How to Choose Truth or Dare Questions for the Group You Actually Have

Choosing a Truth or Dare prompt well comes down to one question: is everyone in this room risking about the same amount by answering it? A question that costs six close friends nothing can cost one person at a work night out a great deal, and the prompt itself is identical in both rooms. Group fit, not prompt quality, is what decides how the evening goes.

That is inconvenient advice, because it means no list is universally good. Plenty of articles will hand you three hundred questions. This one is about the part that happens after you have the list — deciding which of them belong in front of the specific people sitting in front of you, and rewriting the ones that are nearly right.

If you want the banks themselves, they already exist and they are sorted by audience: the general question bank, the friends deck, the couples deck, and separate teen and kids decks. Come back here for how to filter them.

Three tests every prompt should pass

These take about two seconds each once you have used them a few times, and between them they catch most of the prompts that cause problems.

1. The reciprocity test — would you answer it?

If you would not want this question directed at you tonight, in this room, it does not go in the pile. This sounds obvious and it is routinely violated, because the person assembling the prompts is usually thinking about what would be entertaining to hear rather than what would be reasonable to answer.

The test also catches something subtler: a question you would happily answer because you already know your answer is a good one. If the honest response to it makes you look interesting, you have written a prompt for yourself rather than for the group.

2. The symmetry test — does everyone risk the same?

The core one. A prompt is well chosen when the cost of answering it honestly is roughly equal around the circle. It fails when one person has substantially more to lose, and that person is almost never the loudest voice in the room.

Asymmetry creeps in through several doors. Someone in the group is single and everyone else is paired. Someone is a colleague rather than a friend. Someone is fifteen years older, or is the only person who does not already know the story being referenced. None of these make a question unusable — they make it a question that needs checking before it goes in.

3. The residue test — is it over when the turn ends?

A good prompt is finished when the laughter is. A prompt with residue leaves information in the room that someone has to live with afterwards, and the game gets blamed for it long after anyone remembers what was actually asked.

Performed dares are almost always residue-free: an accent, a dance, a daft photo. Disclosure questions vary enormously. "What is the pettiest reason you have stopped talking to someone" is fine, because the pettiness is the joke. The same question about someone in the room is a different object entirely.

Fastest version of all three: would you be comfortable if this exact prompt came up on the turn of the quietest person present? If yes, it is fine for everyone.

Rewriting a prompt that is nearly right

Most prompts that fail are not bad ideas. They are reasonable ideas pointed at too wide a target, and narrowing the scope usually fixes them without losing what was interesting in the first place.

The pattern is consistent enough to be worth learning as a move: keep the subject, shrink the claim.

Three of those six were fixed by moving the target off a person in the room and onto the player themselves, which is the single highest-yield edit available. The interesting content survives; the thing that would have soured the evening does not.

If a prompt needs the group to be in a good mood to be survivable, it is a prompt about the group's mood rather than about the person answering.

The group-fit matrix

Two properties determine which bank a group can use, and they are close to independent of each other. How much shared history exists, and how evenly the risk is spread.

The top-right quadrant is the only one where you can hand the group a list and stop thinking. Everywhere else, the list needs filtering, and the two quadrants along the bottom need it most, because a group with plenty of history can lull you into forgetting that the risk is not shared equally.

The couples row is worth expanding, because it is the one that goes wrong quietly. Once a couple is present, a surprising share of prompts acquire a second audience, and a group that does not notice this ends up playing a game about the relationship while four other people watch. Keeping couple-specific prompts in their own pile and drawing from it deliberately is the whole fix. If that is the actual point of the evening, the couples deck is built for it.

Getting the intensity curve right

Groups tend to fail in one of two directions, and they are not equally common. Opening too hard is the famous mistake. Flatlining is the frequent one.

The orange failure is loud. Someone opens with something enormous, the room laughs nervously, and every subsequent prompt is an anticlimax — you have spent the ceiling in the first two minutes. Groups usually recover from this, because it is obvious enough that somebody says so.

The grey failure never gets named. Every prompt is safe, nobody is uncomfortable, and the game just fizzles after twenty minutes with everyone vaguely agreeing it was fine. This is much more common, and it is what happens by default when nobody has thought about ordering at all.

The practical version of the green line: roughly three light prompts to each heavier one, escalating slowly enough that no single prompt is a big step up from the one before it. The size of the step matters more than the height reached.

Dares follow different rules

Almost all published advice about choosing prompts is really advice about choosing truths. Dares behave differently enough to need their own filter, and applying truth logic to them produces the two classic bad dares.

