AI-Generated & Synthetic Content Disclosure
We believe you should always be able to tell which parts of MandarinMiles were made by a person and which were produced with generative technology. This page describes where we use AI, how those assets are labelled, and what we do not do.
1. What is AI-generated or AI-assisted
- Illustrations and decorative artwork. Some images used in readers, banners and promotional areas are created with generative image tools, then selected, cropped, colour-adjusted and reviewed by a person before publication. We describe these as AI-assisted illustrations.
- Narration audio. Reader audio is synthesised speech (text-to-speech) generated from our own written text. It is not a recording of a human voice actor, and it does not imitate or impersonate any real person.
- Tools used. Illustrations are generated using third-party image-generation services; narration is generated using third-party text-to-speech services. We do not disclose the names of individual vendors as a matter of policy, but will provide them on request for transparency.
2. What is not AI-generated
Reader texts, pinyin annotations and tone-change notes, word lists, translations, segmentation, the level/colour system, grading rules, the interface and the reading experience are authored or compiled by people from our own work and from open linguistic datasets. Those datasets, their licences and our modifications are documented on the Data Sources & Credits page. Automated grading rules assist editors but a person remains responsible for the published level.
3. How we label it
- Visible labelling: AI-assisted images carry a caption or nearby label such as “AI-assisted illustration”, and their alternative text describes the scene rather than implying a photograph or a named human artist.
- Page-level disclosure: this page is linked from the footer and from the Privacy Policy, Terms, EULA and Copyright pages, so a reader can always find the general disclosure even where an individual caption is missing.
- Machine-readable provenance: where a generation tool provides signed metadata or Content Credentials (for example C2PA-style records), we aim to preserve it when publishing and avoid processing steps that strip it. Not every asset or tool supports such metadata, so we do not claim that every image carries one. The obligation to embed machine-readable marks sits primarily with the provider of the generative system; as the publisher of the content we focus on clear visible disclosure and on retaining the provenance information we receive.
- Audio: narration is identified as synthesised speech in the reader interface and in this disclosure.
4. Our editorial commitments
- Every AI-assisted image is reviewed by a person before publication for accuracy, cultural sensitivity and suitability for children.
- We do not use generative tools to create realistic depictions of real, identifiable people, or to fabricate events, quotes, reviews, testimonials or news about matters of public interest.
- We do not present AI-assisted images as documentary photographs, and we do not use synthesised voices to impersonate a real individual.
- We do not use your personal data, saved words or reading history as prompts or training data for generative tools.
5. Legal context
Transparency obligations for certain AI systems under Article 50 of the EU AI Act have applied since 2 August 2026, with a voluntary Code of Practice on transparency of AI-generated content published by the European Commission in June 2026. Those rules distinguish the provider of a generative system (who must enable machine-readable marking) from the deployer (who must disclose deepfakes and certain AI-generated text on matters of public interest). We publish this disclosure both to meet our obligations as a deployer and to give learners straightforward information regardless of whether a specific legal duty applies to a given illustration. This page is a transparency statement, not a certification or a claim of regulatory approval.
6. Copyright and attribution
We claim rights in the selection, arrangement, editing and compilation of AI-assisted artwork, but we do not assert exclusive copyright in raw machine-generated output where the applicable law does not recognise it. Third-party open-data licences (including CC BY-SA 4.0) are recorded on our Credits page. See also the Copyright & DMCA Policy.
7. Corrections and questions
If you see an AI-assisted image without a label, believe something has been mislabelled, or want to know how a specific asset was produced, email [email protected] with the page URL. We will review and correct the labelling. Material changes to how we use generative content will be recorded here with a new date.
MandarinMiles is operated by David Chen, an individual based in mainland China.