Half of Spain's internet users use generative AI, and ChatGPT leads, says the CNMC
IA en un minuto newsroom · Editor: Jon Elgezabal
In 30 seconds
CNMC, the body that watches over competition in Spain, has measured generative AI use among internet users: half use it regularly. ChatGPT leads with 41.3%, well ahead of Gemini and NotebookLM (17.2%) and Copilot (7.1%). The Household Panel also finds that just 8.6% pay and that 15.7% never check the answers they get.
One in two internet users in Spain regularly turned to generative artificial intelligence tools in the second quarter of 2026. That is the headline finding of the latest Household Panel from the CNMC, Spain's National Markets and Competition Commission, the body that oversees markets and competition, which published it on November 6.
ChatGPT tops the list: 41.3% of internet users used it regularly. Next come Gemini and NotebookLM, Google's assistant and its tool for working with documents, which the CNMC counts together (17.2%), and Copilot, Microsoft's assistant (7.1%).
Paying is still rare: only 8.6% of the people who use these apps paid for them.
On trust, 38.9% of users relied on the answers quite a lot (36.2%) or a lot (2.7%); four in ten said they were indifferent and 18.1% distrusted them. In total, 84.1% checked at some point whether the answers were true: 30.1% did so often, 35% now and then and 19.1% rarely. The remaining 15.7% never checked.
The Household Panel is a twice-yearly survey of households and individuals that also covers telecoms, audiovisual services, energy, post and transport. This round was carried out in the second quarter of 2026 with 6,009 households and 8,853 people.
Why it matters · analysis and opinion
The survey confirms that generative AI is no longer a niche for the curious in Spain, but it also shows use is still light: hardly anyone pays and trust in the answers is lukewarm. For a business, that cuts two ways. First, many of its staff and customers already use these tools, with or without permission, so it is worth deciding what data can go into them. Second, the habit of checking what the machine says is not as widespread as it should be: anyone using these answers for work needs a defined human review, not a hope that each person will do it. The dominance of a single tool also matters when choosing what to train a team on.
Official source: CNMC · Written with the help of AI: how we make the news
