AI Transparency
Last updated: 21 October 2026 · version 2026-10
Nudge uses AI to help you build a good calculator, scan or product finder faster. This page explains what the AI does, what it sees, where you stay in control and how we comply with the EU AI Act.
What the AI does
- Building: based on your website, a description or files, the AI suggests questions, results, calculations and a welcome and result screen.
- Editing: you ask for a change in plain language ("make question 2 shorter") and the AI adjusts the Nudge.
- Follow-up: the AI drafts follow-up emails and suggests improvements based on how emails perform.
- Improving: once there are enough visitors, the AI makes suggestions based on counts: where people drop off and which answers are chosen most.
- Translating: with "Duplicate in English" the AI translates a copy of your Nudge.
- Support: the help button can have AI give a first answer.
Which model
Nudge uses Anthropic's Claude models through their business API. Anthropic is a sub-processor of Nudge; see sub-processors.
What the AI sees, and what it doesn't
- Yes: your instructions, the texts and settings of your Nudge, the public texts of the website you enter, the files you put in the knowledge base, your support question, and counts and averages of answers.
- No: names, email addresses, phone numbers and other contact details of leads. Those stay in your Lead Vault.
- No training: neither Nudge nor Anthropic uses your data to train AI models. Anthropic keeps inputs and outputs only briefly to investigate abuse.
You stay in control
- Everything the AI creates is a suggestion. You review it, edit it and decide when something goes live.
- Suggestions are automatically checked for inappropriate content before you see them.
- Calculations get a plausibility check; always verify amounts and promises yourself.
- The AI makes no decisions about visitors. The lead score is not AI but a calculation based on weights you can see and adjust.
What your visitors see
A visitor does not talk to an AI in a Nudge: they answer fixed questions you published. So there is no chatbot that requires a disclosure.
The AI Act does require AI-generated text to be recognisable in a machine-readable way (Article 50(2)). Therefore:
- the element that places a Nudge on your website carries the attributes
data-ai-generated="assisted"anddata-ai-provider="nudge"; - follow-up emails written with AI carry the header
X-AI-Generated: assisted.
Don't remove these markings. If you use a Nudge to inform the public on matters of public interest, also state visibly that the text was created with the help of AI.
What's not allowed
Don't use Nudge for applications the AI Act prohibits or classifies as high-risk without the required safeguards. Think of assessing job applicants, creditworthiness, insurance or access to education and public services, where the result determines what someone gets. If you use Nudge for recruitment, never let the result decide on its own: a human reviews every candidate. See also article 8 of the terms of service.
AI literacy
Everyone at Nudge who works with AI understands how the models we use work and where their limits are. In the builder and the help texts we explain what the AI does, so your team handles the suggestions responsibly too.
Questions
Questions about AI in Nudge? Email info@nudgedigital.nl.
The Dutch version of our legal documents is legally binding.