Voice Interview Not Working
Step-by-step troubleshooting guide when voice interviews are not functioning correctly.
Voice interviews require a working microphone, a supported browser, and a secure connection. If your voice interview is not working, this guide walks you through every common cause and its fix.
Quick Checklist
Before diving into detailed troubleshooting, run through this quick checklist:
- Are you using Chrome or Edge? (Recommended browsers for voice)
- Is your page loaded over HTTPS?
- Have you granted microphone permission to the site?
- Is your microphone working in other applications?
- Is your browser up to date?
If you answered "no" to any of these, that is likely your issue. Read the relevant section below for the fix.
Step 1: Check Your Browser
Voice interviews work best in Google Chrome and Microsoft Edge. These browsers have the most reliable support for the audio technologies Koji uses, including WebSocket connections and the Web Audio API.
Firefox supports voice interviews but occasionally has issues with echo cancellation and microphone selection.
Safari has limited support. Voice interviews may work on recent versions of Safari on macOS, but we recommend switching to Chrome or Edge if you encounter problems.
If you are using an unsupported or outdated browser, voice features may fail silently or behave unpredictably. Check Browser Compatibility for the full support matrix.
How to check your browser version
- Chrome: Click the three-dot menu > Help > About Google Chrome
- Edge: Click the three-dot menu > Help and feedback > About Microsoft Edge
- Firefox: Click the hamburger menu > Help > About Firefox
Update your browser if it is more than a few versions behind the latest release.
Step 2: Check HTTPS
Browsers require an HTTPS connection to allow microphone access. This is a security feature built into all modern browsers -- it is not a Koji-specific requirement.
If you are accessing Koji through:
- koji.so -- This is always HTTPS. No action needed.
- An embedded widget -- Make sure the host page uses HTTPS. If the page embedding the Koji widget uses HTTP, the browser blocks microphone access.
- A development server -- Use
localhost(which browsers treat as a secure context) or configure HTTPS on your development server.
Look at the URL bar. If there is a lock icon, you are on HTTPS. If there is a "Not Secure" warning, microphone access is blocked.
Step 3: Grant Microphone Permission
The most common cause of voice interview issues is a missing or denied microphone permission.
Chrome / Edge
- Look for a camera/microphone icon in the address bar (right side).
- Click it to see the current permission state.
- If it says "Blocked", click the dropdown and select Allow.
- Reload the page.
Alternatively:
- Click the lock icon in the address bar.
- Find Microphone in the permissions list.
- Set it to Allow.
- Reload the page.
Firefox
- Click the lock icon in the address bar.
- Click Connection secure > More Information.
- Go to the Permissions tab.
- Find Use the Microphone and set it to Allow.
- Reload the page.
Safari
- Go to Safari > Settings > Websites.
- Find Microphone in the sidebar.
- Find the Koji domain (koji.so) and set it to Allow.
- Reload the page.
System-Level Permissions (macOS)
On macOS, your browser also needs system-level microphone permission:
- Open System Settings > Privacy & Security > Microphone.
- Make sure your browser (Chrome, Edge, etc.) is listed and enabled.
- If it is not listed, you may need to restart the browser and try accessing the microphone again to trigger the system prompt.
System-Level Permissions (Windows)
- Open Settings > Privacy > Microphone.
- Make sure "Allow apps to access your microphone" is turned on.
- Scroll down and confirm your browser is in the allowed list.
Step 4: Test Your Microphone
Before blaming the browser or Koji, confirm your microphone is working:
- Open your operating system's sound settings.
- Speak into your microphone and check if the input level meter moves.
- Try recording a short clip in a voice memo app or an online microphone tester.
If the microphone does not work in other applications, the problem is at the hardware or OS level, not with Koji.
Common Microphone Hardware Issues
- External microphone not selected. Your system may default to a built-in microphone. Go to sound settings and select the correct input device.
- Mute switch on. Some headsets and microphones have physical mute switches. Check for a switch or button on the device.
- USB connection issues. If using a USB microphone, try a different USB port. Hubs can sometimes cause connection problems.
- Bluetooth connection dropped. Bluetooth headsets can disconnect silently. Check your Bluetooth settings to confirm the headset is still connected.
