Skip to main content

​
Direct answer

Define the event date, freeze a relevant prompt and provider cohort, and compare sentiment distributions before and after the event. Read the underlying answers and report the change as temporal association. The comparison alone cannot show that the crisis or launch caused it.

​
Data required

You need the dated event, unchanged configured prompts, topics or tags, providers and models, comparable pre-event and post-event periods, stored answers, sentiment classifications, SELF/DIRECT mentions, and citation URLs.

​
Workflow

1

Define the comparison

Record the event date and choose pre-event and post-event periods that use the same monitoring configuration.
2

Check cohort stability

Confirm prompts, topics, providers, models, geography, and run coverage. Note any configuration or model changes.
3

Compare and inspect

Compare sentiment distributions with answer counts, then read representative answers from both periods.
4

Document alternative explanations

Record concurrent announcements, source updates, competitor events, prompt changes, and small sample sizes.

​
Interpretation

A difference between periods is an observed change in selected stored answers. It may be associated with the event, but it does not establish causation. Report counts, shares, filters, and examples together; do not label a percentage as good or bad without a stated business baseline.

​
Possible next actions

  • If a verified factual issue appears after the event, test an authoritative clarification and keep the original comparison cohort.
  • If the change occurs only in one provider or prompt, investigate that slice before making a broader reputation decision.
  • If answer counts are sparse, extend observation or combine only genuinely comparable runs.

​
Limitations

Provider models, retrieval behavior, sources, and monitoring configuration can change between periods. Sentiment classification can miss nuance. A return toward the earlier distribution does not prove that communications or content caused the change.

Exports

Export comparable pre-event and post-event cohorts.

Answer Details

Build a time-based answer report.

Shared Links

Share a filtered product view.

Visibility over time

Apply the same cohort discipline to visibility metrics.

Correct misinformation

Address verified factual discrepancies.

Negative sentiment by topic

Locate the topics behind the observed change.

What AI says

Read the underlying answers.