The methodology, stated plainly

How scores are built

• AI readiness — five drivers of whether AI can read and trust your foundation (clarity, source support, readability, point of view, recommendation fit), scored 0–5 each and weighted 25/25/20/15/15.
• AI visibility — a deterministic blend: 70% correct-business appearances on blind discovery questions and 30% recommendation performance in comparison questions. A comparison answer scores 100 when you are recommended first, 50 when you are listed, and 0 when you are not recommended.
• Recognition floor — if at least one completed direct or comparison answer identifies the correct business, Visibility cannot be lower than 15. Prompted recognition never adds above 15 and never substitutes for being found or recommended by a buyer-facing question.
• AI accuracy — how much of what AI says survives verification.
• Daizie Score — the one headline number: 45% accuracy, 30% visibility, 25% readiness.

The scoring math lives in code, against a versioned rubric — AI is used only to read and report what was observed, never to judge freely. Every measurement is logged and reproducible against its rubric version.

Visibility formula version: visibility-70blind-30comparison-prompted-floor15-v4. Visibility is withheld when blind-question coverage is below 80%, an engine completes fewer than half of its expected blind attempts, or a supposedly blind question contains the business name, domain, or alias.

How to read the 1–100 spectrum

Scores are the first finding in every report. The free snapshot opens with AI Visibility. The paid report opens with the Daizie Score, then shows Readiness, Visibility, and Accuracy so you can see what is driving the headline number.

Accuracy uses the same numeric brackets with accuracy-specific names: High risk, Unreliable, Mixed, Reliable, and Highly reliable. A score is shown as unavailable when coverage or required inputs are missing; missing information is never turned into a zero.

How the named competitor comparison works

When you provide exactly two competitors, Daizie can compare all three businesses using the same completed blind buyer answers—the questions that name none of you. Recommendation reach is the percent of those answers that list a business. Strength when listed is 100 when it is listed first and 50 when it appears later, averaged across its appearances. A business that never appears receives zero on both coordinates.

The map requires at least 80% blind-answer coverage. Your business is identity-checked against its website; named competitors are matched from the names and domains you submit, so their identity resolution is less deep. The map describes this dated snapshot—not total market share, product quality, website Readiness, Accuracy, or universal buyer preference. The longer “share of recommendation” list also includes comparison questions and therefore answers a different question: who was named anywhere in the buyer-facing report.

How the Action Plan works

Prompt-level findings are evidence, not one task per prompt. Deterministic rules consolidate repeated observations into no more than three website priorities, each with its evidence, target location, and first move. Separately, structured findings match a reviewed, versioned catalog of practical visibility opportunities for a local, service, or product business. The catalog—not a model—controls each option’s instructions, real links, eligibility, warnings, effort, and completion criteria.

Customers may choose up to three opportunities for a Focus Plan. Suggestions are bounded starting points, not requirements, rankings, or guaranteed outcomes. Daizie records the plan; it never posts, registers, purchases, or edits an outside account on the customer’s behalf.