Troubleshooting & common questions

The most common cause: the extension has to be opened from the Google Meet tab. Click the Interviewary icon in your Chrome toolbar while the Meet tab is the active tab — that single click both opens the side panel and grants Chrome’s one-time permission to capture that tab’s audio. If you open the panel from any other tab, it has nothing to listen to.

After. Join the Meet call first and make sure audio is actually flowing, then click the Interviewary icon on the Meet tab and press Start interview. There needs to be live call audio on the tab for it to capture — opening it on an empty lobby gives it silence.

It captures the audio coming out of the Meet tab — that’s the candidate (and anyone else on the call). It deliberately does not record your own microphone. You already know what you asked; Interviewary only needs to evaluate the candidate’s answers.

Make sure a https://meet.google.com tab is open, then click the Interviewary icon in the toolbar while that Meet tab is focused and press Start again. The audio-capture permission is granted per-tab at the moment you click the icon — so it has to be clicked on the Meet tab, not from the side panel of a different page.

Keep the Meet tab open for the whole interview. If the tab was closed or reloaded, the capture stream ends — just re-click the Interviewary icon on the Meet tab and press Start again to resume.

Yes. Capturing the tab doesn’t mute it — you and the candidate continue to hear each other exactly as normal.

No — only Google Meet running in a browser tab. The Zoom and Teams desktop apps aren’t browser tabs, so their audio can’t be captured this way.

No. Audio is streamed for live transcription and then discarded — there’s no saved recording, and your transcript, answers, and keys are never sent to us. The only thing kept is the report, and only if you choose to download it. (If you opt in to anonymous usage analytics in Settings, the extension sends only event counts — never transcripts or answers.) See the Privacy Policy and Security page for the full data flow.

Open the side panel, click the gear → Settings. Paste your Deepgram key and your chosen LLM provider’s key, pick a fast model and a report model, click Test keys, then Save. Keys are stored on your own device and you only do this once. The full walkthrough is in the Setup section.

Double-check the key was pasted in full with no stray spaces, that the provider dropdown matches the key you’re using, and that the account has remaining credit and the right permissions. Deepgram keys come from console.deepgram.com.

Pick a fast, lightweight model as your “fast model” — it runs after every answer and must stay under the ~2-second latency budget. Pick a higher-quality model as your “report model”, since it only runs once at the end where quality matters more than speed.

First confirm transcription is working (see Audio & recordingabove) — no transcript means nothing for the AI to evaluate. Then check that your LLM key passed Test keys. Finally, give it a moment: ratings and follow-ups appear a second or two after the candidate finishes speaking.

New content is designed to appear within about two seconds of the candidate finishing an answer. Heavier “report” models are slower by design and only run at the end.

Click End interview. Interviewary generates a summary with every Q&A, its evaluation, strengths, concerns, and a recommendation, which you can download as Markdown or PDF.

Only to the providers you configured, using your own keys — never to us. Read the Privacy Policy and Security page for specifics.