Most “hallucination” complaints in retrieval-augmented systems are retrieval failures. If the wrong chunk, an unauthorized chunk, or a stale chunk reaches the model, fluent answers will still be wrong. Fix evaluation around retrieval before you swap models.

01

Citation checks

Require answers to point at retrieved passages. Spot-check whether the cited text actually supports the claim. If citations are missing or ornamental, treat the run as a fail even when the prose sounds right.

Track citation precision over time. A model upgrade that drops citation honesty is a regression.

02

Golden questions

Build a small set of real questions with known-good document IDs. Score whether the retriever returns those docs in the top-k before you score the generated answer.

Include negatives: questions that should return nothing, or should refuse when docs are insufficient.

Where RAG quality usually breaks
03

Permission boundaries

Filter retrieval by the user’s ACL before ranking. Never retrieve then redact in the prompt and hope the model complies. Cross-tenant or cross-team leakage is a product incident, not an accuracy issue.

Test with users who should see nothing for a given corpus. Empty results are the correct answer.

04

Stale docs

Index freshness, document owners, and deprecation flags belong in the pipeline. A perfect embedding of last year’s policy is still a wrong answer.

When source systems change, re-ingest on a schedule you can explain—and show “as of” dates in the UI when it matters.

05

Order of work

Instrument retrieval metrics first. Then tune chunking, embeddings, and hybrid search. Only then compare generators on the same retrieved context.

If retrieval is wrong, a better model mostly writes a better wrong answer.

Sources

  1. Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks Lewis et al. (arXiv), 2020
  2. Evaluating RAG applications LangSmith / LangChain docs, 2025
NEXT ARTICLEOnly 5% get AI value at scale.