GEO in Spanish: how to show up in AI answers
Version 0.1 — living draft. I publish from day one and update with a date. Comments and corrections welcome via contact.
What GEO is
Generative Engine Optimization (GEO) is the discipline of getting content selected, cited and summarised by generative engines such as ChatGPT, Perplexity, Gemini or Google’s AI Overviews. Where classic SEO chases a slot in the result list, GEO chases a slot inside the answer.
Why now
- AI Overviews appear on roughly 45% of Google searches and cut organic clicks by up to 58%: the answer is consumed inside the result.
- ChatGPT, Perplexity and Gemini are already massive discovery channels in Spanish, where the ecosystem of quality citable sources is still small: there’s room.
- Brand over domain: AIs cite third-party sources (press, forums, directories) more often than a business’s own website.
The three levers
1. Extractable structure
Generative engines extract passages, not pages. Every key claim must stand on its own, out of context:
- A definition block at the top (40–60 words is the optimal extraction size).
- Headings that answer real questions, in natural language.
- Tables for comparisons, numbered lists for processes, FAQs for specific doubts.
- Data with source and date: “67% of deepfakes detected in 2025 were tied to legal cases” beats “many deepfakes end up in court”.
2. Citable authority
According to the Princeton GEO study (KDD 2024), measured on Perplexity:
| Technique | Visibility impact |
|---|---|
| Citing sources | +40% |
| Including statistics | +37% |
| Named expert quotes | +30% |
| Authority tone | +25% |
| Clarity | +20% |
| Keyword stuffing | −10% |
The winning combination is fluency + data: on domains with little prior authority, the improvement reaches 115%.
3. Presence where they look
AIs cite where there is conversation and contrast: local and industry press, Wikipedia/Wikidata,
forums with real participation, podcasts (transcribed), YouTube (the text layer). Before asking
to be cited, check that they can cite you: robots.txt open to GPTBot, ClaudeBot,
PerplexityBot and Google-Extended.
What to measure
Five brand queries and five category queries, once a month, on ChatGPT, Perplexity and Gemini: do you appear?, who appears?, which page of yours gets cited? A spreadsheet is enough to see the trend. If you want tooling, it exists (Otterly, Peec, ZipTie), but the manual method teaches more than the report.
What not to do
- Parallel “for AI” content (risk under Google’s scaled-content-abuse policy).
- Fragmenting pages into clickbait crumbs for extraction.
- Fabricated mentions in forums or Wikipedia: reputational and algorithmic penalty.
- Blocking AI crawlers if you want to be cited.
Launch checklist
- Clear definition in the first paragraph of every important piece.
- Visible author, with credentials and a link to their profile (E-E-A-T).
- Visible publication and update dates.
- At least one sourced data point per section.
llms.txtand robots.txt reviewed.- Coherent entity: same name across web, LinkedIn and press.
This guide is part of jquin.net. If you find it useful, link to it: that’s exactly the mechanism it describes.