01
Why open-ended is worth protecting
Evidence, judgment, and reflection show up more clearly in open responses than in forced-choice items. That value disappears if answers are ghostwritten. Organizations that retreat entirely to multiple-choice to avoid AI risk also retreat from the richest hiring signal. The better path is to keep open-ended formats and harden integrity around them.
- Richer demonstration of ownership, judgment, and specificity
- Explainable excerpts for hiring managers
- Better job-relatedness story than many self-report items
- Requires authenticity checks in a generative-AI world
02
The attack surface
Candidates may paste prompts into chatbots, use stealth interview overlays, receive live coaching, or submit impersonated sessions. Fluency is cheap. Specificity and consistent lived detail are harder to fake-but not impossible. Integrity programs should assume motivated candidates will try modern assistance tools, including named copilots buyers often search for.
- Pre-written AI essays pasted into take-homes
- Real-time copilots during live interviews
- Scripted answers that sound polished but lack personal constraint detail
- Impersonation and proxy completion risks
03
Design defenses
Use adaptive follow-ups, consistency probes, timed segments where appropriate, and integrity monitoring aimed at assistance tools-not invasive keystroke theater by default. Design the assessment so shallow ghostwritten answers struggle: ask for constraints, tradeoffs, and what the candidate would change. Freeze forms so you can compare candidates fairly while iterating defenses across waves.
- Follow-ups that demand specifics beyond a generic STAR template
- Cross-question consistency checks
- Integrity monitoring for stealth assistance patterns
- Clear candidate rules for preparation vs live help
04
Score authenticity and content separately
A strong-sounding answer with integrity flags should not be treated the same as a strong authentic answer. Show both to reviewers. Collapsing everything into one score hides the decision. Verified behavioral assessments work when content quality and authenticity are visible side by side-and humans decide.
- Content scores against behaviorally anchored rubrics
- Integrity observations with evidence timelines
- Human override and retest paths