Greater China generative engine optimization study 2026: most enterprise content never gets cited by any AI (Original research)

We presented a report recently to a marketing director at a publicly listed manufacturer. They'd published 430 technical articles, white papers, and press releases over two years. Website traffic was solid. When we tested 200 questions their customers ask against ChatGPT, Perplexity, Gemini, and Claude separately, the outcome was unambiguous: none of those 430 pieces appeared in any answer.

Not three times. Not once. Zero.

This pattern is common across the market. Over six months, we surveyed 240 companies across Taiwan, Hong Kong, and Singapore. We analyzed 18,600 publicly available pieces of content and tested them using 1,200 industry-specific queries against five major generative engines. We tracked which content got cited and how often.

Greater China GEO study: core findings

Only 8.3% of enterprise content we surveyed in Greater China has ever been cited by any major AI engine. 61% of companies have zero citations across all queries.

More than six out of ten companies are invisible in AI-generated answers, regardless of how much content they produce. The search engine era was different. Then, enough content plus tight keyword targeting typically captured some long-tail traffic. Generative engine optimization has no consolation prize. Most AI responses cite three to five sources. If your content is not on that list, you do not exist.

We examined industry patterns, and the gaps are significant:

Financial services and healthcare rank highest. This is not because they understand GEO better. Regulations in these sectors require clear structural information: definitions, data, sources. These are exactly what AI prefers to cite. Manufacturing and logistics rank lowest because their content is mostly product specs and marketing copy. Presentation is heavy, facts are sparse. AI cannot extract usable information from it.

What cited content actually looks like

We reviewed the 8.3% of content that did get cited and found consistent patterns. They have nothing to do with writing quality.

Cited content leads with the answer. One sentence states what something is, how much it costs, and why it matters. No three-paragraph setup before the point. It contains verifiable facts and dates. AI extracts sentences that function as statements of fact. The structure is clear. Headlines ask questions. Paragraphs answer them. No semantic ambiguity. The same fact appears on your site, Wikipedia, and industry publications. AI trusts information it sees corroborated.

Content that receives no citations typically fails the same way: pages full of adjectives, 'industry-leading,' 'breakthrough innovation,' 'trusted by', but no sentence that stands as a fact. AI cites statements it can back up, not claims it can admire.

What these numbers mean for you

To determine your position, test your own content. List 30 questions your customers ask AI most often. Run them one by one. Check the responses for your company. Marketing directors who do this usually move from surprise to focus. Seeing your content's actual value in front of AI changes how you approach the work.

We approach GEO differently from standard content marketing. We establish the citation baseline first, then restructure existing content for clarity and fact density. In GEO, presentation quality does not determine results. Only this matters: AI actually cites you, and people find you because of it.