Any AI would have difficulty mimicking my writing as I am dysgraphic. That means that my writing ability is really clunky with repetitions, missed words, chronic misspellings, along with serious flaws in linear process order over an entire post/article. That means I have to edit and re-edit. Any AI would have to contemplate whether to correct mistakes or leave them with mistakes. If it corrects then it would be a giveaway that an AI were doing it. If it were to create errors I find it unlikely that it would make the "right" type of errors and look like they were considered/artificial.
All types of bio-id are really stupid. You only have one of set of each. If they get compromised then that is your lot - you cannot ask for a new set. I remember laughing my head when fingerprint id came out on phones. What are phones famous for being? Fingerprint magnets. It's like covering your desktop monitor with post-it notes of your password. Either the the silicon valley bros are stupendously dumb or it was a data-collection exercise.
As for Iris scans - well printable contact lenses have been around at least 20 years and smart contact lenses using nfc are under development. Pretty much anything can be projected on them.