Scoring Methodology

How the Score Works

Every business website is evaluated across four dimensions that predict whether an AI agent will choose it over a competitor. Each dimension is scored 0 to 100, then combined using weighted averages calibrated by industry vertical to produce a composite AI Agent Preference Score (0 to 100). The numeric score is the primary metric. Each score falls within one of six named capability bands.

The scoring engine analyzes the live website, its structured data, technical infrastructure, transaction pathways, and booking platform integrations. It cross-references business identity data against authoritative public sources. Scores reflect what an AI agent would encounter when attempting to operate on the site today. No self-reported data is used.

What the score does not measure

The AI Agent Preference Score is not a measure of search engine optimization (SEO), AI answer visibility (AEO), design quality, or business reputation. SEO gets you found. AEO gets you mentioned. This score predicts whether you get booked. A site can rank well on Google, appear in ChatGPT answers, and look beautiful while still scoring poorly on agent preference.

Four Scoring Dimensions

Each dimension isolates a specific aspect of agent preference. Together they answer the question an AI agent asks before acting: can I reach this business, can I finish the transaction, can I trust the data, and how does it compare to everyone else?

Dimension 1

Agent Accessibility

Can AI agents reach and parse this website?

Evaluates the technical foundation that allows AI agents to interact with the site. This includes semantic HTML structure, accessibility features, bot access policies, machine-readable navigation, and the absence of barriers like aggressive CAPTCHAs or anti-bot systems that block automated visitors.

Semantic HTML and structured navigation
AI bot access policies (robots.txt directives for AI crawlers)
Form accessibility and keyboard navigation
CAPTCHA implementation (graduated vs. blanket blocking)
Sitemap presence and server-side rendering

Dimension 2

Transaction Completeness

Can an AI agent actually finish the job?

Maps the full transaction path an agent walks before completing a task: discover the offering, evaluate it, contact the business, book the appointment, and pay. A break at any stage kills the transaction. Scoring combines booking platform detection across scheduling, field service, and industry-specific platforms with transaction path stage coverage, calibrated by vertical category.

Five-stage transaction path (discover, evaluate, contact, book, pay)
Booking platform detection across 30+ scheduling and field service platforms
Vertical-calibrated transaction signals per category
Form quality, payment endpoints, and pricing transparency
Call-to-action clarity at every stage

Dimension 3

Data Reliability

Will the data an agent extracts lead to a successful outcome?

Verifies that operational data is consistent, current, and trustworthy across the site and across the public web. Agents that book with wrong hours or call a disconnected number fail the customer. Data Reliability cross-checks your NAP and identity against Google Places and every public source in our database. It combines entity coherence, operational data structure, and data accuracy into one measure of trust.

Entity coherence against Google Places and public directories
NAP consistency (name, address, phone) across the site
Schema.org markup for services, hours, pricing, and location
SSL, identity verification, and cross-page consistency
Data freshness signals and server-side rendering

Dimension 4

Competitive Position

How do you rank against every competitor in your market?

Placed in the AI agent preference context, the choice between two businesses is rarely a tie. This dimension compares your score against every business in your vertical and city, drawing from our database of 500,000 scored businesses and their historical trend lines. Relative position is part of the composite score, not a separate report section.

Percentile rank across vertical and metro area
Per-dimension competitive deltas
Historical trend comparison over multiple scoring versions
Gap-to-leader analysis
Unique to the full report

How we got here. The original AAO framework launched with six internal dimensions: Agent Compatibility, Transaction Readiness, Agentic Commerce Readiness, Operational Data Structure, Data Accuracy and Currency, and Competitive Position. Methodology v6 reframed those signals into the four-dimension AI Agent Preference Score: Agent Accessibility (from Agent Compatibility), Transaction Completeness (combining Transaction Readiness and Agentic Commerce), Data Reliability (combining Operational Data Structure and Data Accuracy), and Competitive Position. All six internal dimensions continue to be calculated and stored for every scan to power historical trend analysis.

