Are AI-based tools any good for generating API tests yet?

I keep seeing AI testing tools pitched as a way to stop writing test cases by hand, and I can't tell how much is real and how much is hype. My API test suite is the least fun part of the job to maintain, so I'm motivated, but also skeptical.

The thing I actually care about is whether the generated tests reflect what the API really does, or whether they're just plausible-looking cases derived from a spec that drift from reality. Those are very different in practice.

I went through this overview of generative ai testing tools which helped me sort the categories, but I'd rather hear from people who've actually run this on a real project.

Has anyone had AI-generated API tests actually catch real regressions? And do they hold up as the API changes, or do you end up maintaining them just like hand-written ones?