Also, the author using speech to text could pay a person to dictate for them or could use other tools. But I think overcoming physical disabilities and making things more accessible is one of the better use cases for AI. As I understand it, twitch and youtube use generative AI for captions.

I was mostly thinking about the awards and the award eligibility surveys for SFWA she was talking about in the link you posted. Presumably there are blind authors who have participated or want to participate.
But yeah, it is only close to impossible. Using google search nowadays means using gen AI.

Part of the point she is making, that resonates with me, is that it is getting close to literally impossible to avoid using generative AI.
If you are dictating a text, e.g. because you are blind, then you're also using generative AI.
So it is getting harder to separate human and computer output.