What is an AI Trust Score, and how is it calculated?

AI Trust Scanner builds a score by crawling what's actually on a store — the homepage, up to ten product pages, up to five category pages, and one example of every informational page type it can find, from FAQ to shipping policy. Twenty-nine rule-based detectors run against that crawl and record a signal for each one: does a review widget exist, does robots.txt allow GPTBot, is there a returns policy page anywhere. None of those detectors involve a language model making a judgement call — each is a specific, checkable fact about the site, which is what makes the resulting score explainable instead of a black box.

Those signals feed into eight category scorers, each carrying a different weight depending on how strongly it predicts real trustworthiness. Reviews & Reputation carries the most weight, at 18 of the 100 total points, because third-party proof that real customers bought and stayed happy is the hardest signal to fake. AI Readiness, Business Transparency, SEO & Authority, and four more categories fill out the remaining points. The full breakdown of every category, weight, and individual signal inside it is published on the methodology page — nothing about how a store earns its score is held back.

The score is versioned: every audit is stamped with the methodology version it ran under, so a score from six months ago stays comparable to the rules that produced it, rather than silently drifting as the model changes. A score of 70 or higher qualifies a store to display the live badge, and higher tiers add human review and domain verification — see pricing for what each tier includes.