Capability Bands

The AI Agent Preference Score is a 0 to 100 numeric metric. The number is the primary output. Each score also falls within one of six named capability bands that describe what AI agents can do at that level. Band boundaries are fixed and will not change between methodology versions.

Band	Score Range	What It Means
Agent Preferred	90 – 100	AI agents can fully navigate, transact with, and confidently recommend this business. Booking platforms integrated, structured data comprehensive, identity verified across sources. Industry leader in agent preference.
Agent Optimized	70 – 89	Strong foundation with minor gaps. AI agents can work with this business effectively. Schema markup present, transaction pathways functional, data largely consistent.
Agent Ready	50 – 69	AI agents can handle basic tasks but gaps limit what they accomplish. Data extraction and core transactions are possible. A handful of targeted improvements move this tier into optimized territory.
Agent Functional	30 – 49	AI agents find this business but struggle to complete transactions. Navigation works and limited data is available, but reliable structured data and booking integrations are missing.
Agent Detected	10 – 29	AI agents have trouble interacting with this business meaningfully. The site exists but operational data and transaction pathways are largely absent.
Agent Incompatible	0 – 9	AI agents cannot effectively navigate or transact with this business. No structured data, no accessible transaction paths, no meaningful agent preference signals.

Vertical Calibration

The AI Agent Preference Score is calibrated by industry. A plumber, a law firm, and a medical practice have fundamentally different primary transactions, and the scoring reflects that.

Dimension weights and transaction readiness criteria are adjusted per vertical category to ensure fairness. A plumber is not penalized for lacking a patient portal. A law firm is not penalized for lacking online ordering. Each business is scored against what is realistic and relevant for its industry.

The engine recognizes four vertical categories, each with distinct weight profiles and transaction signal calibration:

Trades and Home Services (plumbers, electricians, HVAC, roofers, contractors, cleaners, movers)
Professional Services (legal, medical, dental, financial, insurance, accounting, real estate, veterinary)
Retail and Hospitality (restaurants, retail stores, salons, hotels, food service, cafes)
General (all other business types)

Vertical detection is automatic based on website content, Schema.org types, and business directory classifications. Within each category, the engine maps 183+ specific service verticals from our database.

Scoring Principles

Capability-defined bands

Each band maps to a defined level of agent capability. The boundaries are set by what a business can do at that score, not by how many businesses fall into each band.

Observed, not self-reported

Scores are based entirely on what the scoring engine observes on the live website. No surveys, no self-assessments, no manual input. The score reflects what an AI agent would encounter today.

Stable methodology

Band boundaries are locked and versioned. Businesses can track progress over time knowing the measuring stick has not moved. Methodology changes are versioned and documented.

Industry-fair

Vertical calibration ensures businesses are scored against criteria relevant to their industry. A restaurant and a law firm are held to different transaction standards.

Actionable, not academic

Every dimension measures something a business can improve. The full report includes specific findings and recommendations tied to each dimension score.

Forward-looking

Transaction Completeness rewards early adopters of booking platforms and agent-ready payment endpoints. As AI agent commerce matures, businesses that prepared early will have a compounding advantage.

Data and Coverage

The GradeForAI database includes 500,000 scored businesses across 395 US cities and 183 service verticals. Businesses are discovered through public search engines and business directories. Each business is rescored periodically to reflect website changes.

Competitive Position analysis requires a minimum density of similar businesses in the same city and vertical. When sufficient comparison data exists, the full report includes percentile rankings across all four dimensions.

Methodology version

This page describes methodology v6.2, effective April 2026. The scoring engine, dimension definitions, and band boundaries are versioned together. v6 introduced the AI Agent Preference Score as the headline metric and restructured scoring from six internal dimensions into four public dimensions. v6.2 added four-category vertical calibration with distinct weight profiles per category, price-spectrum detection, and expanded transaction completeness signals. The original six internal dimensions continue to be calculated and stored for historical trend continuity.