Statement on our use of artificial intelligence

Last updated: 25 September 2026

We keep this website evolving with the help of artificial intelligence, for its code and its translations, for some of its illustrations and for the contact form. On our Daily life page, we say that we hand it the things it does better than we do, such as generating code for our small in-house tools, sometimes producing visuals that would have taken us whole days in Photoshop, and copying, pasting, sorting and filtering thousands of lines in one go.

When we call on AI for our articles, our case studies, our glossary or their translations, it makes nothing up. It starts from our own words, written or transcribed, to tidy them up and, at most, link the ideas together. Autonomous AI agents systematically check that the result is faithful to them before we look at it ourselves for final approval.

The website's code

We make most changes to this website's code with Claude Code, Anthropic's tool. To check the values it builds in, Claude reads our mock-ups directly in Figma, through the MCP server that Figma provides.

What AI does to our texts

Most of our French texts come from the team. AI adjusts them lightly, following our editorial guidelines, and reworks the descriptions read by search engines. We write only a small share of them with AI, starting from our notes or from what we said out loud. The alt text of images, which screen readers read out, we also write partly with AI.

The English versions

The English versions of the website are translated with Claude. For the blog, the case studies and the glossary, it is instructed to add nothing, and another AI, OpenAI's Codex, reviews them to look for an addition or a mistranslation.

What we asked image models for

We generated some of the blog's illustrations with Midjourney. The photos of the team and the visuals of the case studies come from us and from our clients. On a few case study visuals, we had gpt-image, OpenAI's image model, anonymise people or produce in-context mock-ups.

What reads your message

When you write to us through the contact form, an OpenAI model reads your message to judge whether it is a genuine enquiry or spam, and its verdict decides whether an acknowledgement is sent to you, without ever deleting a message. A message it files as spam may not be read by the team. Our privacy policy details what it receives.

What the pages do not contain

The pages themselves contain no AI model: you are not chatting with any bot as you browse them, and the form sorting happens once your message has been sent.

We keep learning

We keep learning about AI, and the Daily life page describes how we keep up with the subject and the working group we set up around it.

How this page was written

Claude prepared it from what we had already published in Daily life and in our privacy policy, and from what the website's history and the code of its form show. Sami Lini reviewed and corrected the French version before publication.

For any question about one of these uses, write to hello@akiani.fr. And when one of them changes, we update this page.