AI Citations Barometer: France and the UK

Method

Two panels of 2,998 distinct queries, one French and one British, that come not from the media but from demand. Topics: the most-viewed Wikipedia pages in each country from 5 to 18 August 2026, set against their July audience, so as to keep only what was actually in the news; the nature of each entity is typed through Wikidata to stratify by theme. Queries: for each topic, the queries people actually type into Google and their volumes, through DataForSEO Labs (up to 40 per topic), then a draw stratified by theme (22% cap per theme) and by volume band, with a fixed seed so it can be replayed; the panel was frozen on 18 August for the whole measurement period. Two consequences to keep in mind: these are queries on topics that were in the news, not breaking-news queries in the strict sense (86 French and 77 British queries carry an explicit news marker, the others are about the entity itself); and since the panel is derived from Wikipedia entities, Wikipedia's place among cited sources is partly a construction effect. Every query was submitted once a day around 15:00 UTC, from 19 to 25 August 2026, to Google France and Google UK, at country level, no city, on desktop, through the DataForSEO SERP API with AI Overviews loaded; a second collection on 19 August (an afternoon test) is excluded. For each record we logged whether an AI answer appeared, the full list of cited sources and the top ten classic results: 2,998 distinct queries per country (the draw of 3,000 contains two duplicates), 41,971 records over seven days, no failed query. Cited sources are classified using public reference lists: certified press (ACPM in France; PAMCo, Newsworks and Newsquest in the UK, completed with Reach plc regional titles), broadcast news, TV listings, encyclopedias, social networks (Facebook, YouTube, Instagram, TikTok, Reddit, X, LinkedIn, Threads, Pinterest) and other sites, the latter being the largest category by volume (40.6% of citations in France, 52.3% in the UK). Every published rate comes with a 95% confidence interval.

Limitations

Desktop only, at country level with no city, seven consecutive days in August: these are dated figures. The AI answer rate rises by 3 to 4 points between 19 and 25 August, then levels off in France from 22 to 25 (52.1 / 53.0 / 52.7 / 52.2). Two readings are possible, and our data cannot separate them: Google is widening coverage; or the panel's topics, frozen on 18 August, are cooling down, and Google is removing Top Stories where the AI Overview takes their place. The seven-point gap between the two countries is clear (the bootstrap intervals do not overlap), but the two panels are not identical (themes, volumes, 28% of US topics on the British side): weighted by theme, the gap widens to 10 points (47.9% versus 57.8%). Press reference lists are not symmetrical: ACPM certifies named websites, while the UK lists are reconstructed from several sources, so the France / UK comparison on certified press (16.0% of citations versus 5.9%, 7.8% with regional titles and magazines) is indicative; on topics of national origin only, the gap narrows. The 'all domains' and 'media' perimeters are comparable. The 'other sites' category, the largest by volume (40.6% in France, 52.3% in the UK), contains on the British side about 1,700 citations of non-certified foreign press (people.com 694, forbes.com 345, businessinsider.com 257, nytimes.com 225, variety.com 179), 2.9 points that the press loses by construction; on the French side, dailymotion.com (930) and the Yahoo aggregators. Without YouTube, social networks take 15.5% of citations in France against 16.0% for certified press: a tie. In the UK the gap holds either way (15.1% versus 7.8%). The Daily Mail gets only 17 citations, but dailymail.co.uk does not appear in the organic top 10 of the AI Overview pages in this panel: the AI is not excluding it, organic search is not ranking it. Three cited URLs out of ten do not come from the organic top 10 (29.1% in France, 31.8% in the UK), and a citation is not a click. The intervals on citation shares treat the seven records as independent although the same answers cite the same domains every day; they are therefore too narrow, and the conclusions rest on the bootstrap intervals of the rates. Finally, very high-volume queries trigger far fewer AI answers than the rest: weighted by search volume, the rate drops to 17.6% in France and 13.6% in the UK. Likewise, on topics of French origin alone, certified press and social networks are level (21.5% against 21.2%): the headline describes the full panel and the UK unambiguously, France only narrowly.