Fix GEO optimizer project-site AI discovery false negatives

[Copilot]

The GEO optimizer audit was reporting missing llms.txt and AI discovery files because scanner findings did not account for the /gh-aw/ GitHub Pages project-site base path. Those false negatives caused duplicate follow-up issues for files that are already live.

• Project-site verification

  • Adds authoritative HTTP checks for project-scoped AI discovery URLs:

    https://github.github.com/gh-aw/llms.txt
    https://github.github.com/gh-aw/.well-known/ai.txt
    https://github.github.com/gh-aw/ai/summary.json
    https://github.github.com/gh-aw/ai/faq.json
    https://github.github.com/gh-aw/ai/service.json

  • Persists verification metadata and successful response bodies in the GEO optimizer artifact.

• False-negative handling

  • Updates the audit instructions to treat project-site verification as authoritative for llms.txt and AI discovery findings.
  • Prevents missing-file recommendations from being selected when the project-site verification succeeds.

• Workflow output

  • Adds a structured docs-ai-discovery-verification.json summary:

    {
      "base_url": "https://github.github.com/gh-aw",
      "summary": {
        "llms_txt_found": true,
        "ai_discovery_found_count": 4,
        "ai_discovery_total": 4,
        "ai_discovery_all_found": true
      }
    }

  • Regenerates the compiled workflow lock file.

• Fixes [deep-report] GEO Optimizer scanner reports false 0-scores for llms.txt/AI-discovery files that are live and working #53787

[Copilot]

Pull request overview

Adds project-scoped verification to prevent GEO audit false negatives for existing AI discovery files.

Changes:

• Verifies and preserves AI discovery responses.
• Updates audit guidance to prioritize verification results.
• Regenerates the compiled workflow.

Show a summary per file

File	Description
.github/workflows/daily-geo-optimizer.md	Adds verification and audit instructions.
.github/workflows/daily-geo-optimizer.lock.yml	Regenerates the compiled workflow.

Review details

💡 Add a code-review agent skill for context-aware, tailored reviews. Learn more in the docs.

Suppressed comments (2)

.github/workflows/daily-geo-optimizer.md:284

• The aggregate ai_discovery_all_found gate only handles the all-or-nothing case. If three authoritative checks succeed and one fails, the scanner's blanket “no AI discovery signals” result can still be repeated for the three verified endpoints in the discussion. Evaluate each entry in checks so partial availability does not recreate the false negatives this change is intended to remove.

verification reports `summary.llms_txt_found: true`, do not report llms.txt as missing. When it
reports `summary.ai_discovery_all_found: true`, do not report AI discovery files as missing.

.github/workflows/daily-geo-optimizer.md:373

• In a partial result, “found the file or files” can discard the entire recommendation after any named endpoint is found even when another named endpoint genuinely failed verification. Filter the recommendation per endpoint, and select the next recommendation only when every missing-file claim was disproved.

If the audit's highest-impact recommendation says documentation-site robots.txt, llms.txt, or AI
discovery files are missing, first compare it against the authoritative verification files above.
If the project-site verification found the file or files, treat that recommendation as a scanner
false negative and select the next actionable recommendation instead.

• Files reviewed: 2/2 changed files
• Comments generated: 1
• Review effort level: Balanced

[github-actions]

🎉 This pull request is included in a new release.

Release: v0.87.1