The useful property of a dare is that it produces no information. A player who does a terrible accent for ten seconds has disclosed nothing about themselves, which is why dares work in groups where truths do not — new groups, work groups, mixed ages. Shifting the ratio toward dares is the single most effective adjustment for a group you are unsure about.

That property is destroyed by exactly two things, and they are the two things bad dares have in common.

1. Dares that require disclosure. "Show us your camera roll", "read out your last message". These are truths wearing a dare's clothing — the player is being asked for information, and the fact that it arrives as an instruction rather than a question does not change what it costs.
2. Dares involving someone outside the room. Anything that sends a message, makes a call, or involves a person who did not agree to play. The cost of these lands on somebody who is not there and gets no say.

Worth saying out loud before you start: dares involving other people's phones, other people's messages, or anyone outside the room need that person's agreement first. "It was a dare" is not consent, and it is not a defence the following morning.

The pattern in the dares that consistently work: performed rather than revealed, over in under a minute, and the embarrassment is shared rather than aimed. A dare that leaves one person alone in the spotlight for five minutes has stopped being a game and become a performance nobody asked to give.

Length is underrated as a filter. A ten-second dare is almost always fine regardless of content, because the audience attention it attracts expires before it can become uncomfortable. Anything over about a minute needs to be genuinely funny to justify the airtime.

Reading the room mid-game

No amount of preparation survives every group, so the skill that actually matters is noticing when a round has gone past where people wanted to be. The signal is smaller than most people expect, and waiting for something obvious means waiting too long.

• Answers get shorter. The clearest tell. Someone who was giving three sentences gives four words.
• Laughter arrives a beat late and stops abruptly instead of trailing off.
• Prefacing appears. "This is boring, but…" is almost always a request to be passed over.
• Posture opens outward. Someone who was leaning in is now leaning back, or has picked up their phone.
• Someone starts answering for other people — a protective move, and usually the most socially attentive person in the room doing it.

The response is deliberately unglamorous: take the next turn yourself, pick something light, and let the level drop without commenting on it. Naming the discomfort out loud makes it the centre of attention, which is precisely what the person who flinched is trying to avoid.

If you only remember one thing: move, do not discuss. A light prompt taken by the host fixes the room in about fifteen seconds. An apology takes five minutes and makes it worse.

A shortcut for the work-party case

This one comes up constantly and almost every published prompt bank is unusable for it, so it is worth a section of its own.

The reason is not that colleagues are more easily offended. It is that a group of friends is roughly symmetrical — everyone risks about the same by answering — and a work group is not. People have different amounts of seniority, different job security, different amounts riding on how they are perceived, and the person with the most exposure is rarely the one deciding what gets asked.

A bank that works has three properties. Everything is about preference or opinion rather than history. Nothing touches relationships, family or private life. And there is a single line that does not move: if an honest answer could plausibly change how someone is treated at work afterwards, the prompt is out, regardless of how funny it is.

The one to watch: "harmless" questions about salary, promotions, or what someone thinks of a decision made by someone present. These read as fun and are the ones with the longest tail.

Dares are much safer ground here than truths, because a performed dare produces no information about the person doing it. If in doubt, shift the ratio hard toward dares and keep them short.

Knowing when to stop

Truth or Dare has no end condition. There is no final score and no last round, so every game of it ends either because someone calls it while the room is still loud, or because it slowly dissolves into side conversations and phones.

The second version is far more common, and it does quiet damage. A group whose last memory of the game is twenty minutes of it not quite working will be less enthusiastic next time somebody suggests it — not through any decision, just through the feeling now attached to the name.

So the ending has to be chosen. Watch for the round that gets the biggest reaction of the night, let it finish, and stop there. Someone will ask for one more. Declining that is the most useful thing a host does all evening, and it is harder than it sounds because the room is enjoying itself, which feels like an argument for continuing.

A version that removes the awkwardness: decide the length before you start — "we will play for about forty minutes" — and let the clock take the blame. Nobody has to be the person who called time.

Where to get the prompts themselves

Everything above is filtering. For the raw material, these are sorted by audience so most of the filtering is already done:

• The main question bank — the general-purpose set, mixed intensity.
• Questions for friends — assumes shared history, leans on it.
• Questions for couples — the separate pile referred to above.
• Teen questions and kids questions — age-rated rather than simply milder.
• The rules and safety guide — pass rules, turn order and the setup decisions this article assumes you have made.
• The forfeit ladder — what should happen when somebody declines, which is the other half of choosing well.

If you would rather not assemble anything, the game itself deals prompts by age rating and intensity, keeps a pass rule built in, and works either by passing one phone around or by opening a room everyone joins from their own. Playing remotely has its own failure modes, which the guide to playing over text covers separately.