AI Images on Allegro from August 2: A Seller's Guide

AI images on Allegro must be declared from August 2, 2026 when they were generated or significantly modified by artificial intelligence and meet the definition of a deepfake. Allegro has announced a visible “AI” watermark and new API fields. Sellers should establish where their images came from and update the upload process before August, but they should not automatically label every photograph edited with software that happens to include AI features.

The key distinction is this: Allegro is implementing a platform declaration mechanism based on the AI Act's transparency obligations, not a blanket label for every automated image adjustment. This guide separates confirmed Allegro requirements from cases that still require an assessment of the specific image.

Information current as of July 15, 2026. Allegro says the aiCoCreatedContent.images field is already present in its API structures but temporarily returns an empty array. The platform will announce separately when the mechanism becomes fully functional.

What Allegro has confirmed

In its July 13 announcement for integrators, Allegro set out five rules:

1. Sellers must disclose whether an offer image was generated or significantly modified by AI when it meets the statutory definition of a deepfake.
2. The declaration for a new image is supplied during its first upload. It cannot later be changed through the API.
3. An image included in aiCoCreatedContent.images will be recorded as created using AI.
4. Allegro will add an “AI” watermark to images marked this way.
5. The mechanism will cover offer management, the product catalog, and the image-upload endpoint.

This creates an operational dependency: when a store receives files from an agency, supplier, or integrator, provenance information must reach the person or system that first uploads the image to Allegro.

Generated, significantly modified, or simply adjusted

The AI Act defines a deepfake as generated or manipulated image, audio, or video content that resembles existing persons, objects, places, entities, or events and would falsely appear authentic or truthful. The European Commission describes the obligation as disclosing the artificial origin of such content.

For e-commerce, classify the outcome rather than the name of the tool:

This is not individual legal advice. Businesses producing synthetic visuals at scale should document their classification rules and obtain specialist advice where the boundary remains unclear.

Allegro and Article 50: two layers of responsibility

Article 50 of the AI Act separates obligations for providers and deployers of AI systems:

• providers of systems that generate text, image, audio, or video must make outputs detectable as artificially generated or manipulated in a machine-readable format;
• professional deployers publishing deepfakes must clearly disclose their artificial origin;
• AI-generated text published to inform the public on matters of public interest has separate rules and an exception linked to human review and editorial responsibility.

The rules apply from August 2, 2026. The EU Code of Practice is a voluntary route for demonstrating compliance, not a replacement for the Regulation. In July 2026, the Commission and AI Board assessed it as an adequate instrument supporting compliance.

For most sellers, the practical layer is Allegro's process: the marketplace defines where the declaration is supplied and how it will be displayed to buyers. A seller does not need to design a separate watermark for every file used only in an Allegro offer.

How the Allegro API declaration works

Allegro has announced two methods:

• for offer resources, the aiCoCreatedContent.images array contains the URLs of images declared as generated or significantly modified by AI;
• for direct upload through POST /sale/images, the declaration uses the isAiCoCreated=true/false parameter or a body field with the same name for the supported request type.

Simplified request fragment:

{
  "images": [
    "https://example.com/packshot.jpg",
    "https://example.com/ai-scene.jpg"
  ],
  "aiCoCreatedContent": {
    "images": [
      {
        "url": "https://example.com/ai-scene.jpg"
      }
    ]
  }
}

The URL in aiCoCreatedContent.images must match an image in the product gallery. When an image already exists in Allegro's database, its original declaration controls and a new request-level declaration is ignored. Integrators should therefore store the classification before upload rather than reconstructing it during a later offer update.

A five-step audit before August 2

Do not begin by manually opening thousands of offers without priorities. Use this sequence:

1. Inventory image sources. Separate your own studio, suppliers, manufacturers, agencies, AI generators, and automated editing tools.
2. Start with new and active offers. First review files entering the system under the new process and offers responsible for the largest share of sales.
3. Require a supplier declaration. Add one field to the brief: “generated or significantly modified by AI: yes/no/unknown”.
4. Keep provenance evidence. Retain the original, tool name, edit date, and approving person.
5. Update the integration. Do not set every unknown file to false. Missing information should enter a review queue.

What not to do

• Do not assume that an AI watermark will reduce an offer's ranking. Allegro has announced no such rule.
• Do not promise that a declaration protects an offer from suspension. Transparency does not replace product accuracy or other marketplace policies.
• Do not label the entire catalog “just in case”. Excessive labels make the information less useful to buyers.
• Do not classify images from file metadata alone. Metadata may disappear during export or optimization.
• Do not postpone the decision until upload. The publisher often does not know how the file was made.

What you can implement today

Add an ai_image_status field to your asset library with yes, no, and review values. Store the source, date, approving person, and link to the original alongside it. For new assets, use review until the creator or supplier confirms the production process.

Ask your integration provider three questions:

1. When will it support aiCoCreatedContent.images and isAiCoCreated?
2. Will it store the declaration during the first upload, as required by Allegro's API logic?
3. What happens to an image without a confirmed status: publication block, review queue, or automatic false?

REL8 designs similar process automations so missing data triggers human review instead of a silent assumption.

How much you can gain

There is no credible universal benchmark showing how sales change after AI-image labeling. You can, however, calculate the cost of a sound process:

new images per month × manual classification time × loaded hourly cost

For example, 600 images per month, 90 seconds of review, and a loaded labor cost of €20 per hour equals about €300 of monthly manual work. If the declaration is captured in the photographer or agency brief, later review can focus only on assets marked review.

The largest benefit is not “automated labeling” but a reliable chain of information from image creator to marketplace offer. It prevents every Allegro update from becoming a new manual investigation.

Need to organize the flow between an agency, PIM, store, and Allegro? Contact REL8. We can map the process and identify where the declaration should be captured before it reaches the API.