AI search history · time-oriented evidence

Read historical AI-search evidence as a time series.

Historical is a supported LLM Mentions operation for viewing returned evidence over its available time range. Compare like-for-like timestamps and keep data availability separate from a promise of complete history.

Start free trial AI visibility hub

Direct answer

Match the evidence to the decision.

Keep the selected provider, target, operation, locale, and returned fields attached to every interpretation.

Supported workflow

Inspect this evidence question step by step.

Fix the comparison frame first.

Keep target, provider, location, language, and operation consistent. A changed input can look like a trend even when the underlying evidence is not comparable.

Look for a pattern, not one point.

A single movement can be noise or a bounded data change. Review several comparable observations before deciding whether a pattern deserves investigation.

Choose Delta or New & Lost for sharper questions.

Use Delta when the question is the returned change between comparable points. Use New & Lost when the question is which mentions appeared or disappeared.

Keep the evidence boundary visible.

Record the provider, target, operation, location, language, and returned fields. The result is bounded evidence—not a guarantee, attribution model, exhaustive index, or automatic optimization outcome.

Related evidence

Continue with the operation that answers the next question.

Keep the product boundary explicit.

Is every historical date available?

No completeness promise is made. Use the dates and fields returned by the supported operation.

Does a historical change explain why it happened?

No. Time-oriented evidence shows returned change; attribution requires separate investigation.

Continue the workflow

Inspect a real evidence question in Gavix.

Start free trial