What the score measures—and what it cannot prove.
ActableSite evaluates observable signals on a public website. The score is a repair-prioritization aid, not a simulation of ChatGPT, a search ranking, or a guarantee of commercial outcomes.
Six questions organize the 15 checks
Identity
Can software tell who operates the site, what it offers, and which page is authoritative?
Action
Is the intended next step explicit, crawlable, and labeled for its outcome?
Decision
Are price, commitment, delivery, audience, and material limits visible before action?
Evidence
Are important claims supported by attributable facts, examples, sources, or methodology?
Discovery
Do canonical, sitemap, crawler, metadata, and structured-data signals reduce ambiguity?
Trust
Can a visitor find contact, identity, privacy, terms, and current-information signals?
How scoring works
Each check has a fixed maximum weight. A clearly present signal earns full points; a partial or ambiguous signal earns roughly half; a missing or blocking signal earns none. The weighted total is normalized to 100.
Important limits
- The scan reads HTML and common public discovery files; it does not manually test every page, workflow, device, or assistive technology.
- Automated checks cannot prove accessibility, legal compliance, factual truth, search inclusion, citation, recommendation, or revenue impact.
llms.txtis an experimental convenience signal with low weight. Its presence is not treated as a ranking requirement.- Generated starter files must be reviewed against the business’s real facts and policies before publication.