GEO Assumes Google. Half the Web Isn't.
Every "how to get cited by AI" guide I have read quietly assumes the pipeline ends at Google and ChatGPT. That assumption covers less of the web than the people making it tend to think.
Our directory has 765 skills. Of those, a meaningful cluster is SEO tooling — audits, keyword work, GEO content optimization. Sort that cluster by which search engine it targets, and you get Google, Google, Google, and then exactly one exception.
That exception is fire-your-seo-agency by leopard627, scored 8.1. It is bilingual Korean-English, and it covers Naver and NEO as first-class targets rather than footnotes. 8.1 is not a top score — the seranking family runs 8.6 to 9.0 and technical-seo-checker sits at 9.3 — but it is the only skill we have reviewed that treats a non-Google search ecosystem as the thing it is for.
It also carries one detail the higher-scoring Google skills mostly lack: re-test date discipline. The skill records when its ranking claims were last verified against the live engine. That habit matters more in Korea than it does against Google, and I'll come back to why.
Naver is not a Google translation problem
The reason a bilingual skill is rare isn't language. Translation is a solved problem for an agent — this skill wouldn't exist otherwise.
The reason is that Naver runs a genuinely different ranking pipeline, where content on Naver's own properties (blogs, cafes) historically dominates Naver's own results, and NEO — its AI search layer — inherits that substrate. An SEO skill that "supports Korean" by translating Google advice is worse than useless; it confidently answers the wrong question. What fire-your-seo-agency does instead is carry the engine's actual mechanics as separate knowledge, which is why it earned its score despite modest polish. And that knowledge expires: Naver's algorithms move on their own schedule, unrelated to Google's core updates. Hence the re-test dates. A skill that claims ranking advice without dating it isn't a skill, it's a screenshot of the past.
The gap to the west of Korea
Now extend the map. The Korean exception proves the category is possible. Then look at the Chinese internet — Baidu, WeChat search, and a set of AI assistants that are not ChatGPT and are not reached through it.
In 765 skills: nothing. Not a low score. Zero coverage.
The GEO discourse makes this invisible because of how the story is framed. Tools like geo-content-optimizer (aaron-he-zhu, 8.7) and ultimate-seo-geo (mykpono, 8.3, with its anti-fabrication rules and internal-versus-competitor modes) are built for a specific measurement surface: Google SERPs, ChatGPT citations, an open web that crawlers can reach. That surface is real — for some audiences. But WeChat articles live inside a walled garden the open web's crawlers don't index; Baidu weighs its own ecosystem properties the way Naver does, only differently; and the AI assistants answering questions in Chinese draw on a corpus and a set of index relationships that have no overlap with ChatGPT's. "Get cited by AI" is not one discipline. It's at least two, with different rules and different winners.
So the entire optimization stack — the schema advice, the citation-bait patterns, the llms.txt debates — silently carries a geo assumption that its mostly-anglophone authors never state, because from inside the anglophone internet it looks like a physical law rather than a locality.
Why the vacuum persists
The boring reason: skill authors write for the ecosystem they search in. The person with the domain knowledge to build a Baidu-competent agent skill is usually not the person publishing skills to an English-language directory, and the incentives haven't yet moved anyone to bridge it.
That makes this one of the few genuinely open territories in agent tooling. The Korean precedent shows the shape of the product: engine-specific mechanics as separate knowledge, not translated heuristics, plus dated claims because the substrate moves. Building the same thing for Baidu and WeChat — with the re-test discipline fire-your-seo-agency already models — is not a localization project. It's a from-scratch build with an existing template for how to stay honest.
The GEO narrative says AI search is consolidating the world's questions into one pipeline. The directory's own coverage map says otherwise: the pipelines are diverging, and the tooling has only crossed one of the borders. Whoever crosses the second one enters a market with no incumbents — including, currently, every skill we've scored.