This is a living page. Every row has a number, what it means, the source and the date, and every source has a caveat on the sources page. Spain-specific studies are marked "Spain" in the source column; the rest measure mostly English or US prompts and are here to explain the mechanism, not to size the Spanish market. If you want the same measurement for your own brand, the audit method explains how we run it and the hub explains how it is priced. Terms are in the glossary.
1. Adoption of assistants in Spain
| Number | Meaning | Source | Date |
|---|---|---|---|
| 4% to 14% to 28% | People in Spain who say they use ChatGPT frequently, 2023, 2024 and 2025 waves | Funcas (Spain), online panel, n=1,200 | Jan 2025, fieldwork Dec 2024 |
| Four in ten | Under-45s who use some AI tool daily | Funcas (Spain) | Jan 2025 |
| 11%, 6%, under 2% | Frequent use of Gemini, Copilot, and Perplexity, Claude or Grok | Funcas (Spain) | Jan 2025 |
| 61% | People in Spain who have used generative AI, up from 5% in 2023 | Bain, Spain figures via IT Reseller (Spain) | Feb 2026 |
| 22% | People in Spain who have replaced traditional search with AI chatbots most of the time, reported as the highest rate in Europe | Bain via IT Reseller (Spain) | Feb 2026 |
What to take from it: frequent use of ChatGPT doubled in a year and the gap between "have used" (61%) and "use frequently" (28%) is where the next two years of growth sit. Gemini is a distant second and Perplexity is near zero in Spain by Funcas' measure, which is why our audits report per assistant rather than one blended score.
2. What happens to organic search
| Number | Meaning | Source | Date |
|---|---|---|---|
| 52.02% to 40.70% | Organic search's share of total traffic to Spanish sites, full-year 2025 versus January to April 2026, an 11.32-point drop against 2.59 points globally | SE Ranking (Spain), 101,574 sites | Jun 2026 |
| 0.30% | Share of traffic to those sites that arrives from AI platforms in 2026 | SE Ranking (Spain) | Jun 2026 |
| 70.49%, 12.41%, 11.64% | Share of that AI referral traffic from ChatGPT, Gemini and Perplexity | SE Ranking (Spain) | Jun 2026 |
| −58% | Lower click-through rate for the top organic result when an AI Overview is present, informational keywords, December 2025 data; the previous wave was −34.5% | Ahrefs via Medianama (global) | Feb 2026 |
| 38% | Pages cited in AI Overviews that also rank in the top 10 for the query, down from 76% in July 2025; 31% rank beyond position 100 | Ahrefs via Search Engine Journal (global) | Mar 2026 |
| 161% | How much more likely a page ranking for fan-out sub-queries is to be cited than one ranking only for the main query; 68% of cited pages ranked top 10 for neither | Surfer via Search Engine Land (global) | Dec 2025 |
| About 15% (unverified) | Share of Spanish searches that show an AI Overview, among the highest in Europe, attributed to Semrush | Marketing4eCommerce (Spain); link could not be fetched on 9 Sep 2026 | 2026 |
What to take from it: the click is leaving before the AI referral arrives. Organic lost eleven points of share in a year while AI platforms sent 0.30% of visits. Ranking first still helps, but only 38% of AI Overview citations come from the top ten, and pages that answer the sub-questions (price, timeline, comparison, "how it works") are cited far more often than pages that only rank for the head term.
