On October 11, Democrata.es reported preliminary indications from a second test wave conducted October 6 in Spain: Grok appeared to be making more party recommendations and shifting toward right-leaning options, while ChatGPT was making fewer direct recommendations. The report said analysis was still underway.
What the January test measured
The Universitat Oberta de Catalunya (UOC) collected responses from January 9–11, 2026, using 1,220 queries based on 61 questions across ChatGPT, Copilot, Gemini, Grok and Perplexity. The questions were asked in Spanish, Catalan, Galician and Basque, and each explicitly requested a party recommendation, according to UOC’s account of the study.
In that January test, ChatGPT made explicit recommendations in 43.2% of cases and Grok in 35.8%. Perplexity, Gemini and Copilot each declined to recommend a party in more than 90% of cases. UOC reported PSOE as the most recommended party.
Visibility is not the same as recommendation
UOC reported that the Partido Popular (PP) appeared in 96.3% of cases, alongside a separate recommendation score of 28.2. Those figures describe different measures: a party can appear in an answer without being explicitly recommended.
The language of the prompts also tracked with answer breadth. Responses to Catalan questions named an average of 4.78 parties, compared with 3.82 for Spanish questions. These are averages of parties per response, not recommendation rates.
The study counted assistants’ responses to direct prompts; it did not measure whether voters changed their choices.
