AI SEO: ChatGPT loses ground in Spain, schema markup gains it

ChatGPT’s share of AI-driven traffic in Spain has fallen from 81.44% to 70.49% in just over a year, while Gemini has jumped from 4.39% to 12.41%, according to the latest SE Ranking analysis covering January 2025 to April 2026. At Evicron, an AI and custom software studio based in Barcelona, we read this as confirmation of something our visibility audits had already been showing clients: which assistant cites your business is shifting fast, and betting everything on optimizing for a single one is a mistake. The good news is there’s a technical lever — structured data — that now has solid research behind what it actually delivers.

AI traffic in Spain is being redistributed

AI-driven traffic is still small in absolute terms — in 2025 it averaged just 0.23% of total web traffic in Spain — but its internal composition has shifted more in a year than the classic search market moved in a decade. Gemini is the clearest case: its share grew 274% year over year and it has already overtaken Perplexity as the second AI platform by traffic in Spain. ChatGPT remains, by a wide margin, the main reference, but losing 11 points of share in a single year is a signal no business relying on generative-AI visibility should ignore. Content built purely for “what ChatGPT reads” is starting to leave traffic on the table in Gemini and in Google’s AI Mode, already live in Spain since late 2025.

For a small or midsize business, the practical takeaway is simple: measuring visibility in a single assistant is no longer enough. If your content strategy and your structured data are designed exclusively for ChatGPT, you’re optimizing for a shrinking share, not a growing one.

Structured data: the signal you can actually control

This is where schema markup comes in — the code that explicitly describes what a page is (an article, an FAQ block, a business listing) so a crawler, human or AI, doesn’t have to guess. Several 2026 studies agree on the correlation: content with structured data is roughly 2.5 times more likely to appear cited in an AI response, and recent analyses measure that 65% of Google AI Mode citations and 71% of ChatGPT citations come from pages with schema in place.

That figure deserves a caveat before anyone invests blindly. A controlled experiment by OtterlyAI, presented at the BrightonSEO conference in April 2026, added schema to over 2,000 URLs on a single site without touching the content: citations in Google AI Overviews rose 1,500% and in AI Mode 377%, but fell in ChatGPT, Gemini and Copilot, with no movement in Perplexity. The authors’ own conclusion is that schema correlates with citation but doesn’t cause it on its own across every platform equally: it works as an additional signal layered on top of content quality and entity authority, not as a universal shortcut. An independent academic study published on SSRN ranks structured data as the third strongest factor associated with AI citation (+39%), behind content freshness and metadata (+47%) and semantic HTML (+42%).

Which schema types generative AI actually prioritizes

Not all structured data performs equally. An analysis of citations by schema type puts BreadcrumbList (page navigation trail) at a 46.2% citation rate and FAQPage (question-and-answer blocks) at 45.6%, well ahead of more generic formats. For a services business, that translates into two concrete priorities: structuring site navigation properly (so every page declares where it sits within the site) and turning real customer questions into schema-marked FAQ blocks instead of burying them in body copy.

What to do this week

1. Audit which assistant actually cites you today — not just ChatGPT: run the same queries through Gemini and Google’s AI Mode and compare which sources show up.
2. Implement FAQPage and BreadcrumbList on the pages that already get search traffic before rolling out other schema types: they’re the formats with the best measured citation rate.
3. Don’t treat schema as an isolated shortcut. Pair it with fresh content and clear entities — your business name, location and service stated in the opening paragraph — which the academic study above found to matter more.
4. Revisit the GEO fundamentals we covered in our guide on getting AI to recommend your business: they remain the starting condition.

How we approach this at Evicron

In our SEO and web positioning for AI projects, we start by auditing which assistant actually cites each client — we don’t assume it — and prioritize the schema types that perform best for that specific case instead of rolling out a generic structured-data template. It’s a small difference on paper that, with this year’s data, is starting to show up in citations.

Want to know if your website is ready for the coming redistribution of AI traffic? Get in touch: we start with a real visibility audit across the main assistants, not a generic best-practices checklist.