Introduction
Most candidates treat screen sharing like a light switch. Either it's on or it's off. They click 'Share Screen' when the interviewer asks, hand over control, and hope nothing embarrassing is visible in the background.
That's not how screen sharing works. And in a technical interview where you're sharing your screen for a live coding problem, or using an AI overlay tool to help you answer questions in real time, not understanding the mechanics is the kind of mistake that ends the interview before you've finished your answer.
Display capture, window capture, multi-monitor setups, virtual displays, and OS-level overlay exclusions all behave differently. Zoom, Teams, and Google Meet each handle screen capture with different APIs. What appears on your interviewer's screen depends on decisions you make — or don't make — before you click Share.
This piece covers every layer: the technology behind screen capture, the difference between what you see and what they see, how invisible overlays actually work, and how to configure your setup so you control every pixel that leaves your machine.
How Screen Sharing Actually Works: The Technical Layer Most Candidates Skip
When you share your screen on Zoom, Teams, or Google Meet, the platform doesn't take a photo of your monitor. It captures a video stream frame by frame from either your full display, a specific application window, or a browser tab, depending on which mode you select.
That distinction matters more than most candidates realize. Here's what each mode actually captures:
| Capture Mode | What Gets Shared | Risk Level for Overlay Tools | Best For |
|---|---|---|---|
| Display Capture (full screen) | Everything on the selected monitor — all windows, tabs, notifications, overlay apps | High — everything visible is captured | Presentations with multiple windows |
| Window Capture (application) | Only the selected application window — other apps are excluded | Low — all other windows excluded | Sharing a single browser or IDE |
| Browser Tab Capture | Only the selected tab — nothing outside the browser | Very Low — most isolated | Sharing a specific webpage or doc |
| Virtual Display | A separate simulated monitor — only what you put there | Minimal — completely isolated | Advanced setups with AI overlay tools |
The API Layer: Why Zoom, Teams, and Meet Behave Differently
Each platform uses a different OS-level API to capture screen content. This is why the same overlay tool can be invisible on one platform and fully visible on another.
On Windows, most screen-capture APIs respect the WDA_EXCLUDEFROMCAPTURE flag — a system-level setting that tells the OS to exclude a specific window from any capture stream. Applications that set this flag correctly become invisible to Zoom, Teams, and Meet when those platforms capture using the standard Windows Graphics Capture API.
On macOS, the equivalent is the CGWindowListOption exclusion. Apps that register at the correct layer in the macOS window hierarchy can sit above your visible workspace while remaining outside the capture boundary that screen-sharing software reads.
The catch is that DirectX captures, GPU frame grabs, and kernel-level screen-recording tools bypass these flags entirely. If an interviewer's company uses proctoring software that operates at the kernel level, standard flag-based exclusions don't protect you.
Zoom, Teams, and Google Meet: How Each Platform Captures Your Screen
Zoom
Zoom offers three sharing modes: full screen, application window, and advanced options including portion of screen and second camera. When you select 'Share Screen' and choose your full desktop, Zoom captures the entire display at up to 1080p, 30 frames per second.
The important Zoom-specific setting is 'Advanced capture with window filtering', available in Zoom's share settings. When this is enabled, Zoom uses the Windows Graphics Capture API with window filtering active, which means windows that have set the WDA_EXCLUDEFROMCAPTURE flag become invisible in the share stream. GirGit AI's overlay sets this flag by default.
Without this setting, Zoom falls back to legacy GDI capture, which sees everything — including windows flagged for exclusion.
| Zoom Share Mode | What Interviewer Sees | Overlay Visibility |
|---|---|---|
| Full Screen (default) | Everything on your monitor | Visible unless overlay is OS-excluded |
| Application Window | Only the selected app | Overlay not visible (different window) |
| Advanced: Window Filtering ON | Full screen minus excluded windows | Invisible if overlay sets exclusion flag |
| Advanced: Legacy / GDI mode | Everything, no exclusions respected | Visible regardless of overlay flags |
Microsoft Teams
Teams operates similarly to Zoom but defaults to full desktop sharing in most corporate configurations. The distinction that matters for Teams specifically: enterprise accounts often have policies that force full desktop sharing — you can't switch to window-only mode without admin permission.
This is the environment where AI overlay detection risk is highest. Corporate Teams deployments with IT-managed configurations may also run monitoring software alongside the call that operates at a different capture layer than Teams itself.
For candidates interviewing at enterprise companies using managed Teams environments — financial services, healthcare, government — the safest approach is a virtual display setup, not relying on window filtering alone.
Google Meet
Meet offers three sharing options: entire screen, a window, or a tab. Tab sharing is the most isolated; only the selected browser tab is captured, and nothing outside it is visible. For interviews where you are sharing a coding environment or document in a browser, tab sharing eliminates most overlay detection risk.
