How to use Interview Sidekick
Five features, each walked through step by step. Click any step to see the UI. Click More details, Examples, or Troubleshooting for the depth.
Profile setup
A one-time setup that powers every other feature. The AI uses your resume + the job description + your custom rules to produce answers that quote your real experience — not generic STAR-method filler.
Head to interviewsidekick.com and click Get Started.
What we store: your email, name, and profile photo from Google. Nothing else. We don't access Gmail, Calendar, Drive, or any other Google data.
Single account everywhere: the same Google account signs you into the web app, the Mac desktop app, the Windows desktop app, and your phone. Your resume, JDs, sessions, and past scores sync across all of them.
“I don't use Google.” Email ben@interviewsidekick.com with the email you'd like to use. We'll provision it as email-magic-link auth manually.
“The Google popup is blocked.” Browsers sometimes block the OAuth popup. Allow popups for interviewsidekick.com and click Continue with Google again.
“I signed in but the desktop app doesn't see me.” The desktop app uses the same login but is a separate auth flow. Click the “Sign in” button inside the app, not the web one.
Fill in the role you're interviewing for. The JD is the single biggest input the AI uses to decide what's relevant — copy/paste straight from the company's posting.
JD length: up to 4,000 characters (roughly a full posting). Paste the whole thing — title, responsibilities, requirements, and “nice to haves.” The AI uses all of it.
Multiple roles? Each session has its own profile. When you start a new live session or mock interview, you can swap the role without disrupting your other ones.
Where to edit later: Settings → Profile, or click the role name in the top of any session. Changes apply to new AI answers, not previous ones.
Good JD paste (gets a great result):
Bad JD (too vague):
The first gives the AI vocabulary to mirror back (“card-network integrations,” “platform primitives”). The second gives it nothing to work with.
Upload a PDF or paste the text — either works. The AI references your real bullet points when answering instead of making up generic ones.
Accepted formats: PDF (.pdf) up to 5MB, or plain text up to 4,000 characters. We parse text out of the PDF — no images or charts get analyzed.
What gets stored: the parsed text, on our servers, encrypted at rest. Used only to construct AI prompts for your sessions. Not shared with employers, recruiters, or third parties.
Privacy mode: if your resume has anything you don't want sent to an LLM (e.g. a confidential project name), redact it before upload — paste the text and edit it in the textarea.
Editing later: Settings → Profile → Resume. You can replace, edit, or delete it anytime. Deleting it removes it from our database within 24 hours.
“My PDF won't upload.” Scanned PDFs (image-only) don't parse. Re-save as a text PDF, or paste the content directly.
“The AI's answers don't reference my projects.” The most likely cause is that the JD is generic. If the JD doesn't mention “Python” or “ML,” the AI doesn't know to surface your Python/ML projects. Use Step 4 (Extra Context) to force it.
This is the one most users skip — and the one that separates “generic AI answer” from “sounds like me.” Tell the AI how you want it to answer: structure, length, focus areas, things to avoid.
Format / structure rules:
- “Use the STAR method for every behavioral answer.”
- “Keep every answer under 90 seconds spoken — roughly 200 words.”
- “Always end behavioral answers with what you learned.”
Focus rules:
- “Lean heavily on my Python and ML experience at Reddit.”
- “When asked about leadership, reference my time managing the 6-person platform team.”
- “If asked about gaps in my resume, mention the parental leave directly and pivot.”
Voice / tone rules:
- “Don't use corporate buzzwords like ‘synergy’ or ‘move the needle.’”
- “Be specific and quantified. Always include a metric.”
- “Mirror the interviewer's vocabulary — if they say ‘customer,’ I say ‘customer,’ not ‘user.’”
Deflection rules:
- “If asked about salary, redirect: ‘Happy to discuss once we align on the role.’”
- “Don't answer anything about my current employer in detail — be vague.”
How they apply: these rules are injected at the top of every AI answer prompt. They take priority over the AI's default behavior — so “keep under 90 seconds” will actually keep it under 90 seconds.
Stacking: stack as many as you want, but each one dilutes the others. 3-5 high-signal rules outperform 15 vague ones.
Save and reuse: rules are saved on your profile and reused across all sessions until you change them. Good to set once and forget.
Live Interview Assistant
Real-time AI help during your actual interview. Pick one of three setups depending on whether you can install software, want Stealth Mode, or prefer using your phone as a second screen.
| Feature | DesktopBEST | Phone | Chrome |
|---|---|---|---|
| Stealth Mode | |||
| System Audio Capture | |||
| Compact Floating Window | |||
| Code Capture (screenshots) | |||
| No Download Required | |||
| Works with Zoom Desktop App |
Setup A — Desktop App
RecommendedGo to /download and grab the build for your OS. The page auto-detects whether you're on Mac (Apple Silicon vs Intel) or Windows.
