Methodology

Nothing up our sleeve

This page is the complete scoring rubric for the "Am I dating a bot?" screener — every item, its weight, its source, and exactly how answers become a result. If a future screener on this site can't publish a page like this, it doesn't ship.

What this instrument is

A 16-item screening checklist for romance-scam and AI-persona risk factors, in three sections: how they talk (conversation tells), the story they tell (narrative shape), and money (extraction behaviors). Each item is a pattern documented by the FTC or FBI in their consumer-protection material, or by large recent surveys on AI companionship. It is an evidence-derived checklist, not a validated psychometric scale — see limitations below.

How scoring works

Each item is answered yes (full weight), not sure (half weight), or no (zero). The weighted sum places the result in a band — but money items also carry hard floor rules, mirroring the FTC's framing that a money ask is the defining signal rather than one flag among many:

High — the "pay fees to withdraw" item is yes, or 2+ money items are yes, or the weighted sum ≥ 25.
Elevated — any single money item is yes, or the weighted sum ≥ 15.
Caution — weighted sum ≥ 7.
Low — everything else. Explicitly not a guarantee.

The items, weights, and sources

Limitations, stated plainly

· This instrument has not been through a formal validation study. Items and weights come directly from published FTC/FBI documentation, but the thresholds are editorial judgment, not fitted parameters.
· It is self-report about someone else's behavior, answered by the person with the least distance from the situation.
· Risk factors are not proof. Real people are sometimes deployed overseas, bad at video calls, and fast to say "I love you". The score means "verify", never "convict".
· A low score is not clearance. Patient scammers show few flags early — which is why the money rules floor the result regardless of the total.

Privacy