AI and automation

What is retrieval augmented generation?

Also known as: RAG

Published September 10, 2026 · Reviewed by the Fidiora team

Definition

Retrieval augmented generation is a technique where a system searches a document collection for passages relevant to a question, then supplies those passages to a language model as context before it answers. The model composes its response from retrieved material rather than from memory alone.

Why it matters

RAG is what makes AI support answers checkable. Without it a model answers from training data that may be outdated, generic, or simply wrong about your product. With it, the answer traces back to a specific document you control and can correct, which is the difference between a demo and a deployable system.

What to know

  • Retrieval quality sets the ceiling. If the right passage is not retrieved, no amount of model quality saves the answer.
  • Chunking strategy quietly determines accuracy. Passages split mid-procedure produce confidently incomplete answers.
  • Stale sources are the most common production failure, and they fail silently because the answer still sounds right.
  • Citing the source document lets both customers and agents verify, which converts trust from a claim into a check.
  • Contradictory documents are worse than missing ones, because the system picks one and sounds certain.

An example

Example: a customer asks about the refund window. The system retrieves the current refund policy page, hands that text to the model, and the answer quotes fourteen days because that is what the page says. Change the page and the answer changes with it.

Common mistakes

  • Indexing outdated documentation alongside current documentation
  • Chunking documents so procedures are split across passages
  • Not surfacing the source, which removes the ability to verify
  • Assuming a better model fixes a retrieval problem
Where Fidiora fits

Fidiora grounds every answer in your help centre, website, and past tickets rather than in general model knowledge. When your content cannot answer a question, it says so and flags the gap instead of improvising, which is how documentation quality actually improves over time.

See Pricing
FAQ

Questions

Does RAG stop AI hallucination?
It reduces it substantially but does not eliminate it. A model given the right passage still occasionally overstates or fills a gap. Grounding plus an explicit I do not know path plus source citation is the combination that holds up.
What content should feed a support RAG system?
Current help centre articles, product documentation, policy pages, and high-quality resolved tickets. Audit for contradictions and remove outdated pages first, because the system cannot tell which of two conflicting answers is current.
How often should the index be refreshed?
Whenever the underlying content changes. A daily refresh is a reasonable floor, and anything policy-related should update immediately, because a stale policy answer is the most expensive kind of error.

Related terms

Grounding Grounding is the practice of restricting an AI system to answer only from a defined set of verified sources, such as you... Hallucination A hallucination is an AI-generated statement that is fluent and confident but not supported by any real source. In custo... Knowledge base A knowledge base is a structured collection of articles that answer customer questions about a product or service. It se... Large language model A large language model is a machine learning system trained on very large text collections to predict likely continuatio...
Get Started

See Fidiora resolve a ticket in 60 seconds.

No credit card, no sales call required. Connect your docs and watch it work.

Start Free Book a 15-min Demo

No credit card · Live in under an hour