Meet's capture uses Chrome's screen-capture API on desktop, which respects application-level exclusions in most configurations. However, the 'Entire Screen' option on Meet behaves identically to display capture — everything visible on the selected monitor.
| Platform | Default Share Mode | Window Filtering Support | Tab Sharing Available | Enterprise Policy Risk |
|---|---|---|---|---|
| Zoom | Full screen | Yes (manual setting) | No | Low–Medium |
| Microsoft Teams | Full screen | Limited | No | High in managed environments |
| Google Meet | Full screen | Via Chrome API | Yes | Low–Medium |
Note: sharing behavior may vary based on enterprise security policies and platform updates.
Multi-Monitor Setups: The Setup That Gives You Full Control
A dual-monitor setup is the single most effective screen-sharing strategy for interviews. Not because it is complicated, but because it is simple: you share one screen, you work on the other.
The interviewer sees Monitor 1 — your browser, your IDE, your shared document. Monitor 2 stays off the share. GirGit AI runs on Monitor 2. You answer questions, your overlay surfaces responses, and nothing on Monitor 2 enters the share stream because you never selected it.
No flags, no API exclusions, no technical configuration. One monitor shared, one monitor private. The capture boundary is a physical one.
Single-Monitor Setup: Making It Work Without a Second Screen
No second monitor? The next best option is window-only sharing. Share just your coding environment or browser, not your full desktop. Then position GirGit AI overlay outside the boundary of the shared window.
On a single monitor, this means the overlay window needs to sit in a part of the screen not covered by the shared application. If your IDE takes up the left two-thirds of your screen and GirGit AI sits in a small window on the right edge outside the shared window's boundaries, window-capture mode won't pick it up.
The limitation: if the interviewer asks you to share your full screen at any point, this setup collapses. Always have a response ready — 'I'll need to rearrange my windows for a moment' buys you 20 seconds to minimize the overlay before going full screen.
Virtual Displays: The Advanced Setup for Maximum Control
A virtual display is a software-simulated monitor. It appears to your OS as a second screen, but nothing is physically connected. You share the virtual display in the interview; it shows the interviewer only what you explicitly put there. GirGit AI runs on your real physical display, completely outside the share stream.
Setting Up a Virtual Display on Windows
- Install a virtual display driver — options include Virtual Display Driver or IddSampleDriver.
- The OS registers it as Display 2.
- Open your browser, coding environment, or shared document on the virtual display.
- In Zoom or Teams, select Display 2 to share.
- GirGit AI remains on Display 1, your physical monitor, outside the capture entirely.
Setting Up a Virtual Display on macOS
- Use a display emulator dongle (hardware) or a software tool like BetterDummy.
- BetterDummy creates a virtual display that appears in System Preferences as a real monitor.
- Route your shared content to the virtual display.
- Share that display in Meet or Zoom.
- Your physical display, with GirGit AI on it, never enters the share stream.
How Invisible Overlays Actually Work: The Real Technical Picture
Every AI interview tool claims to be invisible. The mechanics behind that claim vary significantly, and understanding them tells you which tools actually deliver and which ones rely on hope.
There are three ways an overlay can be invisible to screen-sharing software:
| Method | How It Works | Reliability | Risk |
|---|---|---|---|
| OS-Level Flag (WDA / CGWindow) | Sets a system flag telling capture APIs to exclude this window | High — on standard capture APIs | Fails against kernel-level capture or DirectX grabs |
| Off-Screen Positioning | Overlay renders outside the shared window or display boundary | High — physical boundary | Collapses if user shares full screen accidentally |
| Virtual Display Isolation | AI tool runs on a separate display that isn't shared | Very High — architectural separation | Requires setup; no runtime risk once configured |
| Kernel-Level Rendering | Operates below user-space capture layer — invisible to all user-space tools | Highest | Ethical and policy considerations apply |
GirGit AI uses OS-level flag exclusion combined with architectural positioning; the overlay is designed to sit outside the capture boundary while setting the system exclusion flag as a secondary layer of protection. For most interview environments, this is sufficient. For managed enterprise environments with kernel-level monitoring, the virtual display setup is the recommended approach.
What Interviewers Can Actually Detect, and What They Can't
The technical invisibility of an overlay doesn't mean zero detection risk. Experienced interviewers have other signals available to them that don't depend on screen capture at all.
| Detection Method | What It Catches | Can Overlay Tools Avoid It? |
|---|---|---|
| Screen capture | Visible overlays on shared display | Yes — with correct setup |
| Eye movement patterns | Looking off-screen consistently while 'thinking' | Behavioral — depends on how you use the tool |
| Response timing | Unusually fast or templated answers to complex questions | Behavioral — depends on delivery |
| Unnatural pauses | Long gap before a perfectly structured answer | Behavioral — practice helps |
| Proctoring software (user-space) | Running applications, browser tabs | Yes — OS-excluded windows not visible |
| Proctoring software (kernel-level) | Running processes, GPU frame grabs | No — kernel tools bypass API exclusions |
| Code edit history | Idle time followed by an instant perfect solution in coding tools | No — coding tool logs are separate |
The practical conclusion: technical detection risk from screen capture is manageable with the right setup. Behavioral detection risk — how you respond, where your eyes go, how your timing patterns look — is entirely in your control. GirGit AI surfaces answers as a foundation, not a script. The candidate who reads it word-for-word sounds different from the one who absorbs it and responds naturally.
