How Astiva measures AI visibility

Every number Astiva reports is defined, computed, and refreshed on a published cadence. This page documents the full pipeline — query sampling, mention extraction, brand normalization, sentiment scoring, and freshness — so every claim is traceable to a method.

The 7 AISO Metrics

Pipeline: From query to dashboard

  1. Query Sampling — platform-specific query set across informational, commercial, and transactional intent; rotated to reduce cache bias.
  2. Mention Extraction — parses each response for mentions, position, sentence-level sentiment, and the competitive set.
  3. Brand Normalization — resolves mentions across casing, spacing, hyphenation, and misspellings; cross-validated against ground-truth human reviews.
  4. Sentiment Scoring — mention-level scoring aggregated per platform and per window; Sentiment Volatility exposes variance.
  5. Freshness Cadence — daily automated runs across every platform in the customer's plan; dashboard refreshes within 24h; alerts within 24h of visibility shifts.

What "accuracy" means at Astiva

Accuracy refers specifically to brand normalization accuracy — the rate at which the engine correctly resolves a mention to the intended brand entity. Measured against a versioned, human-reviewed evaluation set. Does not refer to sentiment classification or aggregate metric accuracy.

Start Free Analysis | See the Product

← Back to Astiva AI