Generative AI Leader Practice Question: A marketing team wants to generate multiple…

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✗

  Use a zero-shot prompt with a detailed description of the brand voice

  Why it's wrong here

  A zero-shot prompt with a detailed brand-voice description guides tone but supplies no worked examples, so consistency across many generations is weaker than example-driven prompting. Zero-shot suits straightforward tasks where the desired style is easily stated; repeated A/B variants benefit from few-shot exemplars demonstrating the tone.

• ✗

  Use a system instruction that says 'Be consistent'

  Why it's wrong here

  A bare instruction to 'be consistent' gives the model no brand attributes to anchor on, so tone still drifts between outputs. System instructions are the correct mechanism for persistent behavioural constraints, but only when they carry concrete style rules, vocabulary and examples rather than an unqualified directive.

• ✓

  Include few-shot examples of previous ad copy that match the desired tone

  Why this is correct

  Few-shot examples embed concrete instances of the desired brand tone directly in the prompt, giving the model a pattern to imitate across every generated variant. This anchors stylistic consistency more reliably than descriptive instructions alone, satisfying the requirement for uniform tone across all A/B test outputs.

• ✗

  Set the temperature to 0.0 to eliminate randomness

  Why it's wrong here

  Temperature 0.0 makes sampling deterministic, so every run returns the same copy — defeating the purpose of generating multiple variants for A/B testing. Low temperature suits tasks needing reproducible, factual answers; here variety across outputs is required, so the setting directly contradicts the scenario.

Go deeper

Related to this question