AI Interrogation Question Planner

Use this AI Interrogation question planner before you enter the terminal, then update it after every suspect reply. Give each of the ten slots one job: gather, test, press, recover, or close. Mark a row planned, asked, or skipped, write the new fact in a few words, and choose the next question from what actually happened.

The planner is an organization tool, not a guaranteed answer key. AI Interrogation officially gives you ten tries, warns that too much hostility can make a suspect shut down, and warns that excessive friendliness can let the suspect lie. A plan protects the limited budget while leaving room to adapt to the current case and response.

Set up the case header first

Write the case name exactly as it appears in your current session. Add the UTC date for a Daily Case, or label the creator and visible title for a Custom Case. Then summarize the incident in one sentence. Do not preload a suspect, object, witness, or motive from another guide unless the same detail appears in your case.

Create four short notes beneath the header: the suspect’s main claim, the strongest supporting detail, the best current conflict, and the missing part of the confession. Keep each note short enough to scan while the terminal is open. A useful conflict note looks like “denied office entry / later collected documents,” not a copy of the full conversation.

Decide what a complete close needs. The official objective is a full confession, so your last request should usually ask for the action, method, and motive. If the current premise makes one of those terms irrelevant, replace it with the exact missing responsibility shown by the case.

Choose a role for every question

These roles are descriptive rather than scored. A Gather question can expose a contradiction immediately, and a planned Press question may become irrelevant after a new answer. Change the role when the conversation changes instead of forcing the original sequence.

Fill the ten-row planner

The example below gives every slot a purpose without inventing case facts. Replace each target with the time, place, person, object, or statement visible in your run. The state starts as Planned. Change it to Asked only after submission, or Skipped when a stronger route replaces it.

Do not treat the row numbers as a rigid script. If try two creates a strong contradiction, move the Test and Close roles forward. If try six still has no conflict, replace the planned pressure question with a narrower time, access, object, witness, or changed-statement test.

Copy a blank worksheet

Copy this block into a note before you join the case. Keep it outside the Roblox chat and never add account credentials or personal information.

CASE / MODE:
DATE (UTC, for Daily):
INCIDENT IN ONE SENTENCE:
MAIN CLAIM:
SUPPORTING DETAIL:
STRONGEST CONFLICT:
CONFESSION STILL NEEDS:

TRY 1 | ROLE: Gather  | STATE: Planned | QUESTION:
RESULT / NEW FACT:
TRY 2 | ROLE: Gather  | STATE: Planned | QUESTION:
RESULT / NEW FACT:
TRY 3 | ROLE: Gather  | STATE: Planned | QUESTION:
RESULT / NEW FACT:
TRY 4 | ROLE: Test    | STATE: Planned | QUESTION:
RESULT / NEW FACT:
TRY 5 | ROLE: Test    | STATE: Planned | QUESTION:
RESULT / NEW FACT:
TRY 6 | ROLE: Test    | STATE: Planned | QUESTION:
RESULT / NEW FACT:
TRY 7 | ROLE: Press   | STATE: Planned | QUESTION:
RESULT / NEW FACT:
TRY 8 | ROLE: Close   | STATE: Planned | QUESTION:
CONFESSION PARTS PRESENT / MISSING:
TRY 9 | ROLE: Repair  | STATE: Planned | QUESTION:
CONFESSION PARTS PRESENT / MISSING:
TRY 10 | ROLE: Close  | STATE: Planned | QUESTION:
FINAL GAME FEEDBACK:

REVIEW: first weak turn / strongest useful prompt / tone problem / next change

Update the plan during the run

After submitting, mark the row Asked and subtract one from the remaining budget shown by the live interface. Write only the new fact or behavior that changes your next decision. If the answer repeats an old claim, note “repeat” instead of copying it. If it changes, save the old and new versions side by side.

Skip a prepared question when its job is already complete. For example, do not ask a second access question after the suspect clearly admits entry and gives a route. Move to handling, contradiction, or responsibility. Marking the old row Skipped preserves your reasoning without wasting a terminal submission.

When a reply introduces a stronger direction, edit the next planned question rather than adding several improvised turns. One good adaptation keeps the planner useful; abandoning the sheet for three unrelated questions usually hides how much of the ten-try budget remains.

Use the result to choose the next branch

If the suspect gives a concrete fact, ask how it connects to the incident. If two claims conflict, quote both and request an explanation. If the explanation is plausible, test it once instead of treating every clarification as guilt. If the explanation creates a new conflict, replace the weaker planned question with that new conflict.

If replies become short or repeatedly refuse after hostile wording, use one Recover slot. Remove the insult, keep the strongest fact, and ask one neutral question. If replies become agreeable but never identify a personal action, stop reassuring the suspect and ask what they personally did.

If a response admits presence but not the central act, use the Repair slot for the act. If it admits the act but not the reason, ask only for motive. If the game has already shown completion feedback, stop; there is no benefit in spending unused slots simply to finish the worksheet.

Adapt it for Base, Daily, and Custom Cases

For a named base case, prefill only badge-backed identity and details you can see in the current premise. Community routes can suggest an evidence order, but you should leave an object or suspect field blank until your run matches it.

For a Daily Case, put the date and visible title at the top. Use the first attempt to map the premise, then duplicate the worksheet and shorten the opening on attempt two. Do not carry yesterday’s names or objects into a new rotation.

For a Custom Case, record the visible creator and confession target. If the premise never supplies a coherent fact to test, note that as a design problem rather than inventing evidence. A player-made case can be unclear even when your questions are well organized.

Review a failed attempt

Circle the earliest row that failed to add or test a fact. That is usually the best question to replace. Keep the strongest opening, contradiction, and close; remove repeated background, an unverified clue, or a tone-heavy accusation. The next worksheet should reach the productive conflict sooner.

Separate a route problem from a tone problem. A route problem means you never established a useful timeline, access point, object, witness, or changed statement. A tone problem means the facts were useful but the wording led to refusal or comfortable evasion. Fixing the correct problem prevents an identical second failure.

Question-planner FAQ

Must all ten rows be filled before I play?

No. Fill the opening roles, at least two evidence tests, and one complete close. Leave later rows available for the replies you have not seen yet.

What should I do when a planned question becomes irrelevant?

Mark it Skipped and replace the next slot with the strongest current fact. Do not submit it merely because it is written down.

Does the planner calculate a success chance?

No official success formula is published. The sheet tracks roles, states, remaining tries, and missing confession parts; it does not predict a win.

Should I plan an exact Case 1–5 answer script?

Plan the evidence roles, then match every noun against the live case. AI responses can vary, and a dated community route is not a guaranteed transcript.

Where should I go after filling the sheet?

Use Best Questions to improve individual prompts, 10-Try Strategy to rebalance the turn budget, and Tone Recovery when the reply pattern changes.

Rule check, August 30, 2026: the official Roblox experience confirms the ten-try limit, full-confession objective, and the two tone risks. The worksheet roles and sequence are player methods, not an official success formula.