AI search favors local domains across 10 markets

What happened

Per Search Engine Journal, an analysis by Aleyda Solis using Similarweb data across 10 markets shows that AI search clicks disproportionately flow to local domains rather than global ones, with the split varying significantly by industry.

The finding upends a common assumption inside global marketing teams: that an English-language .com property, well-optimized and authoritative, will absorb most AI-driven traffic regardless of where the user sits. It does not. ChatGPT, Perplexity, and Google's AI surfaces are routing users to country-specific domains at rates that vary by sector, with finance, retail, and media showing some of the strongest local-domain bias.

Solis's read is that LLMs are treating ccTLDs and localized subdirectories as proxies for relevance and trust within a given market. That has direct consequences for any brand running a hub-and-spoke domain strategy.

Why it matters for your brand

If you are a CMO at a global bank, an industrial group, or a multilateral, your default content strategy probably consolidates authority on one domain. The logic was sound for classic SEO: concentrate backlinks, avoid duplicate content, push everything through the mothership. AI search breaks that logic.

For financial services, this is the most expensive version of the problem. A wealth management arm with a single .com running localized subfolders may find that ChatGPT, when asked about retirement products in Spain or Germany, is citing local competitors operating on .es and .de domains. The user never sees the global brand. The cited answer becomes the answer. Brand equity built over decades fails to convert into LLM visibility because the model is making a locality judgment the brand never optimized for.

For multilaterals and UN system bodies, the implication is different but sharper. Institutions like UNDRR, CGAP, or WHO publish in English from .org domains and assume universal reach. If LLMs in Brazil, India, or Indonesia are routing policy queries to local government or local NGO domains, the multilateral's framing of an issue (disaster risk, financial inclusion, vaccination) loses to the local interpretation. That is a soft-power problem disguised as a traffic problem. The remedy is not translation alone; it is local domain presence, local citations, and content that LLMs can identify as native to the market.

For major industrial groups, the question is which markets justify a real local domain strategy versus a subfolder. HOLCIM-style operators with manufacturing footprints in 60 countries cannot stand up 60 ccTLDs. But they can identify the 10 markets where AI search now drives meaningful B2B procurement research and prioritize accordingly. The cost of being absent from the cited set in a top-five market is no longer a long-tail SEO concern; it is a deal-flow concern.

For philanthropic and policy institutions, the local-domain bias intersects with credibility. When a foundation funds research on agricultural development in Kenya, the LLM is more likely to cite a Kenyan academic or government domain than a US foundation's .org. Co-publishing with local institutions, ensuring the local partner's domain carries the work, may produce better LLM visibility than self-publishing on the funder's own site.

Content strategy implications are concrete. First, audit which of your local domains or subfolders are actually being cited in AI answers in their target market; most teams have never run this query set. Second, treat localized content as a first-class production track, not a translation afterthought; LLMs detect shallow localization. Third, reconsider redirect strategy: brands that 301 local domains into a global hub may be actively destroying the signals LLMs use to assign local relevance.

The signal in context

The Similarweb finding fits a pattern emerging across AI search research over 2024 and 2025: LLMs do not behave like Google. Google's algorithm spent two decades training marketers to consolidate. AI answer engines reward a different shape of presence, one where citation diversity, local trust signals, and topical specificity matter more than raw domain authority. Reddit citations surging in ChatGPT, Wikipedia's outsized role in Perplexity, and now local domains winning regional clicks are all variations of the same theme: the model is making relevance judgments using signals the SEO industry deprioritized.

The strategic question for senior marketers is whether their domain architecture, built for a search era that is ending, is now an active liability. For most global B2B brands the honest answer is yes, and the fix is neither cheap nor fast. It starts with knowing which of your domains the models actually cite, in which markets, for which queries. Brands that have that visibility now will spend the next 18 months reallocating content investment. Brands that do not will discover the gap when a regional sales lead asks why the AI assistant is recommending a competitor.