Data Gouv Search: explore the catalogue of datasets with the help of an AI assistant

Data Gouv Search: explore the catalogue of datasets with the help of an AI assistant
Updated at: 15/10/2026

The Data Gouv Search catalogue is now open to conversational assistants. It is now possible to search it using everyday language, without needing to know the keywords or the structure of the metadata, making it easier to identify datasets relevant to a research project.

 

Natural language search, independent of the language of description

The catalogue’s search engine accepts queries phrased as questions, for example, ‘data on drought in the Mediterranean region since 2010’. It relies on the meaning of the query rather than an exact match of terms, which makes it possible to find datasets even when their descriptions are written in another language.

The results can then be refined by discipline, organisation, date or author.

 

An assistant capable of searching the catalogue

This development is based on a technical standard, the MCP (Model Context Protocol), which enables an AI assistant to query the catalogue directly and utilise the descriptions of the datasets. The assistant can retain up to 200 records during a single session, which opens up several possible uses:

- identify the teams and key contacts for a particular topic, with a view to collaboration;

 - compare the summaries and methodologies of several datasets, without having to consult them one by one;

 - identify cross-disciplinary links as the discussion progresses.

 

An open testing phase

A dedicated chatbot is available to try out the system, with no installation required. Anyone interested is invited to contact the team via the Recherche Data Gouv portal to join the test phase: contact the team via the portal (select: ‘Provide feedback’ or ‘Get help with the catalogue’) to join the test phase.

 

Find out more: Recherche Data Gouv catalogue page