Which Mac build do I need? If you have M1, M2, or M3 → Apple Silicon. If you have an older Intel Mac → Intel. Not sure? Click the Apple logo (top left) → About This Mac → check “Chip”.
File sizes: ~87 MB (Apple Silicon), ~91 MB (Intel Mac), ~124 MB (Windows).
Signed & notarized: the Mac build is signed by “Cover Letter Copilot, LLC” and notarized by Apple. The Windows build is currently unsigned — Windows SmartScreen may warn the first time. Click More info → Run anyway.
Auto-updates: the app checks for updates ~10 seconds after launch. Updates apply on next restart. No manual upgrade needed.
“Download blocked.” Chrome / Edge sometimes flag electron apps. Click “Keep” in the download bar. The file is signed and safe.
“Wrong build downloaded.” Use the explicit buttons below the auto-detect — pick Mac Intel, Mac Apple Silicon, or Windows manually.
Mac: double-click the .dmg, drag the Interview Sidekick icon into Applications, then launch from Spotlight or Launchpad.
Windows: double-click the .exe and click through the installer. The app launches automatically when done.
Either way, click Sign in with Google in the app — same Google account as the web app.
“Mac says ‘can't be opened, unidentified developer.’” You downloaded an older or modified copy. Re-download fresh from /download. The notarized version opens cleanly.
“Windows SmartScreen says ‘Windows protected your PC.’” Expected — our Windows build isn't code-signed yet. Click More info → Run anyway. Safe to install.
“The app opened but won't sign me in.” The OAuth flow opens your default browser. Make sure you complete the sign-in there, then return to the app — it polls every few seconds and should pick up the session.
The first time you click Record, your OS will ask for microphone access. On Mac you'll also get a separate prompt for System Audio Recording — that's the one that lets ISK hear the interviewer through Zoom/Meet/Teams.
The Mac confusion: macOS splits audio permissions into two sections inside the same Settings pane (Privacy & Security → Screen & System Audio Recording):
- Screen Recording — required so ISK can capture the interviewer's audio coming out of your speakers.
- System Audio Recording Only — a separate section below the Screen Recording list. Easy to miss. Required for the system audio capture to actually work.
If you only enable Screen Recording, the app launches but the AI hears nothing from the interviewer. Both must be on.
Two separate sections. People miss the second one — scroll past the Screen Recording list.
“I clicked Allow but it still says permission needed.” On Mac, sometimes you need to quit and reopen Interview Sidekick after granting. Cmd+Q, then relaunch.
“The interviewer's voice isn't showing in the transcript.” 90% of the time this is the missing “System Audio Recording Only” permission. Go to System Settings → Privacy & Security → Screen & System Audio Recording, scroll past the Screen Recording list, find the second section, enable Interview Sidekick there.
“Windows isn't asking for permission.” Windows handles mic permission silently the first time. If audio still doesn't work, go to Settings → Privacy & Security → Microphone and confirm desktop apps can access it.
ISK supports 31 languages for both live transcription and AI answers — they switch together. English (US) is the default; change it before your interview if you need anything else.
Supported: English (US & UK), Spanish, French, German, Portuguese (BR), Hindi, Japanese, Korean, Mandarin, Italian, Dutch, Russian, Arabic, Bulgarian, Croatian, Czech, Danish, Finnish, Greek, Hungarian, Indonesian, Norwegian, Polish, Romanian, Slovak, Swedish, Tamil, Turkish, Ukrainian, Vietnamese.
How it's set: the language picker lives in your session config — selected when you start a new live session, not as a global default. So you can have one session in English and another in Spanish without resetting anything.
Open ISK and click Record. Then join your interview through Zoom / Meet / Teams like you normally would — ISK doesn't replace your meeting tool, it runs alongside it. Transcript on the left, AI answers stream on the right.
The two panels:
- Transcript — live conversation, both sides, color-coded by speaker. Interviewer questions in blue, your responses in gray.
- AI Answer — auto-generated suggestion for whatever the interviewer last asked. Streams in as the model writes it.
Actions you can take on an AI answer:
- 🔄 Regenerate — get a different angle on the same question if the first answer doesn't fit. Useful when the AI misread the question.
- 📨 Send Direct Question — type your own question to the AI (e.g. “Rephrase that more casually”).
- 📸 Analyze Screenshot — for coding rounds, sends a screenshot of your screen to the AI for explanation and suggested approach (covered in the next sub-section).
