How Zertura measures.
Every score traces back to a logged API call. Click the ⓘ on any tile to see the calls behind it.
Perception Score
The Perception Score samples how consumer-facing AI engines actually describe your destination. Each monthly run executes 18 visitor-intent prompts across a 6-LLM panel, repeated 3 times per engine to capture stochastic variance — 324 logged API calls per destination per month.
Copilot intentionally excluded — its training data overlaps with ChatGPT's; including both would double-count.
Site Score
The Site Score audits how readable your destination's web presence is to the AI source layer. Each monthly run sends 12 distinct user-agents (the major indexing crawlers + emerging AI crawlers) at 20 pages selected from your top GA4 traffic — the pages your visitors actually read, not the pages we guess matter. We also score up to 100 external & earned URLs: 35 standard sources (Wikipedia, Yelp, Reddit, TripAdvisor, Google Business Profile, major editorials) plus up to 65 you supply.
| Dimension | Intentful | Zertura |
|---|---|---|
| Crawlers | 10 | 12 |
| Pages audited | 20 pages they guess matter | 20 pages your visitors actually read (GA4) |
| External sources | 79 | Up to 100 (35 standard + 65 custom) |
Brand Pillar Alignment
Brand Pillar Alignment scores how reliably AI sources reinforce each of your stated brand pillars. It is exempt from the composite DAVI Score — pillar work is a slower-moving, strategic signal that shouldn't whipsaw the headline number. Alignment is scored by Claude Sonnet 4.5 against the citation sources gathered during the External & Earned pass, and refreshed with each monthly panel run.
DAVI Score formula
Default weights are configurable per destination. A destination heavy in earned media and editorial coverage can shift toward Perception; a destination still building source coverage can shift toward Site. Brand Pillar Alignment and DSoP (Destination Share-of-Prompt) are tracked separately and never enter the composite.
Score bands
Severe gaps — fix this first.
Inconsistent or partial coverage.
Reliable presence with room to lead.
Best-in-class for the panel.
DSoP and Brand Pillar Alignment display as plain numbers — exempt from banding. Bands apply to DAVI Score, Site Score, Perception Score, Partner Ecosystem Health, and individual dimension scores.
Audit trail
Every API call is persisted in prompt_results with the full raw response, model ID, tokens in/out, latency, estimated cost, and within-cell variance across the 3 runs. Every score on every dashboard traces back to the exact rows that produced it — click the ⓘ icon on any score tile to see the calls behind that number.
Full Postgres export available on request at no charge. No black-box composites, no "trust us" — the moat is the transparency.