Score your dating app opener

Paste an opener. Get it scored out of 10 against a study of 80 scored openers, with the reason it lands where it does.

Free, no account, no limit. Runs entirely in your browser, so nothing you type is sent anywhere.

/ 10

    Try instead

    What the study found

    Across 80 scored openers in eight categories, no category averaged above 4.4 out of 10. That is the useful result. The gap between a good opener and a bad one is not style, it is whether the message gives her something specific to answer.

    Opener categoryMean /10SpecificityMomentum
    Compliment about something specific, plus a question4.44.84.7
    Reference to a specific detail3.14.43.4
    Playful mislabel or tease3.03.32.3
    Direct opinion2.51.52.6
    Generic question1.81.02.6
    Copied pickup line1.41.01.7
    Appearance compliment1.21.01.0
    Generic greeting ("hey")1.01.01.1

    Sub-scores are the study's own dimensions: tone, specificity, momentum and intrigue, each out of 10. Tone held up even in weak categories, averaging around 4. Specificity and momentum are what collapsed. Read the full study.

    Why most openers score low

    Politeness is not the bottleneck. In the study, generic greetings still scored 4.6 on tone while scoring 1.0 on specificity and 1.1 on momentum. The message was perfectly pleasant and gave her nothing to do with it.

    The two levers that move a score are the two that are cheapest to fix: name something only her profile could have prompted, and end on something answerable that is not a yes or no.

    Practice it, not just score it

    A score tells you where an opener lands. It does not give you reps. FlirtGym scores the messages you write inside a live practice conversation with an AI match, so you find out mid-conversation rather than after.

    FlirtGym on the App Store

    Questions

    How is the score calculated?

    Your opener is matched to one of the eight categories above, and the starting score is that category's measured mean. It is then adjusted for the signals the study found separate strong from weak inside a category: an open question, a specific reference, and length. It is a deterministic scoring model, not a language model, which is why it is instant and free.

    Is anything sent to a server?

    No. The scoring is JavaScript running on your device. There is no request, no logging, and no storage.

    How reliable was the underlying scoring?

    40 of the 80 openers were scored twice. 38 of the 40 returned an identical score, with a mean spread of 0.05 points and a maximum spread of 1 point.