Skip to main content
The Quality tab in study Design combines configuration checks with AI review of the topic guide. Use it to find issues before recruitment and to reassess a guide after editing. Quality review requires access to the editable study design. Automated review supplements your research judgment and test interviews; it cannot prove that a question is unbiased or every participant path works.

​
Run and read checks

  1. Open Design → Quality.
  2. Select Run checks and wait for processing to finish.
  3. Review the findings by study, section, research objective, or question as displayed.
  4. Use the severity filters to focus the list, starting with blocking issues.
Configuration checks cover requirements such as valid questions, routing, and feature settings. AI checks can identify wording or methodology concerns, such as leading, ambiguous, or overly complex questions. Read the explanation and the affected content before accepting a recommendation. If Validation outdated appears, the guide has changed since the last review. Run checks again; the earlier results may no longer describe the current design. A failed or unfinished check is not a clean bill of health.

​
Correct a finding

Edit the affected content directly, or choose Fix with AI where offered. The AI action opens a conversation asking the design assistant to address the finding. Review the resulting pending changes before accepting them. An Addressed badge means the assistant was asked to fix the issue. It does not certify that the issue has been resolved; the next validation checks that. Some configuration issues do not offer an AI fix and require a manual correction.

​
Dismiss a finding deliberately

Eligible non-blocking AI findings offer two dismissal choices:
  • Dismiss until …: hide the finding until the relevant content changes.
  • Dismiss permanently: keep the finding dismissed across later checks, including content changes.
Expand Dismissed issues to review earlier decisions and choose Undismiss when a finding should return to the active review. Dismissing a finding does not change the question or fix its underlying behavior. The helpfulness controls let you give feedback on a finding. Submitting written negative feedback on an eligible non-blocking AI finding also dismisses it permanently. Blocking and configuration findings do not become dismissible through feedback.

​
Complete the review

After accepting edits, rerun checks and test the relevant participant paths, including routing, randomization, and language variants. Resolve remaining blocking findings and review warnings in context before launching. Keep the guide’s background documents current so the assistant has the right research context. Quality review checks the design; it does not retrospectively change interviews already collected.