KnowledgeReady
Prepare your knowledge base for AI
Find stale content, conflicting instructions, duplicate sources and retrieval risks before connecting your documentation to RAG, AI search or internal AI agents
The demo is a complete audit of a synthetic knowledge base. No sign-up needed.
security/customer-log-retention-policy.md
Customer Log Retention Policy
Customer application logs must be retained for 90 days in the primary logging cluster.
platform/logging-standard.md
Logging Standard for Platform Services
Customer application logs are retained for 14 days in the primary logging cluster to control storage cost.
An assistant asked how long logs are kept could answer 90 or 14 days, depending on which page it retrieves.
Example from the synthetic demo data
What KnowledgeReady finds
The documentation problems that make AI assistants give wrong or inconsistent answers, even when the retrieval pipeline itself works.
- Outdated content
- Pages last updated before a threshold you choose, which an assistant would treat as current.
- Missing owners and review dates
- Pages nobody is accountable for, and pages with no record of when they were last checked.
- Conflicting instructions
- Active pages that set different values or rules for the same thing, such as retention periods or limits.
- Duplicate knowledge
- Competing copies of the same procedure that split retrieval and drift apart over time.
- Broken links
- References to pages that are not in the export, so retrieved answers point nowhere.
- Poor document structure
- Skipped heading levels, generic headings and titles, and oversized or multi-topic pages.
- Context-dependent fragments
- Sections that rely on “as described above” and lose their meaning once split into chunks.
- Retrieval-unfriendly content
- Rules that live only inside tables, where chunked text loses which value applies to which case.
How it works
- 1
Upload knowledge base
A ZIP export of Markdown, HTML or text pages, for example from Confluence or Notion.
- 2
Structural and deterministic checks
Rules check metadata, freshness, links, headings, size and chunk independence on every page.
- 3
Semantic risk verification
Likely duplicates and contradictions are selected, then verified before they are reported.
- 4
Prioritized report
Findings ranked by severity, with the reason each one matters for AI retrieval and how to fix it.
KnowledgeReady v0.4
Blind synthetic Benchmark #4
80 synthetic documents. Synthetic benchmark.
- Precision
- 100%
- Recall
- 95.7%
- F1
- 97.8%
Measured on a synthetic benchmark, not on customer data. It is not a guarantee of accuracy on real knowledge bases or of how an AI assistant will answer.
KnowledgeReady performs structural analysis within the application. For semantic verification, selected document excerpts may be sent to the configured AI provider. Do not upload information you are not authorized to process. Privacy
View Demo Audit