# Usability testing with screen share Canonical page: https://www.questionpunk.com/support/usability-testing Run AI-moderated usability tests where respondents share their screen, complete tasks, and think aloud. Time: 15-30 minutes to set up Level: Intermediate to Advanced Audience: UX researchers, Product teams, Designers ## Steps ### 1. Add a Usability Test question In the survey editor, click **+ Add** and choose **Usability Test** from the AI & Interviews category. This is a dedicated question type for AI-moderated usability sessions with screen sharing and task tracking. ### 2. Enable guided tasks In the question settings, toggle **Guided Tasks** on. This converts the AI audio interview into a usability testing session with screen sharing and task tracking. ### 3. Configure task scenarios Add up to 20 task scenarios. For each task, provide: the **scenario text** (read aloud by the AI), a **target URL** (website, app, or prototype), optional micro-steps (up to 50 per task) to track sub-task completion, a **time budget** (default 5 minutes, max 45 minutes), and a **stuck threshold** (default 2 minutes of no progress before the AI intervenes). Mark warmup tasks with the **warmup** flag so they are excluded from analysis. ### 4. Set task ordering Enable **randomize task order** to counterbalance task sequence across respondents, reducing order effects. You can also randomize step order within tasks. ### 5. Configure think-aloud and intervention Think-aloud reprompting is on by default: if the respondent is silent for 20 seconds (configurable, 5-120s), the AI delivers a neutral "keep talking" nudge. Set the **intervention mode** to **Coach** (AI reads each task, guides step by step, answers questions, and moves on when done) or **Silent observer** (AI reads each task and watches quietly, speaking only to advance or debrief). Configure a **help response mode** (**deflect**, **minimal answer**, or **none**) to control how the AI handles direct help requests. Optionally add a reassurance intro, up to 10 post-task questions per task, and up to 5 wrap-up questions asked after all tasks. ### 6. Enable screen sharing Screen sharing is toggled at the survey level in **Settings** or per section. On desktop, respondents share their screen via the browser's screen capture. On Android, the mobile app uses native screen sharing. On iPhone, the session hands off to the QuestionPunk iOS app for native screen sharing. ### 7. Review results The results show per-task breakdowns: outcome distribution (success/partial/fail/abandoned), average duration, navigation friction (revisits, oscillations, back-navigations), and per-step completion rates with stuck and wrong-turn indicators. AI-synthesized insights highlight pain points with timestamped quotes linked to the session recording. Per-task video clips are cut automatically from the screen recording. ## Details Usability testing is a dedicated question type in QuestionPunk that combines AI-moderated voice interviews with screen recording and guided task tracking. Respondents speak their thoughts while completing tasks, and the AI moderator adapts its questions based on what it observes. Screen sharing works across desktop browsers, Android (native app), and iPhone (hand-off to the iOS app). Each task scenario defines what the respondent should accomplish, on which URL, with configurable time limits and stuck detection. The AI watches for signs of difficulty and can intervene with gentle guidance or stay silent, depending on your intervention mode. **Think-aloud reprompting** keeps respondents narrating their thought process. If they go silent, the AI delivers a neutral prompt after a configurable silence threshold (default 20 seconds). A **Single Ease Question** (SEQ, 1-7 scale) is automatically collected after each task during the debrief, giving you a standardized ease metric alongside qualitative observations. AI analysis synthesizes pain points across respondents: **bottlenecks** (slow or repeated steps), **confusion** (wrong turns, hesitation), **errors** (failed interactions), **workarounds** (non-standard paths), and **abandon causes**. These are clustered into cross-respondent themes to surface the most common usability issues. Navigation analytics track every screen visit, computing revisit counts, oscillation patterns (A→B→A), and back-navigation frequency per task — quantifying navigation friction without manual log review. Per-task video clips are cut automatically from the full screen recording, making it easy to share specific task segments with stakeholders. Counterbalanced task ordering randomizes the sequence across respondents, controlling for fatigue and learning effects in within-subjects designs. **Native app testing:** Task targets can be set to **website**, **app**, or **prototype**. For native app targets, configure an Android package name, iOS URL scheme, and iOS App Store ID to launch the respondent directly into the app being tested. Per-language task text overlays let you translate scenario descriptions, step instructions, and post-task questions for multilingual usability studies. ## Related articles - [AI audio interviews](https://www.questionpunk.com/support/ai-audio-interviews.md) - [How AI interviews work](https://www.questionpunk.com/support/ai-how-it-works.md) - [Synthetic testing with AI personas](https://www.questionpunk.com/support/synthetic-testing.md)