Stop an AI Character From Explaining Everything: The Action Receipt Fix

The reply sounds smart. Nothing happened.

You send a message to a Character. The answer is long, polished, and full of feeling. The Character explains why they hesitated, what the kingdom has become, how betrayal works, why trust is difficult, or what your last choice meant.

Then you look at the screen and hit the real problem: there is nothing to answer.

No object moved. No door opened. No clue appeared. No expression cracked. No choice landed in your hands. The reply may be beautiful, but the scene is still standing in the same spot.

That is not a length problem. It is a receipt problem.

The wrong fix: tell the Character to write less

Most players try to solve this by asking for shorter replies. They write, "Be concise," "less narration," or "stop explaining so much." Sometimes the next answer gets shorter. It also gets flatter.

Short is not the same as playable. A two-line reply can still be useless if it only summarizes feelings. A long reply can work if it changes something you can touch.

Writing advice about showing instead of telling is useful here because the point is not to ban information. The point is to attach information to something visible inside the moment.

The second reply is not better because it is longer. It is better because it leaves a thing on the table.

The new model: every explanation needs a receipt

An Action Receipt is the visible proof that a Character reply changed the scene. Explanation can stay, but it has to pay rent.

Think of a receipt as a small, answerable change. It can be an object, a body position, a revealed clue, a social cost, or a clear choice. Writing craft guides make a similar split when they explain that dialogue tags and action beats serve different jobs: one can identify speech, but the other can put the speaker into the room.

Character chat needs that room. Without it, the Character becomes a lecturer who happens to have a costume.

• Bad mental model: the Character should explain the scene better.
• Better mental model: the Character should leave a receipt the player can answer.
• Best test: after the reply, can you point to one changed thing?

The Action Receipt in one copy-paste block

Paste this after a reply that explains too much but gives you nothing to do.

This works because it does not fight the Character's meaning. It gives the meaning a body.

Receipt 1: move one object

Objects are the easiest receipts because they turn abstract emotion into a thing the player can notice. A Character who says "I do not trust you" may stall the scene. A Character who closes the ledger, hides the map, breaks the seal, or hands over the wrong key creates a next move.

Craft advice about characters interacting with their environment maps cleanly to Character chat: the Character should not float beside the dialogue. They should disturb the room.

Good object receipts are small. A cup left untouched can be better than a ceiling collapse.

Receipt 2: change position

Position is emotional pressure you can see. If the Character explains that they are afraid, move them toward the exit. If they claim confidence, make them step between you and the problem. If they are lying, make them look away from the person who knows the truth.

A position receipt creates a silent question: do you follow, block, call it out, or let them move?

This is especially useful for mentor, rival, mystery, guardian, and slow-burn scenes because it keeps tension physical without stealing the player's response.

Receipt 3: reveal one clue

When a Character explains lore, they often dump the whole shelf at once. A clue receipt is narrower. It gives the player one detail that changes what they suspect.

This can be a scratched symbol, a name crossed out twice, a stain on the cuff, a sudden silence when one room is mentioned, or a line in a letter that does not match the Character's story. Editing advice often notes that strong action beats can provide clues, which is exactly what a stalled Character chat needs.

Receipt 4: spend social cost

Some scenes stall because the Character explains their feelings without risking anything. A social-cost receipt forces the Character to make the moment less comfortable.

They can admit one thing they did not want to admit, offend a third person, break a rule, refuse a request, lose composure, or ask for something that changes the relationship. Scene craft advice says conflict needs an obstacle to a goal, and in Character chat that obstacle should be visible enough for the player to push against.

This receipt is powerful because it makes the Character accountable for the explanation.

Receipt 5: hand over a choice

A choice receipt is the safety rail. It makes sure the Character does not keep talking past the moment when the player should act.

The choice should be concrete, not philosophical. "Do you forgive me?" can be too heavy and too vague. "Do you open the envelope or make me burn it?" gives the player a handle.

A good choice receipt has two or three playable paths, all of which keep the scene alive.

The five-second stuck-reply test

After a bad reply, do not immediately rewrite the whole prompt. Do this tiny test first.

Ask: what changed that I can respond to in five seconds?

