Nonverbal behavior is objective "evidence". Then it's mainly interpreted by N, and not by T as is mainly with common behavior and words. N is one of 4 equally important mind functions to deal with the reality.
Typing by intuitive impressions from nonverbal behavior (VI) is one of methods existing in Augustinavichiute's texts. She described such traits of types. Intuitive feeling of types traits is popular approach. Initial
info about VI.
It's not against normal psychology that personality traits can be expressed in nonverbal behavior specifics same as in common behavior specifics. Unlike with physiognomics VI which uses static body traits, what is much doubtful.
That intuitive impressions from nonverbal behavior is working typing method was objectively proved in my
experiment of 2015. It gave real typing matches (by random typers on random youtube bloggers) comparable to IRL interview and much higher than accidental (~17% > 6%).
So "NOT a valid" is your supposing results of this method as not meaningful.
In case you want be typed - make a video, so your nonverbal could be seen.
Besides lack of a part of important data, those who know types theory may semi-intentionally (having wishes to have some types, for example to fit better an occupation or better IR with someone) significantly falsificate such data what reduces the chance for correct type. These 2 reasons together make typing of not well-known personally people by Internet without video as not reasonable practice. Sometimes photos are good enough too, mainly when those are many.