Why is this move a blunder?

I just played a game and found a Blunder in the Analysis:

27. Ng5 threatening Qh7# should the knight on f6 ever move - those locking the rest behind him in place. I thought this is a great position, because it buys me time to lift the rook and finish the game in style...

But no, the engine says wants 27. Ng5 is blunder and prefers e5 instead bringing more pieces to the game but also unlocking the position for black. Does anyone see what I am missing? Why is 27. Ng5 when my plan looked so clear cut and dry (and still does) to me?

Pay attention to your opponent's defensive plans too.
Ng5 allows black Ne5 and Re3 will be met by a knight fork on g4.
e5 is better because after dxe5, Ng5 works now. The black Knight can no longer go to e5

The purpose of e5 is twofold: One is of course to open up your bishop, but the other is to free the e4 square for your knight. For example, 14. e5 dxe5 15. Ng5 with Ne4 coming next. Black can't stop it. If 14. e5 Nxe5 instead, then you can play 15. Rxe5 dxe5 16. Ng5 and it's the same thing, or 15. Nxe5 dxe5 16. Re3 with a similar mating attack.

The reason your move 14. Ng5 is not as good is because Black could've played 14... Ne5, which not only stops you from playing e5, but also stops 15. Re3?? because of 15... Neg4. You're still better after 14. Ng5 Ne5, but you need to find 15. Be2 to stop ...Neg4, and your attack is stopped (for now).

[Site "https://lichess.org/i56kgUJc"] [Date "2026.06.13"] [White "dirkster99"] [Black "TobiiDynavox"] "14. Ng5?? (7.30 --> 2.71) Blunder. e5 was best"
This issue arises regularly. For better or for worse, the machine seems to be willing to use the word, "blunder", even if the chosen move is a likely winner. The machine was reacting to the availability of a move with a much higher evaluation.