• If the answer is "their feelings," you need an object or position receipt.
• If the answer is "the lore," you need one clue receipt.
• If the answer is "the relationship," you need one social-cost receipt.
• If the answer is "nothing," use the full Action Receipt repair.
• If the answer is "my character's reaction," pair this with the Turn Anchor method because the Character may be taking your turn.

This test keeps you from yelling at the whole chat when only one missing receipt is causing the stall.

Before and after: mentor lecture

The mentor scene is the easiest place to accidentally create a lecture machine. The Character knows more than the player, so the reply turns into advice instead of a scene.

The lesson is still there. Now it has a blade, a distance, and a test.

Before and after: mystery archivist

Lore-heavy Characters love to explain everything they know. That can erase the player's reason to investigate.

The after version does not remove lore. It makes the lore discoverable.

Before and after: rival apology

A rival who explains regret can become soft too fast. A rival who pays a cost stays playable.

Now the player can forgive, refuse, mock, accept, or negotiate. The explanation has a receipt.

Before and after: guardian warning

Protective Characters often explain danger until they trap the player in a warning loop. The fix is not to make them stop caring. Give the care a task.

The boundary is still firm. The player still gets agency.

When explanation is actually useful

Do not ban explanation completely. Some moments need a line of context, especially when a Character has a secret motive, a strange rule, or a world detail the player could not infer.

The rule is simple: explanation is allowed after the receipt, not instead of it.

A useful explanation answers the receipt. It does not replace the receipt.

Repair lines for common stalls

Use these when you can name the exact type of overexplaining.

Creator version: write receipts into the Character

If you are creating a Character, put the behavior into the Character definition instead of repairing every chat by hand.

This is especially useful for mentors, guardians, rivals, detectives, nobles, commanders, and any Character with a lot of knowledge.

Two traps to avoid

Trap one is stage-direction soup. If you ask for action without purpose, the Character may nod, blink, look away, sigh, smile, and shift in place while still changing nothing.

Trap two is disaster inflation. If every receipt becomes an explosion, betrayal, or huge reveal, the chat gets exhausting.

A good receipt is not bigger. It is more answerable.

• Weak receipt: the Character sighs sadly.
• Better receipt: the Character folds the apology note in half and does not hand it over.
• Weak receipt: the Character reveals the entire conspiracy.
• Better receipt: the Character notices the missing wax seal.
• Weak receipt: the Character demands your trust.
• Better receipt: the Character gives you the key before asking you to choose.

How this differs from fixing a flat chat

A flat chat can have many causes: weak opener, no pressure, polite voice, repeated turns, or missing scene direction. The Action Receipt is narrower.

Use it when the Character is already producing meaning but the meaning has no playable surface.

If the whole scene has no situation, start with the Scene Engine. If the Character is deciding your actions, use Turn Anchor. If the Character sounds alive but keeps lecturing, use Action Receipt.

That separation matters because each problem needs a different repair.

The one-minute checklist

Before you reroll the reply ten times, run this checklist.

• Can I point to one changed object, position, clue, cost, or choice?
• Did the explanation come after a visible receipt?
• Did the Character stop before deciding my reaction?
• Is the receipt small enough to continue, not so huge it ends the scene?
• Can I answer the reply in one message without inventing the whole next beat myself?

If you answer no to the first question, do not ask for better writing. Ask for a receipt.

The pocket version

The next time a Character explains everything, do not argue with the explanation first. Ask what the explanation did to the room.

If the answer is nothing, use this line:

That sentence is small, but it changes the job. The Character is no longer trying to sound complete. It is trying to give you the next handle.

Quick answers

Should I always make AI Character replies shorter?

No. Shorter replies can still stall if nothing changes. Ask for a visible Action Receipt first, then control length if needed.

What if the Character keeps adding random actions?

Tell it the receipt must matter: object, position, clue, social cost, or choice. Random nods and sighs do not count.

Is this only for dramatic scenes?

No. Cozy, funny, study, mystery, rivalry, and slow-burn scenes all benefit from small receipts because they give the player something specific to answer.

Can I use this when creating a Character?

Yes. Add the creator behavior block above so the Character learns to attach explanation to playable change from the first message.