For coding interviews on CoderPad, HackerRank, LeetCode, or any code editor, click Screenshot code in the Code tab. ISK captures whatever's on your screen, sends it to the AI, and you get an explanation of the problem and a suggested approach in seconds.
The screenshot stays local + ephemeral — sent to the AI for that one answer and not stored.
“Transcript is showing my voice but not the interviewer's.” System Audio Recording permission missing (see Step 3). Or you're using the Phone setup but speakers are muted — turn them up.
“AI is stuck on ‘waiting for question.’” Auto-detect didn't catch it as a question (often happens with statements phrased as questions). Click Generate answer manually, or rephrase the auto-detect threshold in Settings.
“Answers are generic.” Profile setup likely incomplete. Check Settings → Profile and make sure resume + JD + extra context are all filled in.
Stealth Mode uses OS-level content protection to hide ISK from screen share — even if you share your full screen, your interviewer sees nothing of ISK.
| Action | Mac | Windows |
|---|---|---|
| Generate AI Answer | ⌘ + ⏎ | Ctrl + Enter |
| Toggle Recording | ⌘ + ⇧ + R | Ctrl + Shift + R |
| Toggle Stealth Mode | ⌘ + ⇧ + S | Ctrl + Shift + S |
| Focus Input | ⌘ + ⇧ + I | Ctrl + Shift + I |
| Capture Screenshot | ⌘ + ⇧ + A | Ctrl + Shift + A |
| Hide / Show Window | ⌘ + ⇧ + H | Ctrl + Shift + H |
| Expand / Collapse | ⌘ + ⇧ + M | Ctrl + Shift + M |
| Toggle Code Mode | ⌘ + ⇧ + C | Ctrl + Shift + C |
These are global shortcuts — they work even when ISK isn't focused, so you can trigger Generate Answer mid-interview without alt-tabbing.
Compact View shrinks ISK into a tiny floating window (~380px wide) that stays on top of everything. Drag it anywhere on screen. Click the same hotkey to toggle back to full view.
Compact view works with Stealth on — your interviewer won't see this floating window during screen share.
Zoom's default screen-sharing method ignores macOS content protection, which means ISK can still leak into shared screens with Stealth on. One-time fix:
- Open the Zoom desktop app → Settings (gear icon).
- Click Share screen in the left sidebar → Advanced.
- Under Screen capture mode, pick “Advanced capture with window filtering”.
After that, Zoom respects Stealth. Google Meet, Teams, and Webex work out of the box.
“Stealth is on but the interviewer can still see ISK.” 95% of the time this is Zoom-on-Mac with the default share mode — see the Zoom-specific section above. Switch to “Advanced capture with window filtering” and reshare.
“Compact view is too small.” Drag the bottom-right corner to resize. Minimum size is 350×56 (just the title bar).
“Hotkeys don't work.” Another app may have claimed them. On Mac, check System Settings → Keyboard → Keyboard Shortcuts. On Windows, check installed apps that register globals (e.g. Raycast, Alfred, Logi Options+).
Setup B — Phone as second screen
No InstallOpen interviewsidekick.com on your phone and sign in with the same Google account you use elsewhere. Add the page to your home screen if you'll use this often.
Browser recommendation: Chrome on Android, Safari on iPhone. Both work well; Firefox occasionally has audio capture quirks on mobile.
Add to home screen: in Safari, tap Share → Add to Home Screen. In Chrome, tap menu → Add to Home screen. Gives you an app-like icon without an install.
Put your phone on the desk next to your laptop — close enough to clearly pick up the interviewer's voice, but not in front of your camera. The phone's mic acts as the audio source.
Volume: turn laptop volume up enough that the phone clearly hears the interviewer. Phone screen should be at full brightness and screen-sleep disabled (Settings → Display) so it doesn't go dark mid-interview.
Phone position: screen tilted up at ~30°, about 8-12 inches from the laptop speaker, slightly off to your dominant-hand side so you can glance without turning your head from camera.
Phone audio source: the phone listens through its built-in mic. Earbuds/headphones aren't needed (and actually hurt — they isolate your ears from the interviewer's voice that ISK needs to hear).
“Phone is picking up my voice instead of the interviewer's.” Mute the ISK mic on the phone while you speak (button next to the timer). Unmute when the interviewer starts talking again.
“Audio is too quiet.” Turn up laptop speaker volume. If you're wearing headphones for the call, your laptop speakers might be auto-muted — switch audio output to speakers in your meeting tool.
Hit Record on your phone. The discipline: tap the mute button when you start talking, unmute when the interviewer asks the next question. Keeps the AI focused on their questions only.
