AI assisted patches and credits where it is due

I sometimes get patches, which are literally just LLM output dumps.

Even finding out if there is a bug symptom/bug has been delegated to
me.

I'm thinking of followning process when this happens next time:

1. Do all the real work.
2. Retain whatever model was used as co-author.
3. Downgrade submitted from author to reporter.

I've seen also reasonable AI assisted patches and this type of behavior
just downplays them.

Any better ideas? I'm open.

BR, Jarkko