# 公开搜索发现恢复计划（2026-08-18）

生成时间：2026-08-18T05:23:21+08:00

机器记录：`/data/public-search-discovery-recovery-plan-20260818.json`

HTML：`/reports/public-search-discovery-recovery-plan-20260818.html`

状态：`recovery-plan-ready-d1-completed-no-positive-discovery`

## 当前结论

D1 proves execution discipline, not discovery: the target article is crawlable and D1 is fully recorded, but public search did not return the exact/canonical target URL and AI adoption remains unproven.

当前可声明：

- 公开站点上线：可以
- HTTP / crawler 可读：可以

当前不能声明：

- 搜索发现：不能
- AI 采用：不能
- 自然推荐：不能
- 模型认知改变：不能
- 客户付费发布结果：不能

## 诊断

| ID | 发现 | 证据 | 含义 |
| --- | --- | --- | --- |
| diag-01-crawlability-not-main-blocker | Crawler policy and HTTP access are not the current main blocker. | Production full check 282/282; crawler User-Agent matrix 126/126; robots explicit crawler rules 17. | Further work should prioritize discovery signals and external/source routes, not only more static files. |
| diag-02-search-discovery-missing | The target owned URL has not been observed in sampled public search results. | D1 25/25 recorded; S2/S3 0/0; positiveRows 0. | The article must remain classified as crawlable-not-indexed / searchDiscovery not-observed until exact/canonical URL evidence appears. |
| diag-03-blocked-search-surfaces-not-negative-proof | A large share of D1 rows were blocked by captcha/verification surfaces. | Blocked rows 15; valid non-positive rows 10; ambiguous rows 1. | Blocked rows should not be interpreted as not-found; they require later manual/browser or alternative-surface retesting. |
| diag-04-ai-adoption-separate-from-search | Search discovery and AI answer adoption remain separate evidence chains. | Window-pitfall AI retest 0/16; next row requires explicit authorization phrase. | Do not claim model cognition change or AI recommendation until raw AI answers and screenshots are recorded. |

## 恢复假设

| ID | 可能性 | 影响 | 理由 | 验证方式 |
| --- | --- | --- | --- | --- |
| h1-new-domain-and-weak-link-graph | high | high | The pages.dev site is technically readable but the target URL is not found; current evidence shows existing brand signals mainly on winsmo.com and external media, not on the new target URL. | D3/D7/D14 public search retests should show whether discovery improves without external writes; external source route tests should separately measure the value of official-site and media links. |
| h2-search-surface-blocking | high | medium | 15/25 D1 rows are blocked, so the negative sample is incomplete on some search engines. | Retest blocked rows with fresh browser sessions and preserve screenshots; do not count blocked rows as not-found. |
| h3-article-needs-more-query-specific-entry-points | medium | medium | The target article is one page; search discovery may improve if the question cluster has supporting pages, hub links, and concise machine-readable summaries. | Add or link supporting summaries only with evidence boundaries, then monitor whether site: and exact-slug queries surface. |
| h4-external-source-route-needed-before-natural-ai-recommendation | medium | high | Prior evidence indicates existing external sources and official site are discoverable; AI platforms may rely on search/source routes rather than voluntarily reading a standalone llms.txt. | After explicit authorization, test an official-site news route or paid media route separately from the owned pages.dev route. |

## 恢复动作

| 层级 | 状态 | 动作 | 是否需要外部写授权 |
| --- | --- | --- | --- |
| owned-site-machine-readable | done-this-turn | Keep D1 current state synchronized across llms.txt, AI crawl index, RSS, sitemap, D1 gates and objective audit. | no |
| owned-site-discovery | done-this-turn | Add the query-specific hub/summary links for the target article and ensure sitemap/feed/llms/AI crawl index point to the same canonical target. | no |
| public-search-retest | scheduled | Run remaining D3/D7/D14 rows; classify blocked, not-found, exact-target, canonical-equivalent and unsafe-snippet separately. | no |
| official-source-route | requires-explicit-authorization | If the user authorizes official-site work, publish a new evidence-bounded summary on winsmo.com or doctorlucheng.com and link to the owned evidence page. | yes |
| ai-platform-answer | requires-explicit-authorization | Run the next logged-in AI platform test for the target article and save raw answer, screenshot, time, prompt and result gate. | no |

## 后续复测计划

| runId | 日期 | 行数 | 目的 |
| --- | --- | ---: | --- |
| d3-public-search-retest | 2026-08-21 | 25 | Check whether the target owned URL appears after a short crawl/indexing delay. |
| d7-public-search-retest | 2026-08-25 | 25 | Check whether discovery improves over one week and whether blocked surfaces can be retested. |
| d14-distribution-comparison | 2026-09-01 | 25 | Compare owned-site-only discovery against any authorized external/official source distribution route. |

## 边界

- This recovery plan does not prove search indexing.
- This recovery plan does not prove AI platform adoption or model training ingestion.
- Do not claim natural recommendation, stable recommendation, model cognition change, ranking change, customer conversion or paid publishing outcome.
- Do not operate logged-in AI platforms, official site CMS, search-console tools, external media, DNS, ICP, D1/KV/Workers or payment systems without separate explicit authorization.