How to get the best results working with FactCheck

Last updated: July 20, 2026

Here are some best practices to help you make the most of FactCheck.

What makes good FactCheck prompts

FactCheck prompts should have a realistic chance of triggering an LLM response that mentions your brand and makes a verifiable, objective claim: a price, a feature, a spec, or a policy. Vague and subjective prompts ("Is Brand X good?") belong in Sentiment. Comparisons ("Brand X vs Y") can work as long as they include verifiable claims.

Branded vs unbranded prompts

Branded prompts — prompts that mention a brand directly. For example: "Does Profound track brand mentions across AI search engines?"

Unbranded prompts — prompts that ask broader category questions and may surface a brand naturally. For example: "What are the best tools for tracking brand visibility in AI search and why?"

Branded prompts are recommended but not required. They make it easier to filter for relevant responses and reduce noise. Unbranded prompts are fine when the category is specific enough that the brand is likely to surface anyway (for example, "What is the best AEO tool?").

Examples

Good: factual, specific

Bad: vague, subjective

  • Can Profound track brand mentions across ChatGPT and Perplexity

  • How much does Profound's Pro plan cost?

  • What is Profound's refund policy?

  • Is Profound available in Australia?

  • Is Profound a good company

  • Tell me about Profound.

  • What do people think of Profound's customer service

  • What is the best CRM tool?

What makes a good Knowledge Base for FactCheck

Content best practices

  • Make sure the information in your Knowledge Base answers the prompts you're tracking with FactCheck. For example, if you're tracking prompts about pricing, include a pricing page or FAQ that covers it.

  • Make sure the information in your Knowledge Base covers all the content relevant to the prompts you're tracking. If coverage is partial, results degrade in two ways:

    • Claims about topics your Knowledge Base doesn't cover return as "not relevant" because FactCheck has nothing to verify them against.

    • Claims about topics it covers only partially may be misclassified. For example, a claim about an old pricing tier might be marked accurate if the Knowledge Base doesn't include the updated pricing page.

  • Add your public-facing website to your Knowledge Base. Knowledge Bases support recrawling on a daily, weekly, or monthly cadence, keeping your source of truth current automatically. Learn more in the 📄 Create and manage Knowledge Bases section.

  • Use live, publicly accessible pages you own, such as help center articles, policy pages, and product pages. Set up an auto-sync schedule to keep the content current.

Limitations

  • Each FactCheck configuration uses exactly one Knowledge Base, and you can't split verification across several.

  • You can't change which Knowledge Base is connected to FactCheck after the initial setup. However, its contents stay fully editable: add, remove, or update files at any time, and the next time FactCheck verifies claims, it will check them against the updated content.