Yes. According to a Weglot and Ellipsis study published November 2026, websites covering a topic in both English and Spanish earned a 98% higher AI citation rate than websites covering it in only one language, averaged across Google AI Overviews, ChatGPT, and Claude.
What the Study Actually Measured
Weglot partnered with AI search agency Ellipsis to analyze 1,136,972 citations drawn from 116,574 usable AI responses. The researchers took a panel of websites operating in Spain and Mexico, turned each site's ranking keywords into equivalent questions in both Spanish and English, and ran all 11,664 prompts through Google AI Overviews, ChatGPT, and Claude. They then checked whether each cited website actually covered the relevant topic in both languages, not just whether a language switcher existed somewhere on the site.
That distinction matters more than almost anything else in this dataset.
The Headline Numbers, by Platform
The overall 98% lift obscures meaningful platform variation. Here is what the study found for bilingual versus single-language topic coverage:
| Platform | Single-language citation rate | Bilingual citation rate | Multiplier |
|---|---|---|---|
| Google AI Overviews | 17.3% | 31.9% | ~1.8x |
| ChatGPT | 6.5% | 19.4% | ~3x |
| Claude | 10.9% | 17.6% | ~1.6x |
ChatGPT shows the sharpest gap by a wide margin. If your AEO budget is limited, the Weglot study suggests ChatGPT is where multilingual coverage pays off most.
A Multilingual Site Is Not the Same as Multilingual Coverage
This is the finding most likely to be misread in the wild. Websites with a multilingual setup but only a handful of translated pages showed no overall AI citation advantage. ChatGPT does not cite a site on a topic because the site has a language switcher. It cites the site because that specific topic is answered in the language of the query.
That framing has real consequences for how you prioritize translation work. A homepage and a few product pages translated into Spanish will not move your citation rate if the blog post answering the customer's actual question exists only in English.
Why AI Models Prefer Same-Language Sources
The study also checked the language of pages that AI systems actually linked to. Among identifiable links, 95% of citations from Spanish-language questions led to Spanish-language pages. For English questions, the figure was 77% leading to English-language content.
Large language models are technically multilingual, but they clearly favor sources in the query's own language when one is available. The mechanism is straightforward: the model retrieves content most relevant to the query, and language match is a strong relevance signal.
One result from the study is worth sitting with separately. On ChatGPT, for questions asked in Spanish, websites that also had English coverage were cited 2.5 times as often as Spanish-only websites (21.8% versus 8.7%), even though both groups had Spanish content. Bilingual coverage appears to signal something about authority or depth that influences citation even when language match is not the differentiating factor.
What This Means for Your AEO Work
For marketers and founders doing AEO work, this study offers a concrete, underused lever. Most English-language AEO guides focus on schema markup, answer formatting, and E-E-A-T signals. Multilingual coverage rarely appears on those lists, yet this dataset puts it on par with, or ahead of, many content-quality optimizations.
A practical starting sequence:
- List the questions your customers ask about your products or category.
- Check which pages answering those questions already exist in each language you target.
- Fill gaps on your highest-priority topics before translating lower-priority pages.
- Track citation rates per language using a tool like Otterly or PromptWatch, which monitor AI mentions and citations over time.
For monitoring multilingual share of voice across platforms, Athena HQ and OmniSEO can segment citation data by platform, which is useful given the wide variation in multipliers across Google AI Overviews, ChatGPT, and Claude.
Study Caveats Worth Acknowledging
The study covered one language pair: English and Spanish, in Spain and Mexico. The results may not translate directly to other language pairs or regions. Domain authority controls are not described in detail in the published methodology, so it is possible that bilingual sites in the panel skew toward higher-authority domains. The query set was derived from each website's own ranking keywords, which could introduce selection effects.
The Weglot study also notes that its 2025 research found up to 327% higher visibility for second-language searches using a different methodology (website-level translation labels rather than topic-level coverage checks). The 2026 methodology is more precise, and the authors interpret the shift as evidence that AI platforms are getting better at evaluating specific content rather than site-wide signals.
More language pairs, more query categories, and more explicit controls for domain authority would strengthen the conclusions. Until that data exists, the English-Spanish finding is the best number available, and 98% is not a result you ignore while waiting for perfect data.