To be sure, clinically, thresholds are useful because services, insurance, and research all rely on clear binary classifications, in our current society. But outside that context, it isn’t obvious that everyday language needs to mirror that line. Self-identification can be a way of making sense of one’s life, not an attempt to claim a clinical label.
But at the same time, the spectrum includes people with very high support needs, and there’s understandable concern that broad or casual uses of the term can hide those realities in ways that impact care.
To me, none of this means people are wrong for wondering about autism. And I do not have the experience to advocate for or against "anxiety disorders" being weighted more heavily in clinicians' priors than they currently are (as the OP article heavily implies with its length). I mean only to highlight the mismatch between a binary diagnostic system and a very heterogeneous spectrum, and the need for language that acknowledges self-understanding without flattening anyone’s experience.
It seems in the DSM 5 the definition was narrowed specifically to focus on two performance deficits: 1) immediate harms either to the inflicted or to those they interact with due to social interactions, 2) catastrophic academic failures due strictly to input/output perception irregularities not otherwise explained by neuro-transmission disorders (things that can be treated with drugs) or low intelligence measures.
That excludes a massive host of social and perception abnormalities that do result in less immediate social rejection and abstract reasoning failures.