Technical health
20 points- Crawl and index controls
- Canonical and redirect behavior
- Titles, descriptions, headings
- Internal linking and mobile performance
Scoring methodology · v2.0.0
Two scores. Two different questions.
AI Presence measures sampled recommendation outcomes. Website Readiness measures whether your site supplies accessible, clear, supportable evidence. Neither score is a guarantee.
AI Presence Score
The current score uses unbranded buyer prompts across ChatGPT, Gemini, Claude, and Perplexity. Mention rate contributes 40 points, owned-domain citation rate 25, competitive share of observed names 20, and platform breadth 15. A branded-accuracy prompt checks whether a platform describes the business correctly but is excluded from the score because a customer already supplied the brand name.
Each observation stores its platform, prompt, location, model, capture time, mention outcome, owned citations, and observed competitor names. One check means one prompt × one platform × one location. Repeated runs retain their timestamps so coverage, repeatability, and variance can be reported without blending older and newer answers.
Prompt demand and citation evidence
Every audit creates a service × location prompt pack organized by buyer intent and stage. When connected demand data is returned, measured monthly volume and CPC inform priority; otherwise the system clearly labels its intent-based priority as modeled. Citation URLs are normalized into a source graph, while branded samples separately verify whether expected identity, service, and market facts were observed.
Website Readiness Score
This deterministic score totals the five 20-point pillars below. Versioned rules—not generated prose—calculate it. AI may explain stored findings or draft proposed edits, but it cannot alter a raw metric or score. The v2 evidence inventory also samples key internal pages and reports robots, sitemap, canonical, schema, heading, title, and description coverage beyond the homepage.
Local relevance
20 points- Service and location clarity
- Page intent and content depth
- Service-area coverage
- Contact and conversion paths
Entity trust
20 points- Business identity consistency
- Organization and LocalBusiness schema
- Contact, address, and ownership evidence
- Credentials and visible proof
Authority + citations
20 points- Local search presence
- Directory and citation coverage
- Competitor visibility gaps
- Verifiable third-party signals
AI answer readiness
20 points- Direct-answer structure
- Decision facts and question coverage
- Machine-readable entity clarity
- Sampled mentions and cited sources
How missing data works
AI Atom Brain never converts a provider failure into a silent zero. If a required source is unavailable, the affected observation is labeled unavailable, confidence is reduced, and the deterministic website assessment can still publish. Unavailable AI observations are excluded from the AI Presence Score and queued for retry.
Local and broader-market competitor categories
Local comparisons use service-plus-location queries and local or map results. Broader-market comparisons use category-level organic results, with the United States as the v2 default. The categories retain separate queries, geographies, positions, and normalized visibility indexes. They are not blended into one market score.
Interpretation boundaries
These scores are prioritization aids, not platform ranking factors or predictions of traffic. AI recommendation observations are timestamped samples from specific systems, models, prompts, and locations; they are not a universal ranking of every possible answer. AI Atom Brain does not promise rankings, citations, traffic, mentions, or placements.
Approval-first Golden Loop
A verified Brand Context Pack and a decision-ready qualified-outcome frame gate proposed work. Generation settings, Knowledge Graph fields, and indexed retrieval are separate evidence surfaces; a successful write to one does not prove the others have updated. For article imagery, the exact final title must appear in the unique hero that is also used for social and featured-image metadata, while inline-image relevance and asset quality are verified independently. The approved organizational author—not an AI guide or fictional persona—owns the content identity.
Every customer change then follows a recorded chain: evidence-linked proposal, human approval, controlled deployment with a rollback plan, independent live verification, outcome measurement, and a retain, revise, or roll-back decision. Installing AI Atom Brain means activating measurement and workflow modules, connecting authorized data sources, and implementing approved fixes—not placing an autonomous rewriting script on the customer’s website.