Skip to main content
Use change review to decide which AI suggestions become part of your study. Use History to inspect recorded design versions and, when eligible, restore an earlier design.

Review proposed changes

The study design assistant can propose additions, edits, and deletions. Inspect the changed fields in the guide and accept or reject the proposals individually, or use the available bulk controls after reviewing the full set. A proposal shown in a conversation is not the same as a saved design change. Verify the resulting guide after acceptance. Related changes can depend on each other, so check the whole question, section, or stimulus rather than only its highlighted text. Direct manual edits use their own save controls or autosave status. Do not assume every keystroke appears as a separately restorable History version.

View history

Open History in Design to inspect recorded versions and compare their changes. The version toolbar also supports undo and redo when available. Viewing an older version does not itself restore it. The history records authoring state, including sections, questions, choices, routing, translations, and template variables. It is not a backup of every account setting, recruitment transaction, or research output. Accepted study-design assistant changes record a version; history should not be treated as a complete audit log of every action in the product.

Restore an earlier design

Restoration requires study editing permission. Choose a version, review the proposed differences, and confirm Restore this version only when the preview matches the intended change. Restore applies the recorded design as a whole, rather than selectively undoing one field. The current design is recorded before restoration, so another eligible restore can return to that state. Restoration is blocked if any non-test interview record exists for the study. The restriction is broader than completed interviews: an ongoing, abandoned, hidden, or deleted non-test record can also prevent restore. Removing visible responses from the grid does not make restoration safe or necessarily available. A version referencing an asset that no longer exists cannot be restored. Review the blocking explanation instead of treating history as a way to recover a deleted media file.

Changing a live study

Editors can unlock a topic guide where the controls permit it. A warning appears when completed non-test interviews exist. Unlocking is permission to edit the design, not a guarantee that existing interviews will be restarted or updated uniformly.
  • New interviews use the design available when their interview flow is initialized.
  • An existing interview has an initialized question sequence and stimulus assignments. Do not assume adding, reordering, or randomizing content rebuilds that sequence.
  • Section and question routing read current rules when pending questions are evaluated. Moving a question can change which section rule applies to that evaluation.
  • Some content is translated, generated, or cached separately. A saved edit does not prove every participant screen or already generated output has refreshed.
  • Completed responses are not rewritten into answers to a newly worded question. A comparison across the edit can combine different participant experiences.
Use a new study when a substantial change needs a clearly separate research population or questionnaire. For a necessary live correction, record what changed, test the affected path, and review results with the change in mind. Question types and other settings can have additional restrictions after answers exist. Follow the editor’s validation and the relevant routing, stimulus, and recruitment guidance.

Recruitment changes from the design assistant

Design and Recruit share the study setup conversation. You can ask for recruitment work from either page, including screener changes, quota groups, panels, or response limits, then inspect the result in Recruit. Recruitment changes made through the assistant can be saved directly. They do not all appear in the topic guide’s pending-change review, so rejecting a design proposal does not undo recruitment settings the assistant already changed. Review the assistant’s result and the actual panel, quota, or limit configuration before launching. If the assistant reports a limitation or an incomplete action, do not assume the change was made. Use Recruit for the final launch and any cost or payment decisions. Study version history is not a rollback mechanism for recruitment transactions.