We analyzed 1,610 sources cited by ChatGPT, Gemini, Google AI Overview, and Perplexity in three Spanish-speaking markets (Argentina, Spain, and Uruguay). The conclusion: there is not a single strategy for AI positioning for the Spanish-speaking world. Each market has its own rules, and here are the data for you to adapt yours.
When you ask ChatGPT what sneakers to buy, where does it get the answer? Does it read blogs in Spanish? English forums? Instagram Reels? The manufacturer's website?
The answer matters, a lot. Because if you understand where the AI extracts information in your sector, you know exactly where you need to be for it to mention you. That is, at its core, all about GEO (Generative Engine Optimization).
So instead of theorizing, we measured it. We took a specific niche —running shoes—, ran it in three Spanish-speaking markets of different digital maturity (Argentina, Spain, and Uruguay) and classified each of the 1,610 sources cited by ChatGPT, Gemini, Google AI Overview, and Perplexity. This is what we found.
There is no single GEO strategy for the Spanish-speaking world. The ecosystem of each country reconfigures which engine leads, which platform matters, and which content format the AI actually cites.
The goal was not to measure brand positions, but to understand the mechanics: where the models get their information based on the country. For that, we designed a replicable protocol.
We entered with two strong hypotheses. Both fell through. And that's where it got interesting.
First surprise. Even if you ask in Spanish, about a product sold in your country, the models do not respond only with local content. We measured what percentage of the cited URLs were in Spanish versus English/global.
| Market | URLs in ES | URLs EN/global | % EN | Ambiguous UGC |
|---|---|---|---|---|
| Argentina | 313 (65%) | 104 | 22% | 64 |
| Spain | 315 (57%) | 145 | 26% | 93 |
| Uruguay | 281 (49%) | 132 | 23% | 165 |
Distribution of the language of the cited sources, by market.
The pattern is clear: the more mature the market, the higher the proportion of content in English. Spain leads with 26% because the models bring technical content from the USA and UK as a complement to local. Uruguay has less pure English, but its ambiguous language UGC skyrockets to 165 sources: the Uruguayan user consumes much more content without clear localization.
The practical conclusion hurts a bit: producing only in Spanish leaves you without covering a quarter of the citation territory. Technical content in English does not get lost against local Spanish. It coexists with it.
Here was our second hypothesis: we thought that more than 90% of the cited YouTube videos would be Shorts, purely due to the volume of publication. And yes, many Shorts are cited... but they hardly count.
| Market | Long (URLs / citations) | Shorts (URLs / citations) | Long advantage |
|---|---|---|---|
| Argentina | 2 / 58 | 12 / 15 | 23× |
| Spain | 2 / 56 | 9 / 9 | 28× |
| Uruguay | 2 / 48 | 11 / 15 | 18× |
In all three markets, 2 long videos accumulate more citations than 9–12 Shorts combined.
Read it well: in Argentina, two long videos gathered 58 citations. Twelve Shorts gathered 15. The citation multiplier per URL for the long format ranges from 18× to 28× depending on the market. And the pattern repeats identically in all three countries.
We also found something useful for any regional brand: there are pan-Hispanic assets, long videos that cross borders. A single well-positioned piece appeared cited in Argentina (22 citations) and Spain (14). Another dominated Spain (42) and Uruguay (48) at the same time. A good long video does not perform in one market: it performs in several.
The format /watch?v= remains the most efficient visibility asset on YouTube for LLMs. Shorts provide volume; long videos provide authority. For GEO, prioritize long and deep pieces over the quantity of clips.
There is a widely repeated belief in GEO: "all social media content feeds Google AI Overview". We measured it. It is true... partially. Each social platform has its engine, and it is not the same.
| Platform | ChatGPT | Perplexity | AI Overview | Gemini |
|---|---|---|---|---|
| Instagram (3 markets) | 0 | 0 | 84 URLs | 3 |
| TikTok (3 markets) | 0 | 0 | 45 URLs | 0 |
| YouTube (3 markets) | 0 | 3 videos | 35⟧ | 0 |
| Reddit (3 markets) | 166 | 0 | 54 | 5 |
Social URLs cited by each engine, aggregated across the markets.
The map turned out this clear:
The phrase "UGC = AI Overview" is ~75% true: it applies to Instagram and TikTok. But the social play needs to be segmented by engine. Instagram/TikTok drive AIO; Reddit mainly feeds ChatGPT; YouTube plays in both. There is no single social action that covers all engines.
If Instagram feeds AI Overview, the logical question is: what format within Instagram? The video Reel or the static post with copy? We count the citations of each.
Cross-market and compelling: AI Overview prefers Reels at least 8 to 1 over static posts. And the more mature the market in Instagram usage, the more it leans towards the Reel (Uruguay reaches 93%).