3. What AI answers cite
| Number | Meaning | Source | Date |
|---|---|---|---|
| 84% | Citations in AI answers that point to earned media; the figure has stayed between 82% and 89% across three editions since July 2025 | Muck Rack, 25M+ links, ChatGPT, Claude, Gemini (mostly English) | May 2026 |
| 27% | Citations that are journalism specifically | Muck Rack | May 2026 |
| 0.3% | Citations that are paid or advertorial content | Muck Rack | May 2026 |
| 96%, 82%, 55% | ChatGPT, Gemini and Claude responses that include citations | Muck Rack | May 2026 |
| 3.5 times | How much more often press releases appear in "industry trend" answers than in other answer types | Muck Rack | May 2026 |
| 63% | Citations pointing to listicles, across about 400M citations and six engines; 71% to 86% of those listicles are ranked lists | Evertune via Search Engine Land (mostly English) | May 2026 |
| 62% | AI citations that are ghost citations: the URL is cited, the brand is not named. ChatGPT cites 87% of the time and names the brand 20.7%; Gemini names 83.7% and cites 21.4% | Semrush and Growth Memo, 115 prompts, 14 countries | Jun 2026 |
| 43.3% versus 18% | Brand mention rate on comparative prompts versus informational prompts | Semrush and Growth Memo | Jun 2026 |
| 15.2% | ChatGPT categories with a clear brand owner; owners kept the spot in 90.4% of month-over-month comparisons | Semrush and Growth Memo, 1,094 US categories (US only) | Jul 2026 |
| 5,806 | El País citations on ChatGPT in Spanish prompts, the fifth most-cited domain overall; 5,143 on Google AI Mode; El Español leads Spanish media on Gemini with 1,434 | Ahrefs ES (Spain), 2,262,194 prompts | Jul 2026 |
What to take from it: assistants cite what other people publish about you far more than what you publish yourself. Paid placements are statistically absent. Ranked lists and comparisons are the formats that get cited and, separately, the formats that get a brand named rather than ghost-cited. In Spain the media that assistants trust are the national newspapers first.
4. How unstable answers are
| Number | Meaning | Source | Date |
|---|---|---|---|
| 56% | Google AI Mode's cited sources replaced every week; ChatGPT replaces 74% | SISTRIX tracking, reported by Kevin Indig in Search Engine Land | Jun 2026 |
| 2.2% | ChatGPT citations that remain after running the same prompt three times | AirOps, 815,000 prompt-page pairs, reported in the same article | Jun 2026 |
| 5 runs per prompt per platform per week | Indig's recommended minimum for a report with confidence intervals; a single run is not a measurement | Kevin Indig, Search Engine Land | Jun 2026 |
| 10.13%, no effect | Domains with an llms.txt file, and no measurable relationship between the file and citation frequency | SE Ranking via Search Engine Journal, about 300,000 domains | Nov 2025 |
What to take from it: any single screenshot of an AI answer is an anecdote. A share-of-voice number is only meaningful with repeated runs, per assistant, with an error bar. That is why the audit method runs five passes per question and reports a range, and why nobody honest can guarantee a placement.
5. What it means for a Spanish brand
Three plain conclusions, each tied to the rows above.
- Measure before you spend. Frequent ChatGPT use went from 4% to 28% in two years and 22% of people in Spain already prefer a chatbot to a search engine most of the time. Whether your brand appears in those answers is a knowable fact, but only with repeated runs per assistant (56% of AI Mode sources change weekly; 2.2% of ChatGPT citations survive three runs). Start with a baseline that has an error bar.
- Earned media is the lever, and PR already produces it. 84% of citations point to earned media and 0.3% to paid content. A press mention in El País, an industry outlet or a review site does more for AI visibility than any file or markup on your own site. The Spanish ad market fell 2.6% in 2025 (InfoAdex, the first drop in five years); the budget that moves is the one that buys coverage, not the one that buys the answer box, because the answer box is not for sale.
- Publish the pages that answer sub-questions, in ranked and comparative formats. 63% of citations go to listicles, comparative prompts get a 43.3% mention rate against 18% for informational ones, and pages ranking for fan-out sub-queries are 161% more likely to be cited. Only 15.2% of ChatGPT categories have an owner in the US data, and owners keep their place 90% of months. In Spain most categories are still open; the first brand with real depth tends to keep the spot.
6. How this page is kept current
Every figure here is a published study, named, dated and linked, with its sample and its caveat on the sources page. Spain-specific studies are marked as such, so market context is never confused with mechanism. Each quarter the links are fetched again and compared with the study, new waves replace old figures, and the previous figure stays in the note so the trend is visible. Anything we cannot open and verify is labelled unverified or left out. For the same measurement on your own brand, the audit method sets out the design: five runs per question, results per assistant, and a confidence interval on every rate.
