Published on 04/07/25

How does our AI personalize your browsing experience?

Arthur
Arthur Guillermin Hazan
3 min
Share:

AI at the service of the user experience

On a content-rich site, the hardest part isn't producing content... it'shelping users find what's really useful. Articles, case studies, expertise: it's all there, but you still have to find the right content at the right time.

That's where our AI recommendation module comes in. It analyzes the pages already consulted to suggest the most relevant content to be explored next. Discreetly integrated into the interface, this effective support is designed to enrich the experience without disruption.

How does it work?

  • Page view detection

    During browsing, the URLs of the pages consulted are recorded directly in the browser, via local storage. No personal data is collected or transmitted: this approach makes it possible to analyze browsing locally, without tracking or server calls.

  • Search for the most relevant content

    When the module is displayed, a query is sent to our vector database (Qdrant), which contains all the site's structured content (articles, case studies, expertise, landing pages). The AI identifies the semantically closest content among those not yet proposed.

  • Generate personalized message

    Once the URL has been selected, a call is made to AI to dynamically generate the message displayed in the module: an introductory sentence and button text, adapted to the browsing context and recommended content.

  • Repeat management and call limitation

    The same content cannot be recommended twice. What's more, after three dynamic suggestions, the module automatically switches to a static version with a fixed message encouraging contact. This logic limits the number of AI calls while maintaining a consistent experience.

Data secure

A data-secure approach

The module is based on a strong principle: no personal data is collected.

  • Everything is stored locally, and nothing is shared on the server side.
  • No need for cookie banners.
  • No retargeting, no profiling, just instant, contextual recommendations.

We wanted useful personalization, not surveillance in disguise.

AI Ecodesign

An "eco-mastered" greenapproach

AI is a powerful resource, but it also consumes energy. To avoid technical abuse and remain consistent with our commitments:

  • The Qdrant database is updated once a day, not every time it is modified.
  • After 3 dynamic displays, the module switches to static content.
  • This system is hidden if no relevant suggestions are found.

Our thinking

Efficient AI doesn't need to be constantly in demand.
It's a question of balancing performance, impact and common sense.

 

Our AI on the site

We open the doors of our AI to you with complete transparency: what it does, what it doesn't do.

And tomorrow?

What are the possible uses?
Contact

This module is just the first example of what this technological brick can achieve. The same system could feed an intelligent search engine, a "you'll like it too" block, or even personalize home pages according to usage.

If these topics interest you, let's talk!

These articles may be of interest to you

Would you like to talk about it?

Contact us!
Contact

According to my analysis

Are you interested in our projects and our expertise? Why don’t we have a chat?