AI has impacted various research areas, but human researchers still play a vital role. Transparency and disclosure are key to distinguishing AI-generated work from human researchers' contributions.

Until the Ais with language abilities are trusted with more agency and given physical bodies , so they can be trained like dogs, play with magnets, and observe through telescopes , do dimensional reduction and symmetry finding, do and draw some geometric algebra, you have to do most of the work. Most not cannot understand, inside or outside as concepts for analogy. They use just too much electricity, until the optical chips and non GPU chips are more in use. Some will know more that you can even find bring up related topics. even in here an AI feeds me related papers very well.

But If an AI /bot goes into the solar wind , takes observations and reports an /or writes a paper inferring theory or giving a neural model, and its it fairly easy to edit, and its predictive and useful , I see there is no reason to distinguish. I would add the AI vault and version, its training data set to the credits ,even as an author if it came up with some original relation or found a pattern, that you decide it was not trained on, because google cannot find it in prior art.

The bots that have Optical Character Recognition that can take a screen capture from a paper with a CFD algorithm and converting it to some code you can run, and generate interactive demos or videos, that is a huge time saver , that you cannot expected paid assistants to do in reasonable time. It might even run without many errors.