← All features

AI claims

Discover recurring claims about your brand and inspect their source evidence and verification status.

Request a subscription Explore this feature

An advance subscription is required to use the service. Send your request using the form below.

One measurement system
OpenAI APIGemini APIPerplexity APIGoogle AI OverviewGoogle AI ModeChatGPT

API measurements and separately measured interfaces can produce different results. Measurements describe the configured prompts and retrieval context.

AI Visibility

What question does it answer?

Discover recurring claims about your brand and inspect their source evidence and verification status. Processed answers are required. Source verification depends on project settings and available evidence; automated results are not a comprehensive fact-checking guarantee.

Sample project

Explore this feature with sample data

Northstar · illustrative project · comparable 30-day periods. These examples explain how to use and interpret the measurements.

Claim groups

30

Supported groups

20

Contradicted groups

4
Claim groups by evidence status

Figures shown are illustrative sample data.

Owned-source supported groups in the sample workflow
0 7.5 15 3 Week 1 6 Week 2 8 Week 3 12 Week 4

Cumulative illustration ending at twelve groups, not a processing-time promise.

Figures shown are illustrative sample data.

From measurement to a decision
  1. Choose a period and topic
  2. Inspect recurring claims and evidence
  3. Record accurate brand information where needed
  4. Review subsequent occurrences
Detailed sample breakdown
Metric / scopeSample valueInterpretation
“Offers small-business packages”Supported by owned sourcenorthstar.example/services
“Only serves large businesses”ContradictedThe package description provides contrary evidence
“Fastest provider”UnresolvedNo adequate comparison evidence

Figures shown are illustrative sample data.

How should you read the result?

The chart counts thirty claim groups, rather than answers or brand mentions. Twenty have supporting sources, four are contradicted, and six are unresolved. Inspect evidence origin and text.

Which decision does it support?

Prioritise repeated commercially important inaccuracies for review of the original answers and current web information.

What should you check first?

Frequency does not prove truth. Owned-source support is not an independent external verification.

Request a quote for your project

What do you see?

  • Meaning-based claim groups and occurrences
  • Source evidence and supported, contradicted, unresolved, or pending states
  • Customer-provided information and owned versus external evidence

How do you use it to decide?

  1. Choose a period and topic
  2. Inspect recurring claims and evidence
  3. Record accurate brand information where needed
  4. Review subsequent occurrences
Request a subscription

An illustrative use case

AI claims Northstar serves only large businesses, while its service page describes a small-business package. Reviewing the claim and evidence together helps investigate inaccurate brand portrayals.

An illustrative scenario, not a customer result or a promise of results.

What does measurement require?

Processed answers are required. Source verification depends on project settings and available evidence; automated results are not a comprehensive fact-checking guarantee.

Frequently asked questions

Is a claim the same as a mention?

A mention establishes that the brand appears; a claim describes what the answer says about it. One answer may contain several claims.

Does unresolved mean false?

No. Available evidence is insufficient to support or contradict the claim.

Can I provide accurate information?

Important information and an optional source URL can be recorded for subsequent verification. This does not directly change platform answers.

What should I check before comparing periods?

Check the selected question or URL, filters, sample size, source coverage, and data freshness. Different scopes can produce differences unrelated to a real change.

Does missing data mean zero performance?

No. A source connection, retrieval, or processing step may be missing. Check availability before interpreting an empty block.

Can I use this to guarantee an AI recommendation?

The feature supports evidence-based investigation. It does not control provider answers or guarantee recommendations or citations.

Related features

Make your AI presence measurable

Request a quote for your project

Tell us which brand and website you want to measure and what question you need to answer. We will discuss the subscription and data connections your project requires.