Like many have said, I try to stick with things I am certain of, to avoid scenarios of “the blind leading the blind”. Especially since this can all end-up as training data for the image recognition system, I want to only be as specific as I am confident of an ID. For example, I just saw some kind of thistle, I recognized it from the structure of the blooming flower, I double checked images of thistles in bloom, and the teasel which I’ve seen in the area, and I was quite confident it was some kind of thistle. However, since I couldn’t guarantee what type of thistle or related variety it may have been, I went up to the Family. That way, at least someone who is familiar with that family will be able to see it, review it, and possibly identify it appropriately.
Similarly, per someone’s suggestions for non-experts (https://forum.inaturalist.org/t/useful-inaturalist-tasks-for-non-experts-wiki/35034), I have been going through observations listed as Needs ID / Unknown, and identifying them by kingdom. While it might seem silly to identify something as a plant, remember that you’ve just eliminated billions upon billions of other organisms from the other major kingdoms. It’s actually a pretty huge first step in identifying an unknown specimen, and brings it to the attention of well-studied specialists who may be able to give a more specific identification.
Of course, even in cases where I’m being more specific, (as in the case of the thistle) if someone who is a specialist in that family of flowers has a different ID from a commonly misidentified plant, they can then offer a better ID–so I’m not super worried about being wrong, because there’s still a chance for it to be corrected by the community–but I do my best to be correct.
TL;DR I only identify to the degree of specificity that matches my confidence level.