Setup C — Chrome dual tab
Browser OnlyThis setup only works if your interview runs inside Chrome — Google Meet works great, Zoom-in-browser works, but the Zoom desktop app won't. For the desktop Zoom app, use Setup A or B.
What works in Chrome dual tab: Google Meet, Microsoft Teams web, Zoom web, Webex web, Slack Huddles, Discord stage. Any browser-native call.
What doesn't: Zoom desktop app, Teams desktop app, FaceTime, Phone-only calls, anything where the audio doesn't live in a Chrome tab.
In ISK, click Connect Audio. Chrome shows a tab picker — select your meeting tab and check the “Also share tab audio” box. That box is critical; without it, ISK gets a video feed but no audio.
“I don't see the ‘Also share tab audio’ checkbox.” The checkbox only appears when you select a tab (not a window or entire screen). Click the “Chrome Tab” tab in the picker, then select your meeting tab.
“Tab picker is empty.” Chrome blocked screen sharing. Check the address bar for a permission icon, click it, and allow.
“Audio works but tab freezes.” Sharing a tab uses memory. Close other heavy tabs (Figma, Notion, YouTube) for a smoother session.
ISK pulls audio from the meeting tab. Note: no Stealth Mode in browser, so don't share your full screen unless you're prepared to hide the ISK tab manually.
Mock Interview Practice
Practice with an AI interviewer before the real one. Two modes: Lite (one question at a time, record-and-review) and Advanced (live voice conversation with follow-ups). Both use voice — no typing.
Lite — Record and review
QuickHead to /mock-interview. Pick a category (Behavioral, Product Sense, System Design, Coding, Case…) or a company, and click any question to start.
Custom sessions: if you want a multi-question set instead of one-off, click Start Custom Session → pick a company + role + length, and the AI builds you a 3-, 5-, or 8-question pack from the JD.
Plan gating: mock interview practice requires a paid subscription. Free users can browse the question bank but answering/scoring is paid-only. Paid users get up to 10 mock questions per day.
One question, one answer. The AI transcribes you in real time. When you're done, tap the mic again to stop.
Pause mid-answer: pause/resume button next to the mic. Useful if you trail off and want to gather thoughts.
Restart: not happy with your answer? Tap Re-record — overwrites the previous take, no penalty.
Time limit: 3 minutes hard cap. Forces you to be concise (which is how real answers should be).
The AI grades you on four dimensions and returns specific keep/improve notes. Scores are saved to your profile so you can watch your trend over time.
Four sub-scores (each out of 100) plus an overall score:
- Structure — did you have a clear opener, middle, and close? STAR-style framing for behavioral answers.
- Clarity — did you avoid jargon and stay on topic? Were pronouns and references clear?
- Specificity — did you use concrete examples, names, projects, dates? Or did you stay abstract?
- Impact — did you quantify outcomes (numbers, %), make it clear what you personally did (vs. “we”)?
Each score is saved per question so you can see your trend over time.
Every mock session is saved. Replay any past answer, see how scores trend, and identify which categories you score lowest on — then practice those more.
Advanced — Live voice conversation
RealisticOpen any question in the Mock Interview library and choose Live voice mode (paid plans only — Lite plans get the record-and-review flow instead). Pick a voice style, then start. The AI speaks the question out loud and listens for your answer.
Voice options: two styles — feminine and masculine. Both use ElevenLabs voices.
Single-question vs custom multi-question: opening one question = a single voice-to-voice exchange with feedback at the end. For a full multi-question simulation, use the Custom Session flow (Quick / Standard / Comprehensive — 3 / 5 / 8 questions).
Plan gating: live voice mode requires a paid plan. Lite plan users see the record-and-review mode instead.
Unlike Lite, this is back-and-forth. If your answer is vague, they push: “Can you give me a specific example?” If you mention a metric, they ask: “How did you measure that?”
The AI listens between turns: when you stop talking, the AI processes what you said and either asks a follow-up or moves on. Silence triggers a follow-up — don't panic if there's a pause.
End early: close the session anytime. You still get the feedback for whatever you covered.
Saved as “Live voice”: sessions are tagged in your past-sessions list so you can tell them apart from record-and-review.
When the session ends, you get scored feedback for every question — with the AI flagging where your answers got stronger or weaker as the conversation progressed.
Interview Prep — JD-generated questions
Paste a JD, get a tailored set of questions you're likely to be asked. Pick how long a session you want — the AI generates 3, 5, or 8 questions specific to that role and company.
Head to /mock-interview and click Start Custom Session. This is the same flow as Mock Interview — but instead of picking a pre-written question, the AI builds a question pack from your job description.
