Assessment design
Randomized pools, focused timing and applied questions help make outside assistance harder.
Reduce the risk of undisclosed AI assistance with advanced test design and reviewable signals. Explore the simulated Integrity Score to see how different signals change the result.
No simulated concerns. This is not a guarantee of unaided work.
Toggle a signal to see what changes.
Did an interruption or accessibility tool move focus?
Was the task familiar, misunderstood or interrupted?
What tools were permitted for this task?
What other evidence supports the concern? Detectors are unreliable.
The approach brings together prevention, evidence and a human response.
Randomized pools, focused timing and applied questions help make outside assistance harder.
Focus, timing and other contextual signals help the reviewer ask specific questions.
Give the candidate a chance to explain. One signal is never proof of misconduct.
Assessment design combines randomized question pools, per-question timing and formats that ask candidates to apply information. Single-use invitations and server-side delivery help limit exposure of test content.
Some assessments may restrict copying, pasting or leaving full-screen mode. Those settings must be explained before a candidate starts. They reduce opportunities for assistance; they do not establish that assistance is impossible.
Depending on enabled settings, integrity evidence can include browser focus changes, paste events, typing bursts and unusual answer timing. Optional webcam snapshots may support face-present checks when enabled with appropriate notice and consent.
Device and network patterns, similarities between answers and AI-text likelihood can contribute context. AI-text likelihood is a weak signal because detectors are unreliable. Face matching is a later-release capability, not a live promise.
The illustrative score runs from 0 to 100 with three bands: High confidence, Review suggested and Serious concerns. Higher scores indicate fewer concerns under the configured rules, not certainty about what happened.
The interactive demo uses illustrative weights. Production rules require versioning and validation. Never reject a person automatically or treat a single signal as proof of misconduct.
Browser-based assessments cannot reliably detect every second device, overlay or outside helper. We do not claim to detect invisible overlays. Normal interruptions, assistive technology and network problems can also create signals.
Ask for context, review the actual work and give candidates a fair opportunity to explain. The goal is better evidence, not an accusation generated by software.
Camera features, where enabled, require appropriate notice, consent, restricted access and a disclosed retention schedule. No camera or identity data is collected by the interactive demo on this website.
Employers must communicate the retention schedule, applicable monitoring settings and alternative assessment routes before participation. Read the Candidate Privacy Notice and Biometric Data Policy.
When AI-assisted applications make screening harder, a score summary may not be enough. Compare integrity evidence and reporting in Starter with the Free plan.