I killed one of my favorite NPCs once. Not because the story demanded it. Not because the villain earned it. Because I had written myself into a corner and I didn’t know how else to get out.

The table went quiet. A player set down their pencil. Someone said “okay” in that flat voice that means it’s definitely not okay. And I sat there realizing that I had just spent six sessions building someone they actually cared about, and then burned that person down like a barn full of hay because I ran out of ideas for what to do with them next.

That is not a storytelling choice. That is a GM having a bad night and using narrative death as a trapdoor.

Here’s the thing nobody says out loud about NPC deaths: players are not upset because the character died. They are upset because it didn’t feel fair. And “fair” in D&D doesn’t mean soft — it means the fiction earned it. It means they had a chance, even if they failed. It means the death felt like it came from the world and not from the DM getting bored. When an NPC death lands wrong, it’s almost never a pacing problem or a story problem. It is a trust problem.

So let’s talk about that.

I believe the philosopher Mike Tyson once said that everyone has a plan until they get punched in the mouth. Game Masters have a similar (but far less violent) mantra: no plan survives contact with players. Not even a little bit.

You spent three hours this week building out the merchant quarter. You have NPC names, a pickpocket encounter, a rumor table, and a whole subplot about a corrupt city guard. You are prepared. You are ready. You are a professional. And then your rogue announces that, actually, they’d like to defect to the enemy faction they met for eight seconds in session one. Suddenly, you’re sitting behind your screen with the blank eyes of someone who has just been asked to do calculus in a language they don’t speak.

That moment has a name. It’s called the freeze, and every GM who has ever run a game has felt it. The silence stretches. You glance at your notes. Your notes don’t help. You say “interesting” like you meant to do that.

The good news? The freeze is not a skill problem. It’s a tools problem. You don’t need to become some kind of improvisational genius overnight. You just need five questions in your back pocket that work like a crowbar when your brain locks up. Questions you can ask yourself in ten seconds, flat, that give you enough to keep going.

That’s what this is. Five improv prompts that actually work, with examples of how to use them mid-session, so you can stop white-knuckling it every time your players decide the dungeon you prepped is less interesting than the goblin merchant they met on the road.

With that setup, let’s get into it.