The same move, but with different feedback displayed

As a beginner, which analysis is better to use for reviewing my games? When I use the analysis view, it shows castling as a good move, but when I check the review, it says it’s a mistake. I’m a bit confused

Yes. It can be confusing. The computer is trying to answer a wildly complicated question.

The computers gave two different lines after castling. The position is only at move 8 in an opening (Caro-Kann) which is still very complicated when you attempt to look out more than a few moves.

The descriptive evaluations (good, mistake) look at the change in estimated winning chances. (This is somewhat involved. You should read the Help topic on it. https://support.chess.com/en/articles/8572705-how-are-moves-classified-what-is-a-blunder-or-brilliant-etc. )

They show two different depths of search. (24 ply on one; 26 on the other) One shows Stockfish 17. The other doesn't show what engine is used.

One retreats a bishop to e3 after the bishop is challenged with h6. This results in a plus 1.04 evaluation.

The other trades the bishop for a knight on e7. This line results in the +1.55 evaluation.

I can't see beyond the first several moves but I would attempt to get the same evaluations and follow the shown lines to see what the computer was looking at.

Okay! As a beginner, which analysis would you recommend I use? I feel like I shouldn’t go too deep into the analysis at my level, but I still want something fair and helpful to improve my games. What do you think?

I am not sure what to tell you. The cloud analysis goes with a different move, 8 Bh6.

I think that you should explore it on your own.

In general, avoid the game review, not only because of this. It tends to give not so useful comments that don't have anything to do with reality. Game review is work in progress, the aim is for it to be able to tell you why some move is good or not, but this is something game review can't really do at present time, in most cases.

Terms good move, inaccuracy etc are just interpretations of what is the engine evaluation of the position. You should generally forget about them. The only interest you have is how much the evaluation shifted after a certain move. I think that the difference here is that analysis uses bigger depth than game review.

All that being said, castling here is completely fine. The reason why Bh6 is better is because if black takes, it invites your queen on h6, and can't castle. And if he castles instead, you have some kind of an attack with h4 and so on. But to figure it out, you either have to have a bit more knowledge, or to go through the engine lines yourself manually, and sometimes both are required.

In any case, you shouldn't take terms like good move, excellent move etc to heart but look at the evaluation yourself. Use analysis, try to figure things out yourself and later with the help of the engine evaluation as well, and generally don't take this "AI coach" inside the game review too seriously.
Perhaps in many years it could, to some extent, replace human perspective, but that day is certainly not today.