AI and automation

What is grounding?

Also known as: knowledge grounding

Published September 10, 2026 · Reviewed by the Fidiora team

Definition

Grounding is the practice of restricting an AI system to answer only from a defined set of verified sources, such as your help centre and product documentation, rather than from general model knowledge. A grounded system either finds support for its answer in your content or declines to answer.

Why it matters

Grounding is the control that makes AI support auditable. An ungrounded model will answer a policy question with something plausible drawn from the internet, which is indistinguishable from a correct answer until a customer acts on it. Grounding converts an unbounded risk into a content management problem you can actually solve.

What to know

  • True grounding includes a refusal path. A system that always answers is not grounded, it is confident.
  • Source attribution is the practical test: can you see which document produced the answer?
  • Grounding does not fix bad content. It faithfully reproduces whatever you gave it, including errors.
  • Permissions matter. A grounded system must respect who is allowed to see which document.
  • Gap reporting, where the system logs questions it could not ground, is the most useful by-product.

An example

Example: a customer asks whether a feature works on a legacy plan. Nothing in your documentation covers legacy plans. A grounded system says it cannot confirm and routes to a human. An ungrounded one invents a confident answer that may be wrong.

Common mistakes

  • Assuming a vendor is grounded because they say the word
  • Grounding on documentation nobody has audited in a year
  • Removing the refusal path because it reduces the resolution rate
  • Ignoring the log of ungroundable questions, which is your content roadmap
Where Fidiora fits

Fidiora answers strictly from your own content and says so when it cannot. Every ungroundable question is logged as a documentation gap, so the questions Fidiora could not answer this month become the articles that let it answer next month.

See Pricing
FAQ

Questions

How do I test whether an AI support tool is grounded?
Ask it something your documentation does not cover and see whether it admits the gap or invents an answer. Then ask it something your documentation covers incorrectly and check whether it reproduces your error, which proves it is reading your content.
Is grounding the same as RAG?
Retrieval augmented generation is the common technique for achieving grounding, but grounding is the goal and the policy. A system can use retrieval and still be permitted to answer from general knowledge when retrieval fails, which is not grounded.
What happens when the content is wrong?
A grounded system repeats your error faithfully and confidently. That is a feature in the sense that it is traceable, and it is a reason to audit your documentation before launch rather than after.

Related terms

Retrieval augmented generation Retrieval augmented generation is a technique where a system searches a document collection for passages relevant to a q... 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... Knowledge gap analysis Knowledge gap analysis is the systematic identification of questions customers ask that your documentation does not answ...
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