AI & Acceptable Use Policy

Where AI is used

• Ad and social creative generation

  What the AI does

  Writes ad copy and, in some cases, synthesizes the background image.

  Human control

  Every creative is reviewed and approved by hotel staff before publication.

• Campaign strategy

  What the AI does

  Proposes audiences, budgets and schedules from the hotel's own performance data.

  Human control

  Staff approve before anything is published to an ad platform.

• Guest message drafting

  What the AI does

  Drafts replies to reviews and guest messages.

  Human control

  Drafts are queued for staff approval. A low rating is never answered automatically.

• Analysis and classification

  What the AI does

  Summarizes performance, classifies inbound messages, and extracts facts from the hotel's own website.

  Human control

  Advisory only. No outward-facing action follows from it on its own.

Hostli does not use AI to make decisions that produce legal or similarly significant effects on individual guests.

Disclosure of AI-generated images

Where a creative’s background image is synthesized by AI, Hostli embeds a signed C2PA Content Credentials manifest into the published image. The manifest is machine-readable, travels inside the image file itself, and declares the action as a creation, the IPTC digital source type "trainedAlgorithmicMedia" (created using generative AI), and the software agent that produced it. This is our EU AI Act Art. 50(2) marking mechanism.

What we mark

• creatives whose background image was generated by an AI image model;
• photographs that a person has deliberately altered with AI, whether in a campaign creative, an organic social post, or a flyer. These are marked in every case, because a person asked for AI pixels on purpose.

What we do not mark, and why

Creatives built on a real photograph supplied by the hotel, where AI has only placed text, brand elements or a template around it. None of the photograph’s pixels are synthesized. Marking these would tell a viewer that a genuine photograph of a real hotel room is AI-generated, which is itself a false disclosure. We take the honest boundary to be pixel synthesis, not "an AI system was involved somewhere in the workflow".

Marking is fail-closed. If the manifest cannot be written, the render fails and the unmarked image is never uploaded or published. We do not fall back to publishing an unmarked AI-synthesized image.

Limits you should know about

• Our manifests are signed with a certificate issued by Hostli. They are standards-valid and any C2PA-compatible tool will read the provenance, but our root is not yet on the C2PA Trust List, so a verifier may present the issuer as unrecognized. Trust List membership is a voluntary reputation layer rather than an Art. 50(2) requirement, and submission is on our roadmap.
• We do not rely on invisible watermarks such as SynthID. Where an upstream model embeds one, our compositing step re-rasterizes the image and the upstream watermark cannot be relied upon in the published asset. The C2PA manifest is the mechanism we stand behind.
• Story videos and reels produced by Hostli do not currently carry a C2PA manifest. We are stating this plainly rather than leaving it to be discovered.

Acceptable use

Hotel customers and their staff may not use Hostli’s AI features to:

• generate imagery that misrepresents the property (rooms, views, facilities or amenities the hotel does not have);
• depict identifiable people who have not consented to appear in marketing, or generate imagery of guests;
• produce content that is discriminatory in its targeting or its message, or that targets individuals on the basis of a protected characteristic;
• make claims about price, availability or terms that the hotel cannot honour;
• remove, strip or alter the Content Credentials manifest from a Hostli-generated image;
• present AI-generated imagery as documentary photography of the property where a guest's booking decision would turn on the difference.

Hostli may suspend AI features for an account that breaches this section. Publication remains the hotel’s decision and the hotel’s responsibility: we generate and disclose, our customers approve and publish.

Human oversight

No AI-generated creative reaches an advertising platform or a public social account without a person approving it. Guest-facing replies are drafted, not sent. A review of three stars or below that carries no written text is routed to staff with a fixed, neutral acknowledgement rather than an AI-drafted apology, because a model given a silent one-star rating will invent a grievance the guest never described.

Models and sub-processors

The AI providers Hostli uses are listed, with their roles and locations, on our sub-processor page. Hostli does not permit its AI providers to train their models on customer or guest data.

Changes

We will update this Policy when our AI use or our disclosure mechanism changes materially, and will record the effective date above.