The implication is direct for any brand: the copy of a static post contributes very little to GEO. What the models extract is the audio and the caption of a Reel. If you are going to invest in Instagram for visibility in AI, the vertical video format is what pays.
Reddit es una de las fuentes más golosas para la IA, especialmente para ChatGPT. Pero cuando desglosamos el idioma real de esos hilos, apareció otra sorpresa.
| Mercado | EN crudo | EN traducido (tl=) | ES nativo | /user/ perfiles |
|---|---|---|---|---|
| Argentina | 8 | 15 | 4 | 0 |
| España | 27 | 17 | 21 | 12 |
| Uruguay | 45 | 24 | 31 | 21 |
Idioma y tipo de URL de Reddit citadas, por mercado.
Argentina depende casi por completo de la traducción automática de Google: el 65% de su Reddit citado son URLs con tl=es-419, es decir, hilos en inglés que el modelo tradujo. España consume Reddit en tres formatos balanceados (inglés crudo, inglés traducido y español nativo) sin que ninguno domine. Y Uruguay es el "Reddit power user": más inglés crudo, más perfiles de usuario y más subreddits en español. El patrón más diverso de los tres.
Posicionar contenido en Reddit en español tiene techo bajo: los modelos ya resuelven el idioma traduciendo hilos en inglés. La oportunidad real está en participar en los subreddits en inglés de alto tráfico que después se traducen, más que en levantar comunidad hispana desde cero.
Más allá de la plataforma, ¿qué categoría de fuente cita la IA? Clasificamos cada URL en editorial, UGC, marca (fabricante), retailer (multimarca) o prensa.
| Categoría | AR | ES | UY |
|---|---|---|---|
| Editorial | 46% | 47% | 38% |
| UGC | 18% | 26% | 38% |
| Marca | 30% | 10% | 13% |
| Retailer | 4,6% | 16% | 10% |
| Prensa | 1,6% | 1,4% | 1% |
Peso porcentual de cada categoría de fuente, por mercado.
Three things stand out. First: editorial content is the only stable category, around 40% in all three markets. It is the most reliable ground, rain or shine. Second: UGC doubles from Argentina to Uruguay (18% → 38%); the less local editorial ecosystem, the more UGC to fill the gap. And third, the most curious: brand and retailer are inversely proportional.
After crossing everything, we found the master variable. A single metric predicts the complete behavior of a market in relation to AI: the percentage of URLs in local domain (.uy, .ar, .es).
| Market | % local domain | % UGC | Reddit in ChatGPT | Top engine |
|---|---|---|---|---|
| Spain | 53% | 26% | 60 URLs | ChatGPT |
| Argentina | 18% | 18% | 11 URLs | Perplexity |
| Uruguay | 6% | 38% | 95 URLs | ChatGPT |
Localization of the ecosystem vs engine behavior.
The correlation is clear: the less localization, the more weight of UGC and Reddit; the more localization, the more retailers and editorials. Spain, with 53% local domain, gives its visibility to solid retailers and editorials. Uruguay, with only 6%, gives it to UGC and Reddit. Argentina, in the middle, gives it to direct brands.
Seven findings, translated into concrete decisions you can apply tomorrow:
The big lesson from the study is also the most uncomfortable: you cannot copy the GEO strategy from one market and paste it into another. What works in Spain (retailers + editorial + ChatGPT) fails in Uruguay (UGC + Reddit). The ecosystem dictates. Your job is to read it before acting.
According to this study of 1,610 citations, AI combines four major groups of sources: editorial content (~40%, the most stable category), user-generated content or UGC (Reddit, Instagram, TikTok, YouTube), brand sites and retailers, and press. The weight of each group changes depending on the country and the engine: ChatGPT relies heavily on Reddit, while Google AI Overview prioritizes Instagram and TikTok.
It is each URL that an AI engine shows or references when building its response. When ChatGPT, Perplexity, AI Overview, or Gemini respond to a query, they often rely on specific pages. Each of those pages is a citation. We analyzed 1,610 in total.
The exact numbers vary by vertical, but the structural patterns — the weight of the editorial, the social segmentation by engine, the importance of localization — hold true across most consumer niches. The recommendation is always the same: measure your own market before investing.
Yes, absolutely. Spanish still accounts for the majority of sources (between 49% and 65% depending on the country). The point is that producing only in Spanish leaves a quarter of the territory covered by technical content in English unaddressed. Ideally, both should be covered.
You need an AI visibility tracking tool that monitors your presence within generative responses, not in traditional search engines. CreceRank was designed for that, focusing on Spanish-speaking markets. You can start with a 14-day free trial without a credit card.
CreceRank tracks exactly how your brand appears in ChatGPT, Perplexity, Google AI Overview, and Gemini, with data from LATAM and Spain. What we did with running, do it with your sector.
No credit card required · Prompts in Spanish · Data from Spain and LATAM
