---
language: en
title: "Factual accuracy prompt"
description: "A question used to check AI claims about a product against appropriate documentation, distinguishing factual errors, omissions, and unverifiable information."
canonical: "https://tropk.ai/en/resources/glossary/factual-accuracy-prompt"
last_modified: "2026-10-12T16:12:40.174Z"
---

# Factual accuracy prompt

Reviewed on 2026-10-06.

## What it means

A factual accuracy [prompt](/en/resources/glossary/prompt) is a question used to examine whether a product description matches verifiable references. In Profound’s classification, an accuracy prompt addresses claims about features, terms, or capabilities. The evaluation compares each claim with documentation appropriate to the fact, version, region, and date; the response itself is not confirmation.

## Why it matters for marketing

For marketing, this check identifies information that could mislead someone making a purchasing decision. Favorable copy might describe a nonexistent integration or outdated plan terms. The team can refer a specific discrepancy to the product owner and revise the corresponding public information, without confusing praise, recommendations, and corrections to the facts presented.

## Example in practice

Imagine an equipment supplier asking whether its current model can operate offline. The response says it works fully in this mode, but the manual requires a connection to export files. The technical owner locates the relevant passage and manual version. Marketing records the omitted limitation and revises the description of exporting used in sales materials.

## How to apply or verify

Define the question and gather references before evaluating the response. For each claim, record the passage, the source used to check it, its validity period, and the judgment: supported, contradicted, or insufficient evidence. Consult the person responsible for the specification if documents conflict. Any [citation](/en/resources/glossary/ai-citation) provided by the model must also support exactly the claim attributed to it.

## Important distinctions

Factual accuracy is not a classifier’s statistical accuracy or an overall quality score. A feature’s absence from the response may be an omission, without constituting a false claim. Asking AI to check its own facts can help identify questions, but does not replace appropriate references. Variable commercial information needs to be checked again when it changes.

## Sources and references

- [Chapter 3: Promptology](<https://www.tryprofound.com/aeo-guide/chapter-3>) — Profound
- [Reduce hallucinations](<https://platform.claude.com/docs/en/test-and-evaluate/strengthen-guardrails/reduce-hallucinations>) — Anthropic

## Continue a leitura

- [AI hallucination](</en/resources/glossary/ai-hallucination>)
- [Primary source](</en/resources/glossary/primary-source>)
- [Factual accuracy score](</en/resources/glossary/factual-accuracy-score>)

## About the author

[Luiz Gardelin](</en/team/luiz-gardelin>)

Luiz Gardelin is the founder of Tropk.ai and works in GEO/AEO, product vision, strategy, and growth, helping brands understand and improve their presence in AI responses.

---

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