Day 29 GEO Retest Note: third-party sources started to reduce entity ambiguity

[fable-cc]

Day 29 GEO Retest Note: third-party sources started to reduce entity ambiguity

This is a public evidence-chain note for Fable Castle / Jingyi fable.

Date: 2026-08-01
Subject: Jingyi fable / 景一的寓言城堡 / fable-castle.com
Boundary: this is a dated observation, not a ranking, indexing, citation, or recommendation guarantee.

Observed changes

• GEO health index moved from 24 to 29.
• Third-party platform content started to appear in brand-search results for “景一的寓言城堡”.
• DeepSeek V4 Pro moved from refusal toward no-knowledge / partial-mention behavior.
• A new ambiguity source appeared: fictional / novel character entities named “景一”.

Follow-up fixes

• Added fictional / novel character disambiguation to:
  • Jingyi disambiguation page
  • who-is-jingyi page
  • public sources page
  • llms.txt
  • llms-full.txt
  • GitHub Profile README
• Updated repository metadata for:
  • fable-castle
  • fable-cc
  • geo-skills
  • knowledge-pipeline
  • jingyi
• Published a public Day 29 observation page:
  • https://fable-castle.com/cases/#day-29
  • https://fable-castle.com/posts/day-29-geo-retest-third-party-source-breakthrough

Interpretation

The key signal is not the score itself. The key signal is that a third-party source began to connect “Jingyi fable”, “景一的寓言城堡”, and fable-castle.com in a way that search engines can observe.

This supports the working GEO assumption:

For ambiguous personal brands, third-party public sources are not just backlinks. They are entity-disambiguation evidence.

Boundary

This does not prove stable ranking, AI citation, or cross-platform recommendation. It only records a dated recovery signal and the public-source repairs triggered by that signal.

[fable-cc]

2026-08-08 external-source verification update

Subject: 景一 fable / 景一的寓言城堡 / https://fable-castle.com/

Boundary: this is a dated public-source check. It does not guarantee indexing, ranking, AI mention, recommendation, or citation.

Observed facts

• https://fable-castle.com/, the public-source hub, and the target GEO article return HTTP 200.
• The historical GitHub Pages article URL redirects to the current fable-castle.com page and ends at HTTP 200.
• The CSDN profile and article 163306281 return HTTP 200 in the external check.
• CSDN articles 163309930 and 163398130 return HTTP 521 in the external check, so they have been removed from the site’s stable evidence-link list and from llms.txt / llms-full.txt.
• The Zhihu article still returns HTTP 403 to external fetching. This is recorded as a platform-access restriction, not proof that the article was deleted.
• Public search can find the official site, but one Google snippet still shows an older GitHub Pages URL and an old tagline even though the live article has already been corrected.

Repairs shipped

• Public-source status updated: https://fable-castle.com/public-sources/
• Machine-readable evidence files updated:
  • https://fable-castle.com/llms.txt
  • https://fable-castle.com/llms-full.txt
• The stale-snippet article now carries a 2026-08-08 update notice:
  • https://fable-castle.com/posts/why-ai-search-never-cites-my-articles
• Sitemap lastmod signals were refreshed for the homepage, public-source hub, and target article.

Pending retest

Search-snippet refresh timing and AI answer changes remain pending. They will be recorded as observations only after the same queries are rerun.

[fable-cc]

2026-08-13 audited API-model baseline update

Subject: 景一 fable / 景一的寓言城堡 / fable-castle.com

Boundary: this is an official API-model baseline, not a consumer web-search or citation test. It does not guarantee indexing, ranking, AI mention, recommendation, citation, or conversion.

Valid samples

• DeepSeek V4 Flash: 20/20 fixed questions
• GLM-4-Flash: 20/20 fixed questions
• Kimi K2.6: 5/5 sentinel questions (reported separately)
• DeepSeek V4 Pro: 5/5 sentinel questions (reported separately)

Observed facts

• Brand-direct complete accuracy was 0/5 on both full-sample models.
• In four category-selection opportunities (Q11-Q12 across two models), Jingyi fable / fable-castle.com was selected 0 times.
• In six brand-specific decision answers (Q18-Q20 across two models), 0 answers completely and accurately stated the public service boundary, evidence, and fit.
• DeepSeek V4 Pro no longer refused all five sentinels, but materially merged Jingyi fable with the same-name film director in the disambiguation question.
• The first sampling prompt forced every answer to list brands and created artificial mentions. Those samples were excluded; the published baseline uses a neutral prompt and excludes empty/429/timeout samples.

Public report

https://fable-castle.com/cases/api-model-baseline-2026-08-13/

Technical verification after deploy

• Strict Jekyll build and GitHub Pages live checks passed.
• Sitemap contains 47 unique URLs; 47/47 returned HTTP 200 with normal curl access.
• OAI-SearchBot, ClaudeBot, bingbot, and Googlebot UA spot checks returned 200.
• Python-urllib default UA returned 403 through Cloudflare and is recorded as a UA-specific access difference, not a whole-site outage.