Shift At Midnight Doppelganger Investigation Guide

The official Steam and Xbox pages do not publish a complete doppelganger answer list. You can ask up to five questions and examine the customer's ID, computer record, and behaviour. Use those checks together, but do not treat a demo customer order or one supposed visual tell as a guaranteed answer in randomized Story Mode.

For: Full release (July 22, 2026) · Updated: July 23, 2026

Key takeaways

• Use questions, ID details, the computer record, and behaviour as separate customer checks.
• The game provides up to five questions, but official sources do not require all five or publish an optimal order.
• Serve or release a customer judged human; a suspected non-human can be executed on site, followed by cleanup.
• The official pages provide no universal visual tell or complete doppelganger answer list.

Available customer checks

The game does not require these checks in a published order. Use more than one check before making a high-impact decision.

A careful investigation routine

This workflow organizes the confirmed tools without claiming to be the game's only or optimal sequence:

1. Ask useful questions. Up to five are available; use the ones that can add information for the customer in front of you.
2. Inspect the ID. Record what it actually shows before comparing it with another source.
3. Cross-check the computer. Compare available customer details rather than assuming the ID alone settles the case.
4. Observe behaviour. Keep behaviour as its own channel instead of retrofitting every movement into a guaranteed clue.
5. Make the branch decision. Apply what you found during the current randomized shift, not a memorized demo script.

This routine uses every check listed by the official pages, but those pages do not explain every anomaly rule.

The two official outcome branches

The official product description supports the high-level choice:

• If you judge the customer human, continue the service or release path shown by the current interaction.
• If you judge the customer non-human, the product description says they can be executed on site. The station still needs cleaning afterward.

A missed non-human customer can return later in the same shift as a Hunt. That makes the decision consequential, but it does not justify inventing an anomaly answer key. If the Hunt begins, use the Hunt preparation guide.

Clues not confirmed as universal

The official pages do not establish a universal rule for:

• a particular face, eye, reflection, or animation;
• one line of dialogue or one question response;
• a fixed named-customer outcome;
• a laser, mirror, or other single-check shortcut; or
• a complete list of ID and computer mismatches.

A video can show a clue in one randomized run without proving that the same clue always has the same meaning in another run.

Before making a decision

Before committing, make sure you can identify which questions, ID details, computer information, and behaviours you checked and which branch you selected.

The official pages do not explain:

• the reliable anomaly set;
• what each of the five questions is designed to reveal;
• whether the UI provides a definite pre-decision correctness signal;
• the exact penalty for executing a human; and
• whether any clue rules vary by shift, customer, platform, or patch.

New players should start with the beginner guide; station duties are covered in store and quota, or return to Guides.

Quick answers

How many questions can I ask a customer?

You may ask up to five questions. Official sources do not say that every customer requires all five or that one question order is always best.

Can one visual clue prove a customer is a doppelganger?

The official pages do not say that one visual clue is always conclusive. Use the available questions, ID, computer record, and behaviour together.

What can happen if a non-human customer gets away?

A missed non-human can return later in the same shift as a Hunt. This is a possible consequence, not proof that a particular clue or decision was correct.