AI Agent Cites Links That Are Dead or Moved: Causes and Fixes

IssueCitation AccuracyFrequencyOccasionalCategoryKnowledge RetrievalPublishedView source on GitHub ↗

Issue: AI agent’s citations point to documents that have been moved, deleted, or made inaccessible since they were indexed.

Frequency: Occasional

Symptoms

  • Citation URLs return 404
  • Document has been moved or deleted
  • Permission denied when following citation
  • Link format incorrect
  • Commonly reported in RAG pipelines (LangChain, LlamaIndex) that index document URLs without periodically re-validating them

Root Cause

  • Documents moved after indexing
  • URLs changed without index update
  • Permission changes not reflected
  • Temporary documents cited

Example

Agent response: "See the policy document [1]"

Citation [1]: https://internal.company.com/docs/policy-v2.pdf

User clicks: "404 Not Found - This page doesn't exist"

Reality: Policy renamed to policy-v3.pdf last month

Result: User can't access cited information

Mitigation Strategies

  1. Link validation: Check URLs are accessible before citing
  2. Canonical URLs: Use permanent, stable identifiers
  3. Freshness checks: Verify links periodically
  4. Fallback content: Include enough context that link isn’t critical
  5. Version tracking: Detect when documents are updated/moved
  6. Archive links: Maintain accessible copies of cited content

Detection

  • Automated link checking
  • Track 404 rates on citations
  • Monitor user reports of broken links
  • Regular crawl of cited URLs

How to fix it: validate citation links before serving them and use stable, canonical identifiers instead of URLs that shift — see Mitigation Strategies below.

Mitigations

Prevention

  1. Implement query-answer consistency validation: Decompose complex queries into atomic components and verify each component is addressed in the answer before returning. Use RAGAS Answer Relevancy metric (target: >0.75) with automatic re-generation for scores below threshold. Root cause mitigation: Prevents context-anchoring by explicitly binding answer generation to parsed query intent.

  2. Apply multi-source consensus verification: Require answers synthesized from multiple sources to explicitly cite which sources support each claim and flag unresolved contradictions. Implementation: Use semantic similarity checks across source fragments to detect cherry-picked evidence patterns. Root cause: Ensures balanced representation of evidence when sources conflict.

  3. Enforce comprehensive coverage checks: Implement structured extraction requiring explicit treatment of caveats, limitations, exceptions, and counterevidence for each claim type. Use template responses with mandatory caveat sections. Root cause: Prevents omission of qualifying information that would change user decision-making.

Detection & Response

  1. Answer completeness monitoring: Measure coverage of query intents in generated answers. Track query decomposition rate (% of query components explicitly addressed) and flag responses with coverage <85%. Instrument RAG pipeline to log query-answer similarity scores per component. Alert on sustained scores <0.70.

  2. Evidence balance scoring: For each answer, compute evidence distribution across sources and flag one-sided responses (>70% from single source on multi-source queries). Implement automated extraction of caveat/limitation mentions and track inclusion rates by query type. Target: >80% of medical/financial answers include relevant caveats.

Architecture Patterns

  1. Query Intent Decomposition Graph: Parse complex queries into a DAG of atomic intents before retrieval. Each retrieved document is mapped to specific intent nodes. Answer generation must satisfy all leaf nodes. Validation layer computes coverage before response generation.

  2. Evidence Consensus Engine: Maintain a fact graph where each claim is attributed to specific sources with confidence scores. Multi-source claims require consensus computation (intersection of sources supporting claim). Flagging layer surfaces contradictions to generation model.

  3. Structured Response Templates: Use task-specific response schemas that enforce inclusion of: primary answer, supporting evidence, relevant caveats/exceptions, alternative interpretations, confidence bounds. Auto-flag template violations before user delivery.

Key Metrics

MetricTargetAlert ThresholdMeasurement Method
Answer Relevancy Score>0.75<0.70RAGAS metric on generated answers vs. original query
Query Coverage Rate>90%<85%Percentage of query components explicitly addressed in answer
Evidence Balance Index>0.6<0.4Distribution of citations across sources (Gini coefficient, 0=balanced, 1=single-source)
Caveat Inclusion Rate>80%<70%Percentage of medical/financial answers including relevant limitations
User Clarification Rate<5%>10%Percentage of answered queries requiring follow-up clarification

Alerts & Escalation

AlertConditionSeverityResponse
Low Answer RelevancyAnswer Relevancy Score < 0.70 for >5% of queries in 1-hour windowHIGHPage on-call; trigger re-generation with query reinforcement prompt
Single-Source DominanceEvidence Balance Index < 0.4 on multi-source queries for >3 consecutive queriesMEDIUMLog event; audit cherry-picking patterns in retrieval/synthesis
Rising Clarification DemandUser Clarification Rate exceeds 10% (vs. 5% baseline) over 24-hour windowHIGHInvestigate query decomposition or answer template effectiveness

References

Related Patterns