Step 5: Check for Conflicting Applications
Some applications lock the microphone and prevent the browser from accessing it:
- Video conferencing apps (Zoom, Teams, Meet) may hold an exclusive lock on the microphone, even when minimized.
- Voice assistants (Siri, Cortana) may be listening in the background.
- Recording software (OBS, Audacity) may be capturing the input.
Close any application that might be using the microphone, then reload the Koji interview page.
Step 6: Check Network and Firewall
Koji's voice interviews use ElevenLabs Conversational AI for real-time voice processing. This requires a stable WebSocket connection.
If voice does not connect or cuts out frequently:
- Firewall or proxy: Corporate networks may block WebSocket connections. If you are on a company network, ask IT to allow WebSocket traffic to the ElevenLabs API endpoints.
- VPN: Some VPNs add latency or block WebSocket connections. Try disconnecting your VPN temporarily.
- Slow connection: Voice interviews need a stable internet connection with reasonable latency. If your connection is slow or unstable, voice quality may suffer.
How Voice Interviews Work
Understanding the architecture can help with troubleshooting:
- When you start a voice interview, Koji establishes a WebSocket connection to the voice AI service (powered by ElevenLabs Conversational AI).
- Your microphone audio is captured via the Web Audio API in your browser and streamed over the WebSocket connection.
- The AI processes your speech, generates a response, and streams the audio back to your browser.
- The response audio plays through your speakers or headphones.
This means your browser needs:
- WebSocket support for the real-time connection
- Web Audio API for capturing and playing audio
- Microphone access for recording your voice
What Happens When Voice Can't Start
When a voice interview can't start, Koji tells the participant why instead of leaving them on a loading screen. The message names the cause it found: microphone access was denied, no microphone was found, the microphone is busy in another app, no suitable microphone was found, or access is blocked because the page isn't on a secure (HTTPS) connection. If the cause is something else, such as a dropped connection, the message says voice mode could not be started.
Next to the message is a Try again button. It asks for the microphone again, so it covers the common fixes: you dismissed the browser prompt by accident, you plugged in a headset, or you changed the microphone setting to Allow in your browser after the message appeared. Fix the cause using the steps above, then press Try again. You don't need to reload the page.
What happens next depends on how the researcher set up the study.
Studies that accept typed answers
If the study offers text as well as voice, a failed start moves you straight into the text interview, so you can keep going without a microphone. You'll see the first question, the reason voice couldn't start, and a Try again button above the reply box. Press it once you've fixed the microphone to go back to voice. While a voice interview is running in one of these studies, a Text button next to the microphone also lets you switch to typing at any point.
On the interview landing page of these studies you can also pick Text before you begin. Text interviews don't need a microphone and work in all modern browsers. See Text Interview Experience for how they work.
Voice-only studies
Some studies are set up for voice answers only. In those there is no text option: no Text choice on the landing page, no Text button during the call and no automatic switch to typing. If voice can't start, you'll see the reason and Try again on the voice screen. Work through the checklist and steps above, then press Try again. If it still won't connect, try Chrome on a different device or network, or contact the researcher who sent you the link.
For researchers
If many of your participants are likely to be on work laptops, shared devices or locked-down networks, offer text alongside voice. Participants whose microphone fails can then finish the interview in text instead of dropping out. Voice-only suits studies where hearing people speak matters more than reaching everyone.
Still Not Working?
If you have tried all of the above and voice is still not working:
- Try a different browser (Chrome is most reliable).
- Try a different device to rule out device-specific issues.
- Check the Voice Interview Experience guide for additional context on how voice mode works.
- Contact Koji support with:
- Your browser name and version
- Your operating system
- Whether you see any error messages
- Whether the study let you continue in text
Further reading on the blog
- AI-Moderated Interview Platforms Compared: Which One Actually Works? (2026) — Not all AI interview platforms deliver real qualitative depth. This guide compares the top AI-moderated interview platforms in 2026 — Koji,
- Best AI Customer Interview Tools in 2026: The Complete Buyer's Guide — AI has fundamentally changed how product teams conduct customer research. Here are the best AI customer interview tools in 2026 — ranked by
- Best Customer Churn Interview Tools (2026): The Top 8 Compared — A 2024 study found exit surveys match the real churn driver in only 31% of cases. The right interview tool fixes that. Here are the 8 best c