ProntoWeb WhatsApp

Version ki-2026-11-07 · As of 7 November 2026 · Deutsche Fassung

This is the English version. Consumers receive these terms in the language of their conversation, and that version applies to them. For businesses, the German version prevails if the versions differ (Terms § 2 (5)): Hinweis zum KI-Einsatz

Notice on the Use of AI

Annex 6 to the General Terms and Conditions of ProntoWeb · for information

§ 1 Where AI works for us

StepWhat the AI does
ConversationAn AI assistant answers your WhatsApp messages and asks about what the website still needs.
Voice messagesAn AI transcribes your voice messages.
Photos and videosAn AI describes your photos and videos so that they are placed in the right spot. Of a video, it only sees individual still images, never the sound.
Stock photosAn AI selects the suitable motifs from the results of the Pixabay search.
Documents and linksAn AI reads documents that you send and pages whose link you send, and summarises them. We adopt the text of a page only if you name the page to us as your own.
Texts and designAn AI writes the texts from your information and designs the page.
Generating images (once offered)An AI generates images, but only with your yes.
ChangesAn AI carries out your change requests.
ReviewAn AI assesses the draft before you see it.

§ 2 What no AI does

These things are determined by our system according to fixed rules, not by an AI:

Decisions about your contract are not made by an AI either. Where our system acts automatically, for example in the case of an outstanding payment (Terms § 4 (6)), it follows the deadlines of these terms.

§ 3 Labelling

  1. At the start of the conversation, we tell you that you are writing with an AI assistant. The texts of your website are created with AI from your information. You decide which of them appear on your website.
  2. Paragraphs 2 to 4 apply once we offer AI images (Terms § 3 (8) no. 3). We visibly label AI images that may look like photos on the image itself, for example with an abbreviation in the image or a notice directly below it. Images that are recognisably not photos do not need this. In the legal notice of your website, we name every AI image. All of this is applied by our system according to fixed rules, not by the AI.
  3. Every AI image also contains a machine-readable label: signed provenance data under the C2PA standard and a watermark. Both are preserved when we reduce the size of the image.
  4. Whether an image was created with AI can be determined with freely available C2PA verification tools.

§ 4 Who is responsible

  1. Your website is your presence. You are its provider, and its legal notice names you. You are responsible for the accuracy of the content (Terms § 11 (3)).
  2. We are responsible for our service: that the AI uses only what you have said and sent, and that the page works technically. We are liable under Terms § 11.
  3. Our system therefore checks, among other things, whether names and places on the page come from your information. Errors can nevertheless occur. Please check every version (Terms § 7 (2)).

§ 5 Humans

At your request, or if the assistant cannot proceed, a human looks at the conversation (Terms § 9 (1)).

§ 6 Providers and data

  1. For the steps in § 1, we use AI models from various providers. Which ones they currently are, with their registered office and the basis for a transfer to third countries, is set out in our list at prontoweb.io/ki-anbieter. We add to it before a new provider receives your data.
  2. We use only providers

    • with which we have a contract under Art. 28 GDPR, directly with the provider or with Vercel as intermediary (paragraph 3),
    • that process data outside the EU only on a basis under Art. 44 et seq. GDPR and
    • that do not use your content to train their models if we require this (paragraph 4).
  3. We route our requests to the language models through an intermediary service, the Vercel AI Gateway of Vercel Inc. (USA). It is also included in the list. Vercel is certified under the EU-US Data Privacy Framework. Your voice messages are transcribed by Mistral AI (Paris, France). These requests go directly to Mistral, not via Vercel.
  4. We require all our AI providers not to use your content for training: via Vercel with every request, and for Mistral through the contract and the settings of our account. Via Vercel, we also require for all requests that the providers do not store your content. Mistral may retain requests for up to 30 days in order to detect misuse. We do not guarantee immediate deletion.
  5. xAI (USA) is not certified under the EU-US Data Privacy Framework. For data sent to xAI, we therefore rely on the EU Standard Contractual Clauses in our contract with Vercel. The same applies to any other provider in the USA without this certification. Mistral AI has its registered office in France. With Mistral, its data processing agreement applies.