Your Visibility Index is a 0–100 diagnostic summary of how AI engines treated your brand across one defined set of tracked prompts. It is not a site grade, competitor rank, or percentile.
Start with the outcomes
Before interpreting the index, read the two outcome rates beside it:
- Presence — answers that named your brand, shown as a count and a percentage of measured answers
- Own-site citations — answers that linked to your own site, using the same measured-answer denominator
Those counts tell you what happened. The index compresses four answer-level signals—mention, own-site citation, portrayal, and answer position—so you can follow the same measurement set over time.
How the index is calculated
Stride scores each measured engine independently, then averages the engine results equally. One engine cannot dominate the index merely because it returned more answers.
| Signal | What it measures | Weight within each engine |
|---|---|---|
| Mention | How often the answer names your brand | 40% |
| Own-site citation | How often the answer cites one of your pages | 25% |
| Answer portrayal | Sentiment when your brand is mentioned | 20% |
| Answer position | How early your brand appears when mentioned | 15% |
Missing or failed provider responses are not measured. They do not become negative answers. Only full sweeps update the index; priority checks refresh individual prompt evidence without changing the headline.
Diagnostic cards are not extra score weights
The dashboard Foundation rail explains where to investigate—inputs such as External Brand Presence, Answer Portrayal, Crawl & Indexability, and Structured Data Coverage, paired with AI Readiness. Those rings are site-health and crawl-structure checks, not extra weights in the Visibility Index.
Presence and engine reach already live in the hero and the Coverage panel, so they are not repeated as extra tiles on the home view.
If Stride has not collected a diagnostic input, the card reads not yet measured rather than 0.
Why one run can move
AI engines are non-deterministic. The same prompt can return different brands, sources, and wording on different runs. Providers also update models without notice.
Stride therefore separates three states:
- This run — the measured counts and rates now
- Within normal variation — movement too small or too sparse to interpret
- Repeated comparable movement — a direction observed across scans with the same prompt, engine, competitor, and scoring configuration
When the normalized prompt-text or engine work set changes, Stride starts a new Dashboard Index comparison epoch. A competitor-only change does not necessarily restart that Index trend. Action proof uses a stricter measurement epoch that also protects target and competitor evidence, so it can become not comparable even when the Index trend continues. Stride does not redraw old readings with a new formula or connect unlike scans with a misleading trend line.
Don't over-read one scan
A single movement is an observation, not proof that a change worked. Wait for repeated comparable scans and check the action's expected measurement date.
Mentions and citations need different work
A mention means an engine recognizes your brand. An own-site citation means one of your pages was used as a source. You can have either without the other:
- Mentioned but not cited: strengthen first-party evidence and source-worthiness
- Cited but not named: make the page-to-brand relationship clearer
- Competitors named while you are absent: inspect the exact prompts and sources shaping those answers
How Stride checks completed work
When you record work as implemented, Stride records what changed, the target prompt or page, the expected leading signal, the likely waiting period, and the next measurement date. Resolving an Action without changes closes the recommendation without creating implementation proof or a measurement marker. The proof timeline then separates:
- implementation recorded;
- the declared live check passed, when one is available;
- a positive action-specific signal observed on comparable evidence;
- repeated comparable visibility movement; and
- target-aligned traffic or search movement when an integration has complete, sufficient 28-day before-and-after windows.
The strongest valid historical stage is preserved separately from current evidence health, so a later provider failure, insufficient evidence, changed measurement epoch, confounding, integration unavailability, or source deletion does not rewrite history. These states fail closed: none can promote proof or become a zero, an absence, or attribution. Stride uses “observed after” language rather than causal language. A score change alone is never treated as proof.
Choosing what to do next
Actions are prioritized from their scan evidence, recurrence, connected search demand when an exact query match exists, and implementation effort. Stride does not promise a fixed number of score points or rank work with invented ROI. Evidence labels describe the source coverage—deterministic check, repeated comparable scans, single scan, or insufficient evidence—not statistical certainty.