Walk through the picker: company, job title (or job family), then session length. The three preset lengths:
- Quick Practice — 3 questions (~10-15 min). Focused warm-up.
- Standard Session — 5 questions (~20-30 min). Most common pick.
- Comprehensive Prep — 8 questions (~35-50 min). Full mock feel.
If you completed Profile Setup, your JD is already saved and pre-fills the form. Otherwise paste it here.
Claude reads your JD and writes questions specific to the role — behavioral, technical, product, leadership — categorized so you know what kind of round each maps to.
Inputs: the job description (or job title fallback), company name, and any context documents you select (resume, prior projects, extra-context notes — up to 5).
Model: Claude Sonnet 4.6 via OpenRouter, with a 60-second timeout for the generation call.
Reroll: if a question feels off, you can regenerate the whole set — same JD, different angles. Cheap.
Click the play button on any question to start recording. Same scoring as Mock Interview Lite — STAR structure, clarity, relevance, impact — and your scores are saved to your profile so you can see which question types still trip you up.
Question Bank
Browse 10,000+ real questions asked at Google, Meta, Amazon, Apple, and 150+ other companies. Search and filter by company, role, category, or seniority.
Open /question-bank. Browse by company logo, search any keyword, or filter by category/seniority.
Sidebar filters: company (Google, Meta, Amazon, etc.), category (Behavioral, System Design, Product, Coding…), and job type (PM, SWE, Data, etc.). Stack them.
“I'm a senior PM interviewing at Meta.” → Filter: Meta + Product Sense + Senior. ~80 results.
“I have a SWE coding screen at Google tomorrow.” → Filter: Google + Coding. ~250 results, sorted by frequency.
“I just want behavioral questions for FAANG.” → Filter: Google OR Meta OR Amazon OR Apple OR Netflix + Behavioral. ~600 results.
Click a question to record an answer. The AI grades you the same way as a regular mock interview, and the score is saved to your profile.
Free vs paid: the question bank itself (browsing, searching, filtering) is free. Recording and scoring an answer requires a paid plan — paid users get up to 10 mock questions per day.
Meeting platform specifics
These steps are for the Chrome dual tab setup. With the desktop app, system audio is captured automatically — just join your meeting normally.
Setting up with Zoom
- Click on your Zoom meeting link
- Don't open the Zoom app when asked
- Choose "Join from Your Browser"
- If you can't see the "Join from Your Browser" option:
- Click "Launch Meeting"
- Say no to opening the Zoom app
- Keep doing this until you see "Join from Your Browser"
- Come back to this page and click "Connect"
- Select the Tab where the interview is taking place
- Click "Share"
Setting up with Microsoft Teams
- Click on your Microsoft Teams meeting link
- Select 'Continue on this browser'
- Come back to this page and click "Connect"
- Select the Tab where the interview is taking place and select the "Also share tab audio" option
- Click "Share"
Setting up with Google Meet
- Open your Google Meet in your web browser as usual
- Come back to this page and click "Connect"
- Select the Tab where the interview is taking place and select the "Also share tab audio" option
- Click "Share"
Setting up with Webex
- Click on your Webex meeting link
- When prompted to open the Webex app, click "Cancel" on the dialog
- Look for "Join from your browser" near the bottom of the page and click it
- Allow microphone and camera permissions, then click "Join meeting"
- Come back to this page and click "Connect"
- Select the Tab where the interview is taking place and select the "Also share tab audio" option
- Click "Share"
Setting up with Skype
- If joining through a meeting link, click cancel when prompted to open the Skype app
- Proceed to join the meeting in the browser
- If logging into Skype, go to Skype.com and click "Use Skype online"
- Come back to this page and click "Connect"
- Select the Tab where the interview is taking place
- Click "Share"
Tips for every setup
Use headphones
Prevents echo and helps the AI distinguish the interviewer's voice from yours — the single biggest accuracy improvement.
Run a test first
Do a 2-minute test session with a YouTube mock interview. Confirms your audio + ISK both work before it matters.
Resume + JD are mandatory
Without them, answers are generic. With them, they quote your real experience back at you.
Add extra instructions
"Use STAR method", "Highlight my Python work", "Keep answers under 2 min" — the AI follows these as hard rules.
Open ISK before joining
Enable Stealth Mode + Record before joining the call, not while the interviewer is waiting on you.
Mute when speaking (phone)
On phone, mute ISK's mic while you talk. Keeps the AI focused on the interviewer's questions.
Ready to try it?
Free 2-minute test session — no credit card. Get the desktop app for full features.