How Interview Signal works, from job description to client submittal

Interview Signal is a desktop app for Windows and macOS that runs beside Zoom, Microsoft Teams, Google Meet or a phone screen on the interviewer's own computer. It captures your microphone as "You" and your computer's sound as "Candidate", so no bot joins the call. Each role gets a question kit and scorecard built from the job description; during the call the guide ticks itself off, follow-ups are suggested and the scorecard fills with the candidate's own words; when you end the interview it writes the client submittal in your template. This page describes version 0.9.10, the current release as of October 2026.

What does an interview look like in Interview Signal, step by step?

1. Sign in with your work email. The app emails a link and a six-digit code; either works. An address with no account yet can choose "Start free — 20 minutes, no card".
2. Make a role. Press "New role", paste the job description, and generate the kit: questions, what a strong answer contains, and a weighted scorecard.
3. Start. On the home screen, the "Start an interview" bar has the role you used last already picked, an optional candidate name, and a "Start" button (or press S).
4. Check your audio on the pre-interview screen, add the candidate's details, and pick a consent option if you need one.
5. Run the call with the guide, transcript, follow-ups and scorecard on one screen.
6. Press "End interview", review the scorecard and submittal, and export it.

How do I build a role from a job description?

"New role" opens the kit screen. Give the role a title, then either paste the link into "Paste the posting link" and press "Fetch", or paste the text into "Job description". "What the client actually wants" is optional and outranks the job description wherever the two disagree. Choose the interview language, the submittal template, the interview type and the difficulty, then press "Generate the kit".

Every question is editable. Tick "Must ask" on the questions the client will ask you about, and "Keep" on the ones that should survive "Regenerate questions". The scorecard areas each carry a weight from 1 to 3 and a rubric for a 1–3, 5–6 and 9–10 answer, and the kit ends with a "Consent notice" you can edit to sound like you. "Save role" stores it on your account. The full walkthrough is in how the interview kit is built from a job description.

What happens before the call starts?

The pre-interview screen has four panels, and only the first is required:

• 1. Check your audio. "Test audio" shows live meters for You and Candidate; "Record 5 seconds" records both sides and plays your clip back. See how the audio check works.
• 2. The candidate. Name, "Company or client (optional)", their CV (paste it or attach a PDF or text file), and "How are you speaking?": Video call, Phone or In person.
• 3. Interviewing as a panel (optional). Open a panel or join one with a six-character code, then split the questions. Panels are part of the Team plan.
• 4. Consent (optional). The read-aloud line, where you are interviewing from, and a record of what you chose. See how the consent features work.

Press "Start interview" and capture begins.

What do I see during the interview?

The top bar shows the elapsed time and the time now, "Mic & language", "AI", how many questions are covered, a "?" for help, "Mute me", "Pause capture" and "End interview". Below it are three panes you can resize:

• Guide. One line per question, grouped by category, ticking off as you ask them. Rest on a line to read the whole question, click to open it, and "Pin" it to focus the follow-ups and scoring on it with a clock against its planned time. "+ Add" puts in your own question, or three the model suggests in a category you pick.
• Transcript. Live captions for You and Candidate, the committed transcript above them, "Score as we go" and "Evaluate", and the follow-ups the last answer earned, each with its reason. A box underneath answers your own questions to the model, such as what the right answer is.
• Scorecard. A fit score out of 10, each area as it is scored, a "+" for your own score or note, and "Your notes", which go into the submittal as your observations, never as the candidate's words.

Three optional readings sit on top: an answer check that flags answers that read as AI-written or read out (it never changes a score), a technical accuracy check that marks claims green or red in the transcript, and Dual AI, which runs a second model alongside. All three are switched on or off under "AI".

What happens when I press End interview?

The review screen opens with the submittal written in the role's template. Edit it in place, press "Preview" to see it formatted, or "Rewrite" after changing the candidate name, job title or company at the top. Each score on the scorecard carries the candidate's own quotes, and every quote is checked against the transcript: one that is not there word for word is removed, and an area left with no quote is held at 5 and marked to probe. See the scorecard with verified quotes.

The export row has "Copy", "Save .md", "Export .docx", "Add CSV row", "Email to client", "Copy for Slack", "Copy next round" and "Copy ATS note…", which offers a Bullhorn note, a Greenhouse scorecard note, a Lever note and a JobAdder note to paste by hand. The side panel lists the length, fit, questions and must-asks covered, the consent record, the scorecard, your notes and the transcript, with "Save transcript as .md".

What leaves my computer?

Transcripts, quotes, candidate names, notes and submittals are saved in a folder on your computer and are never uploaded. During the call, audio goes in short chunks to a speech-to-text provider and is not retained, and text is sent to a model provider to produce the kit, follow-ups, scores and submittal. Interview Signal's server keeps counts and durations for each interview, plus your account, roles and templates. The detail is in where your data is stored and on the security page.

Does Interview Signal join the call?

No. There is no bot, no extra participant and no meeting integration to install, so the candidate sees nothing from Interview Signal. That also means the notice is yours to give. How the two audio lanes work is explained in how Interview Signal records without a bot, and how bot-free capture compares with meeting bots in general is in AI note-takers without a bot.

What does Interview Signal not do?

• No ATS integration. Scorecards and submittals are copied or pasted into your ATS.
• Desktop only. Windows 10 or 11 and macOS 13 or newer. There is no web app, no mobile app and no Linux version.
• No video. It does not record video, and it saves no audio file either.
• Phone screens run through the computer. A call on a separate handset is not heard.
• In-person speaker labels are unreliable, because both voices reach the same microphone. The transcript and scorecard still work.
• It needs a connection. Saved roles open offline, but transcription, new kits and scoring do not.

How can I try it?

Every account starts with 20 free minutes of live transcription and no card; hour packs and plans are on the pricing page. On the home screen, "Add practice roles" adds two sample roles whose candidate answers you play from the Try it page, which also has a full walkthrough video. The installers are on the download page.