---
language: en
title: "Information retrieval"
description: "Information retrieval selects documents or passages relevant to a query. Learn how to check the retrieved material before evaluating the answer."
canonical: "https://tropk.ai/en/resources/glossary/information-retrieval"
last_modified: "2026-10-12T16:13:27.283Z"
---

# Information retrieval

Reviewed on 2026-10-06.

## What it means

Information retrieval, or retrieval, is the selection of documents or passages relevant to a query from an available collection. In AI systems, this material may be provided as context to the model that will produce the answer. The source may be an internal database or a web index; retrieving information does not necessarily require querying live pages.

## Why it matters for marketing

For marketing, retrieval helps locate the source of an outdated or incomplete answer. If the system selects an old commercial policy, improving only the [prompt](/en/resources/glossary/prompt)’s style may not solve the problem. The team needs to check which documents reached the model before assessing the wording, and decide whether to revise the collection or the selection mechanism.

## Example in practice

A manufacturer makes installation manuals available to a support assistant. In a hypothetical test about a new piece of equipment, the retrieval log shows three passages from the previous model, even though the correct manual is in the collection. The team decides to review version identification and passage selection because the error occurred before the final guidance was generated.

## How to apply or verify

To verify, use queries that correspond to known documents and examine the retrieved results before the answer. Record the identifier, version, passage, and position of the items, when available. Check whether the necessary document was accessible in the queried collection. The person responsible for the integration should distinguish between access failure, absence from the collection, and unsuitable selection of material that was already available.

## Important distinctions

Retrieval is not training: consulting a document does not demonstrate a change to the model’s parameters. It is also not the final answer; the generator may omit or misinterpret the selected material. [RAG](/en/resources/glossary/rag) combines retrieval and generation. The original RAG paper uses an index of Wikipedia passages, illustrating that this mechanism can operate on a previously prepared collection.

## Sources and references

- [Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks](<https://arxiv.org/html/2005.11401v4>) — Patrick Lewis e coautores, arXiv

## Continue a leitura

- [RAG](</en/resources/glossary/rag>)
- [Passage retrieval](</en/resources/glossary/passage-retrieval>)
- [Search index](</en/resources/glossary/search-index>)

## About the author

[Vinicius Pacheco](</en/team/vinicius-pacheco>)

Responsible for Tropk’s technical content, focusing on crawling, indexing, structured data, and agent access to content.

---

[All pages in Markdown](<https://tropk.ai/sitemap.md>)