How Fakeshop AI calculates its trust score
No black-box guesswork: we disclose which signals we check, how they are weighted and where our data comes from. The score is guidance – an automated assessment, not legal advice.
How the trust score is built
Every check starts at 100 points. Each examined signal adds or subtracts weighted points – serious red flags (e.g. a missing legal notice, prepayment only, a match on a warning list) lower the score significantly, positive signals stabilise it. The result is a single, traceable number from 0 to 100, plus a reason for every signal.
The score is cross-checked by a second, adversarial AI counter-call: one model deliberately tries to refute the first assessment before any hard verdict ("danger") is issued. This prevents premature false alarms from AI hallucinations.
Clear fraud signals – we advise against purchasing.
Unclear or mixed signals – caution and further checks advised.
No relevant red flags found – no guarantee, but a good sign.
The signals we check
More than 30 signals feed into every assessment. The most important groups:
Completeness under German law: company, address, trade register, VAT ID. Missing mandatory details lower the score significantly.
Very young domains (days to a few weeks) are a strong warning sign for short-lived fake shops.
Prepayment only is one of the strongest single signals. Buyer-protection options are stabilising.
Bundling several independent sources to tell fake review waves from genuine feedback.
We check whether a seal image actually links to a valid, official certificate – not just whether it is shown.
Unrealistically high permanent discounts on branded goods are a classic lure and warning sign.
Encryption is mandatory but no proof of trust on its own; suspicious hosting patterns feed in.
Cross-check against community- and AI-fed lists of known fraudulent shops.
Where our data comes from
We evaluate publicly available signals only: the legal notice and the website itself, public trade-register data, WHOIS/domain information, SSL certificates, publicly visible reviews (incl. Trustpilot, Trusted Shops) and public fake-shop warning lists from consumer-protection bodies. We buy no opaque data pools and invent no metrics.
Free, ad-free, independent
For consumers, Fakeshop AI is free and ad-free. We earn nothing from rating a shop well or badly – operations are funded by the paid brand-protection monitoring for businesses and primarily cover server costs. There are no paid "buy-yourself-out" options for rated shops.
Found an error? You can object
Automated assessments can be wrong. Affected shop operators can object to a rating at any time – we review every notice.
To the correction page