
• Origin and production: pasture, mountain farm or handcraft in the image that do not match the place of provenance and the method of manufacture or production.

If one of these applies, the image is a deepfake under the positive example and, above all, information that can mislead. No label fixes that, only a different image. A related case is covered in the guide Labelling AI images in a cosmetics and beauty store.

Is “serving suggestion” enough?

The term does not appear in the FIC Regulation, neither in the original version nor in the consolidated version of 1 April 2025. The yardstick remains Article 7, with the way foods are arranged and the setting in which they are displayed (Article 7(4)(b)). Whether the phrase removes a false impression in a given case is not answered by the Regulation; this guide does not evaluate case law on it. What is clear: “serving suggestion” is not an AI disclosure. If the AI generated the dish, the image needs the AI label as well, and the two together do not make an image permissible that shows more or better than the pack delivers.

In practice (interpretation of this guide): show the pack contents as a real photo next to the serving suggestion, and name side dishes that are not included in the text.

What applies to the pack label and mandatory particulars in the image?

For prepacked food sold at a distance, the mandatory particulars must be available before the purchase is concluded under Article 14(1)(a) of the FIC Regulation, except the best-before or use-by date, and all of them at delivery (point (b)). Under Article 9(1) they include, among others, the name of the food, the list of ingredients, allergens, the net quantity and the nutrition declaration, and all of it must be accurate (Article 7(2)). For AI images this means (interpretation of this guide):

• AI packshot: if the AI generates the pack, it also generates the name, ingredients, nutrition values and seals. If they differ from the real pack label, the image is wrong, with or without an AI label.
• Photo of the pack label as mandatory information: if you show ingredients or nutrition values as an image of the back of the pack, use the real photo; a tool that redraws lettering has no place on the pack label. The safest way is to add the particulars as text.
• Third-party pack with an AI motif: if you photograph a manufacturer’s pack that carries a printed AI motif, you do not use an AI system yourself; your photo is real.

AI MODIFIED or AI GENERATED?

AI MODIFIED is the regular case as soon as a real product or pack photo is in the image: AI serving suggestion next to the real pack, food styling on a real photo, AI flavour variant. AI GENERATED belongs on images with nothing real in them, such as a packshot, dish or farm scene from a prompt; if the origin is unclear, choose AI GENERATED, the stronger disclosure. More in the guide The EU icons for AI images.

What applies in Germany, Austria and Switzerland?

As an EU regulation, the FIC Regulation applies directly in every Member State (Article 55).

Germany: Section 11(1) of the German Food and Feed Code (LFGB) prohibits the operator responsible under Article 8(1) of the FIC Regulation from placing food on the market, or advertising it, with food information that does not comply with Article 7(1), also in conjunction with Article 7(4).

Austria: Section 5(2) of the Austrian Food Safety and Consumer Protection Act (LMSVG) prohibits placing on the market or advertising food with information liable to mislead, with largely the same characteristics as Article 7(1) of the FIC Regulation; under Section 5(4) this applies accordingly to presentation. The AI Act applies directly in Austria. More in the guide Labelling AI images in Austria.

Switzerland: Swiss food law applies to the Swiss market. Under Article 18 of the Food Act (LMG), presentation, labelling, packaging and advertising must not deceive; Article 12 of the Ordinance on Foodstuffs and Utility Articles (LGV) expressly names images and prohibits presentations suggesting a value above the food’s actual condition (paragraph 2(e)). The AI Act only reaches a Swiss store if the image is used in the EU (Article 2(1)(c) AI Act). More in the guide Labelling AI images as a Swiss shop.

How to proceed

1. Sort images: pack, contents, serving suggestion, scene, origin, pack label. Ask manufacturers and agencies where AI was involved. For AI images from the manufacturer that you use unchanged, you are usually not the deployer in the interpretation of the guide AI stock images in your online store, but you remain responsible for a misleading depiction.
2. Check against the goods and the pack label: quantity, ingredients, appearance after preparation, origin, mandatory particulars. If the image differs, change the image.
3. Check the file: a provenance marker (C2PA or IPTC) is evidence of AI origin; a missing marker is no statement. Tools made specifically for food styling have not been measured on the generator test bench. Check your originals in the free AI image checker, not the store copy: on the copies Shopify delivers to visitors, only 6 of 38 marked images still carried the marker on 30 September 2026.
4. Label on the image, visible without a click wherever the image appears (Article 50(5), para. 143), with a text alternative. An app such as Easy AI Act Image Labels reads markers on the original and suggests labels; you label images without a marker yourself with one click, an AI serving suggestion next to the real pack usually with AI MODIFIED.
5. For ads, marketplaces and social media, a copy with the label burned in, for example with the free AI Label Generator. What the platforms require is covered in the guide on AI images on marketplaces and in ads.
6. Keep originals and decisions: the file from the tool in its native format, plus source and reasoning per image.