---
language: en
title: "Prompt set"
description: "A prompt set is a selection of questions used to track AI responses. Its composition defines the scope and limits of visibility comparisons."
canonical: "https://tropk.ai/en/resources/glossary/prompt-set"
last_modified: "2026-10-12T16:12:59.280Z"
---

# Prompt set

Reviewed on 2026-10-06.

## What it means

A [prompt](/en/resources/glossary/prompt) set is a defined selection of questions used to track [AI responses](/en/resources/glossary/ai-response) about a market, a brand, or a need. Each question should have identifiable wording and context to allow it to be run again. The set’s composition determines the scope being observed: results from selected questions do not automatically represent all conversations among potential buyers.

## Why it matters for marketing

For marketing, choosing this set is a decision about what will be measured. A selection dominated by questions that name the company assesses different behavior than questions about the category. If the team changes the composition between periods, a change in the metric may reflect the new scope, making it harder to assess changes in presence.

## Example in practice

Imagine an equipment manufacturer tracking twenty questions that mention its models. In conversations with distributors, it identifies questions about maintenance and compatibility that do not mention brand names. The team creates a second set for these needs and keeps the first as a historical reference. This makes it possible to investigate discovery without mixing the comparison with the previous scope.

## How to apply or verify

Record the wording, topic, language, audience, source, and reason for including each question. Check for repetitions and gaps with people close to customer service or sales. Keep versions of the set and compare periods using the same selection, surface, and collection conditions. When adding questions, report the change and present the scopes separately when necessary.

## Important distinctions

A list created by the team is not equivalent to a record of questions actually asked by users, nor does it indicate the volume of demand. Profound distinguishes its own prompts from queries sourced from its database. LangSmith documents versioning for evaluation sets. These are examples of specific practices and products, and do not establish a universal ideal size for monitoring.

## Sources and references

- [Answer Engine Insights: #1 AI Search Visibility Platform](<https://www.tryprofound.com/features/answer-engine-insights>) — Profound
- [Manage datasets](<https://docs.langchain.com/langsmith/manage-datasets>) — LangChain

## Continue a leitura

- [Prompt sample coverage](</en/resources/glossary/prompt-sample-coverage>)
- [Branded prompt](</en/resources/glossary/branded-prompt>)
- [Unbranded prompt](</en/resources/glossary/unbranded-prompt>)

## 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>)