What Interviewers Can (and Cannot) See on a Screen Share in 2026

The question every candidate wants a straight answer to: what can the interviewer actually see when I share my screen? There is a lot of folklore about this. Here is the concrete answer for Zoom, Google Meet, and Microsoft Teams in 2026, plus the specific behaviors that give people away.

What's captured

Screen sharing captures the pixels of a region you selected. That is it. There is no telemetry, no keystroke logging, no clipboard hook. What determines what the interviewer sees is which pixels are in the selected region.

Three regions are possible:

• Entire monitor. Every visible pixel on the display, top to bottom.
• Application window. Only the pixels of the specific window you selected — nothing overlapping in front, nothing behind.
• Browser tab. Only the tab's rendered contents. Nothing from other tabs, nothing from the browser chrome.

What always shows (entire-screen share)

• Every open window on that monitor (unless flagged as capture-excluded — see below)
• Taskbar and system tray icons
• Notification banners as they appear
• Mouse cursor position and every click
• Focus changes (title bar dimming, window flashing)
• Any second monitor if you selected "all displays"

What does NOT show

• Windows on other virtual desktops (Windows "Task View" workspaces / macOS Spaces)
• Windows that set the OS WDA_EXCLUDEFROMCAPTURE flag on Windows 10 2004+ or the equivalent macOS NSWindowSharingNone flag
• Anything on a second monitor you didn't share (if the interviewer picked "primary display" only)
• The contents of your clipboard until you paste
• Browser history, tab titles other than the shared one, or bookmarks
• System audio (unless the interviewer explicitly enabled "share sound")
• Your webcam if it's a separate window

The capture-exclusion flag

Since Windows 10 build 2004, an app can callSetWindowDisplayAffinity(hwnd, WDA_EXCLUDEFROMCAPTURE) to tell the OS: never include this window's pixels in any capture pipeline. Apps using it are invisible to:

• Zoom, Google Meet, Microsoft Teams (entire-screen and application-window modes)
• Windows Snipping Tool and Snip & Sketch
• Print Screen and Win+Shift+S
• OBS, XSplit, and other screen-capture tools using DirectX / DWM

Apps that use it include some legitimate ones (DRM overlays for streaming services, password managers with reveal-on-hover) and some purpose-built stealth apps like Interview Helpers. See the full stealth breakdown for the exact call site.

What the interviewer notices (that isn't your app)

Even a perfectly hidden overlay can be inferred if your behavior gives it away. The tells:

• Eyes drifting to a specific dead spot on the screen. If your overlay is in the top-right and you keep glancing there while there's "nothing" in the shared view, an experienced interviewer notices.
• Answers arriving before you've thought. Real thinking has friction. Perfect stream-of-consciousness is a tell.
• Perfect complexity analysis stated instantly. Normal engineers pause and count. If you say "O(N log N) time, O(1) space" the second you see the problem, it's suspicious.
• Sudden focus changes. Your IDE title bar dims and re-brightens even though nothing visible was clicked — the interviewer sees the flicker.
• Notifications from apps you shouldn't have open. Turn on Focus Assist / Do Not Disturb.

Behaviors that keep you safe

• Use hotkeys. Global hotkeys don't change window focus, so nothing flickers.
• Overlay in a natural gaze location. Top-right, adjacent to where the IDE puts its file tree, is where your eyes go anyway to read code.
• State the pattern first, verify second. "This looks like sliding window — let me check my complexity" is human. "It's O(N) time, O(K) space" without a pause isn't.
• Type the code yourself. Rhythm gives away transcription faster than any word choice.
• Turn off Windows notifications. Focus Assist → Priority Only or Alarms Only.

Interviewer perspective: what they actually look for

Interviewers know AI tools exist. Most don't care as long as you demonstrate real thinking. What they're grading:

• Can you pattern-match a problem to a category?
• Can you propose an approach and defend it?
• Can you write correct code in a language of your choice?
• Can you enumerate edge cases and state complexity?
• Can you communicate under time pressure?

A stealth AI helper gets you a boost on the first three. Nothing lets you skip 4 and 5. Which is fine — those are the parts you should actually own.

Want the exact tool that stays invisible across every screen-share mode? Download Interview Helpers, or read the deeper stealth documentation. There's also a first-run wizard that walks you through a Zoom self-test so you can verify the invisibility before the real interview.