Appendix C. Exam day guide

Before you start

Prepare a quiet location, stable internet and a charged device.
The exam has 30 questions in 45 minutes: an average of one minute 30 seconds per question.
Check that you have passed question 11 at 15 minutes and question 22 at 30 minutes. Reserve the last five minutes for flagged questions.
Analysis and judgment questions carry three points, so marking the conditions and reading to the end is better for your score than skipping them because they are long.

Read the question first. Check for “most appropriate,” “not appropriate” and “choose two.”
Mark every condition. Bold conditions in scenarios—what alone is needed, what must be controlled, who must not see what—determine the answer.
Apply conditions one by one. For each option, check that both condition A and condition B are satisfied. An option meeting only one is a trap.
Use official definitions for product names. Do not infer from everyday meanings of role or setting names. Even the lowest sharing role can see materials, and ChatGPT app permission settings only change when confirmation is requested; they do not grant or revoke data access. Check each setting's effect against its official definition.

Look at what an option relies on, rather than its length or tone.

What to check	Common assumptions in wrong answers
Control mechanism	Instructions substitute for permission control
Permission scope	Hiding or low roles are treated as blocked access; shared-credential scope is overlooked
Timing	Setting changes are assumed to revoke existing shares and connections
Evidence	Citations or completion reports are treated as proof of accuracy
Approval	High accuracy is used to remove approval for hard-to-reverse actions, or approval is placed at every step
Remedy	Model replacement or emphasis is proposed without identifying the cause
Data	Training use is equated with retention/deletion; business-plan exceptions are missed
Scope	More app connections than needed, or a high delegation level from the start

Single choice: the longest or most cautious option is not necessarily correct. Select an option only if it satisfies all stated conditions.
Multiple choice: judge each item in the <Statements> box separately, then choose the combination containing all and only correct items.
Statement judgment: verify the reason following “because” separately. A correct conclusion with an incorrect reason is not a correct statement.
Ordering: follow each step's inputs to determine order. Analysis and criteria setting come early, full rollout late; harm containment precedes cause analysis.
Matching: eliminate certain matches first. If told there is an unused option, first find the option meeting none of the items' core requirements.