AI transparency
Last updated: 21 October 2026
Feedova produces marketing content with artificial intelligence. This page explains what is generated, how it is flagged, who is responsible for publication, and how to label your content. It supplements the Terms of Use.
1. Feedova uses artificial intelligence
Feedova produces images, videos, voices, avatars, text and analyses with artificial intelligence models supplied by third parties (the list of subprocessors names them), based on the information that brands provide and on what their market says.
What is generated. Every image, video, voice, avatar and text produced by Feedova is generated or modified by artificial intelligence, even when it starts from a real photo or a real video of the brand. The people in the avatars offered by default are fictional: they are not real customers.
2. How it is flagged
Files exported from Feedova will carry machine-readable metadata indicating that they are generated or modified by artificial intelligence (the C2PA standard, known as "Content Credentials", or an equivalent). Some models also add an invisible watermark. We undertake to have this marking in place as soon as the paid plans open, and no later than 2 December 2026. No content is generated for a customer brand before these plans open.
Do not remove these markings when you edit or recompress your files.
Reviews, testimonials and staged people. Feedova is set up never to invent a review, a rating, a first name attached to a quote or a customer result figure, and an automatic check reads ad copies and the text of static creatives before returning them to you. When a generated or acted person speaks as a customer, the text is marked "to disclose: AI-generated content / staged" and the texts screen shows a notice: it is up to you to tell the public where you publish (Terms of Use, section 7.7). These checks are not exhaustive.
3. Who is responsible for publication
The brand that publishes a piece of content is its publisher and advertiser. The brand reviews each piece of content, verifies its claims, and informs the public that content is generated by artificial intelligence when the law or the platform requires it (section 5).
4. When you talk to Feedova
In the application, in emails and in the chat, Feedova writes to you in the first person, like a colleague. These messages are produced automatically, by rules or by artificial intelligence: they are not written by a person. When you ask to speak to a human, a member of our team replies to you.
5. Labeling guide for brands
This guide summarizes the rules known as of 21 October 2026. It does not replace legal advice in your country.
5.1 The essentials
- Never present a person generated by artificial intelligence as a real customer, expert or employee. This is the safest rule everywhere. In the United States, the Federal Trade Commission considers such fake testimonials deceptive.
- In the European Union, an image, a video or a sound generated by artificial intelligence that could pass for real must be disclosed (Regulation (EU) 2024/1689 (AI Act), Article 50(4), applicable since 2 August 2026). For a photorealistic advertisement featuring a generated person, assume that it must be disclosed.
- In France, specific labels exist: « Photographie retouchée » ("retouched photograph") for commercial photographs of models whose body shape has been altered, and « Images virtuelles » ("virtual images") and « Images retouchées » ("retouched images") in influencer content.
- Meta may label your ads as created or modified with artificial intelligence, and requires a disclosure for certain ads.
- Keep the markings in your files.
5.2 Rules by context
| Where, when | Rule | Possible label |
|---|---|---|
| European Union, image, audio or video content generated or modified by AI that resembles existing persons, objects, places or events and could appear authentic | whoever distributes it discloses it clearly, at the latest at the first contact with the public; for an evidently artistic, creative, satirical or fictional work, the disclosure may be discreet but is still required | "AI-generated image", "AI-generated video", "AI-generated voice" |
| European Union, text published to inform the public on matters of public interest | must be disclosed, unless it has undergone human review under editorial responsibility; ordinary advertising copy is in principle not covered | "AI-generated text" |
| France, commercial photograph of a model whose body shape has been made slimmer or fuller | label that is legible and clearly distinguished from the message, under the advertiser's responsibility | « Photographie retouchée » ("retouched photograph") |
| France, content distributed by an influencer | face or body shape produced by AI, or retouched: label visible for the entire duration of the content | « Images virtuelles » ("virtual images"), « Images retouchées » ("retouched images") |
| State of New York, advertisement that shows a synthetic performer created by AI | visible disclosure by whoever produces or creates the advertisement | "This ad features an AI-generated synthetic performer" |
| United States, any advertising | no fake reviews, fake testimonials or fake endorsements; a testimonial reflects the actual experience of a real person | do not present an avatar as a customer; otherwise "Fictional person generated by AI" |
| European Union, reviews and testimonials | banned in all circumstances: fake consumer reviews or endorsements, distorted reviews, falsely presenting oneself as a consumer (Directive 2005/29/EC, Annex I, points 22, 23b and 23c) | quote only real reviews, unchanged; an acted or generated person is disclosed as such |
| United Kingdom, reviews and testimonials | banned in all circumstances: a review that purports to be based on genuine experience but is not, created by any means, software included (DMCC Act 2024, Schedule 20, para. 13) | same rule |
| Meta, all ads | Meta may display an "AI info" label; the rules on health and appearance prohibit idealized before-and-after images and images that create a negative self-perception | no label required outside political and social issue ads |
| Meta, ads about social issues, elections or politics | mandatory disclosure of any realistic image, video or audio that has been digitally created or altered | disclosure checkbox in Ads Manager |
5.3 Avatars and voices of real people
- The person's written, informed and revocable consent is required, covering use by artificial intelligence, variants and the countries of distribution.
- In France, image and voice fall under personality rights; in the United States, under the "right of publicity" and, in several states, under specific laws on digital replicas.
- Keep proof of consent. Feedova may ask you for it.
6. Your data and artificial intelligence
What we do with your content to improve our models, and what our artificial intelligence providers do with the data they receive, is explained in the Privacy Policy and in the list of subprocessors.
7. Reporting content
To report content generated with Feedova that you believe to be unlawful: contact@feedova.ai, subject line "Report".