Kechuang No.1
On 15 June 2026, Kechuang No.1 was mentioned in 15/15 fixed Yuanbao questions and first-recommended in 12/15. Raw answers and screenshots are archived.
Internal archive: fixed 15-question set, Yuanbao JSON answers and 15 screenshots from 15 June 2026. Client screenshots are not published without approval.
Client
Kechuang No.1 is a science park services operator and a client of the AI growth (China) line, delivered as a GEO advisory engagement.
The problem
Buyers increasingly put the question to an AI assistant instead of opening a search engine and working through links. The problem was never the website ranking — it was whether a model answering directly would mention the operator at all, and whether it would describe it correctly.
What we did
The standard GEO sequence: define a reproducible test question set and measure the pre-optimisation baseline; work on entity consistency, quotable factual content and third-party corroboration; then re-run the same question set and compare.
Limits of this case
- The 100% / 80% figures apply only to Yuanbao on 15 June 2026 and a fixed set of 15 questions. They do not describe Doubao, DeepSeek, Kimi or ChatGPT.
- A later 13 August two-platform question-level union was 80% mention (12/15) and 60% first recommendation (9/15), showing why the June peak must carry its date.
- Models keep changing. GEO results have to be re-tested on a cycle; it is not a one-off project.