The Optimal Screen-Sharing Setup for Each Platform
Zoom Interview — Recommended Setup
- Enable 'Advanced capture with window filtering' in Zoom share settings before the interview.
- Share a specific application window, not your full desktop.
- Position GirGit AI outside the shared window boundary on a single monitor, or on Monitor 2 in a dual-monitor setup.
- Test the setup on a private Zoom call before interview day — record yourself and check the recording.
- If asked to share full screen, minimize GirGit AI first, then share.
Microsoft Teams Interview — Recommended Setup
- Check whether the company uses managed Teams policies before the interview; if corporate-managed, assume full desktop sharing only.
- Use a virtual display setup if the role is at a large enterprise or in a regulated industry.
- In unmanaged Teams environments, window sharing works — share only your coding environment or document.
- Keep GirGit AI on a separate monitor or in an off-screen position.
- Assume kernel-level monitoring may be active in high-stakes enterprise interviews.
Google Meet Interview — Recommended Setup
- Use tab sharing when the interview involves a browser-based coding environment or shared doc; this is the most isolated option.
- For full-application sharing, use window capture, not 'Entire Screen'.
- GirGit AI runs outside the browser tab in all tab-sharing scenarios — zero detection risk.
- Test on a private Meet call and check what appears by asking a friend to screenshot what they see.
How GirGit AI Handles Screen Sharing: What Actually Happens
GirGit AI is a desktop application. It doesn't live inside your browser. That distinction matters for screen capture because browser extensions are among the easiest overlay tools to detect — they appear inside the browser chrome, and the toolbar icon shows up the moment you share a browser window.
As a desktop application, GirGit AI's window exists outside the browser stack entirely. It registers with the OS, sets the appropriate exclusion flag for the platform you're on, and renders in a position designed to stay outside standard capture boundaries.
| Interview Platform | GirGit AI Behaviour | Interviewer Visibility |
|---|---|---|
| Zoom (window filtering ON) | OS exclusion flag active — invisible to capture stream | Not visible |
| Zoom (window filtering OFF) | Positional exclusion — share app window only | Not visible if window-sharing used |
| Microsoft Teams (standard) | OS exclusion flag active | Not visible in standard capture |
| Microsoft Teams (enterprise managed) | Recommend virtual display — flag may not apply | Use virtual display for guarantee |
| Google Meet (tab sharing) | Desktop app outside browser entirely | Not visible — different capture surface |
| Google Meet (full screen) | Positional exclusion on second monitor | Not visible with dual-monitor setup |
The Pre-Interview Screen-Sharing Checklist
Run this before every interview. It takes four minutes. It removes 95% of screen-sharing risk.
24 Hours Before
- Confirm which platform the interview uses — Zoom, Teams, or Meet.
- Check if the company is using managed enterprise software (check the meeting invite domain and invite format).
- Set up your sharing mode: window capture or virtual display, based on platform and risk level.
- Update GirGit AI to the latest version — overlay exclusion flags update with OS versions.
30 Minutes Before
- Open a private test call on the same platform and record it.
- Enable GirGit AI overlay and run through a sample question.
- Review the recording — confirm the overlay is not visible in the capture.
- Close unnecessary applications that could appear in notifications or as visible windows.
- Enable Do Not Disturb — notification banners appear in both display capture and window capture.
During the Interview
- Use window or tab sharing, not full-screen sharing, unless specifically required.
- If asked to share full screen, say 'give me one moment to arrange my windows' and minimize GirGit AI before sharing.
- Keep eye-contact patterns natural — glance at the overlay the way you would glance at notes, not the way you would copy from a screen.
- Let GirGit AI surface the answer, then put it in your own words before speaking.
For Recruiting and HR Teams: What AI Overlay Tools Mean for Your Process
After focusing on the candidate side, there's a question on the other side of the call that recruiting and talent acquisition teams are increasingly asking: how do we design interview processes that work in a world where AI overlay tools exist?
The answer isn't to build detection infrastructure. Kernel-level proctoring software introduces legal complexity, candidate trust issues, and false-positive rates that create more problems than they solve. The better answer is to design interviews that test what the candidate actually knows, not how well they perform under artificial pressure.
What This Means for Hiring Processes in 2026
- Conversational follow-up questions that test understanding, not recall — an AI tool can surface a framework answer, but it can't answer a follow-up that references something the candidate just said.
- Live problem-solving sessions where the interviewer guides and asks 'why' at each step — AI answers don't explain reasoning under real follow-up pressure.
- Async video interviews for first-round screening — removes the real-time AI advantage entirely while still capturing communication quality.
- Technical assessments with edit-history logging — AI tools can't eliminate the timing and edit-pattern signals that live coding environments capture.
GirGit AI's B2B team tools give recruiting operations a different angle: using AI assistance to level the playing field across a candidate pool, rather than trying to eliminate it. When every candidate has access to the same real-time support, the evaluation signal shifts to what matters — communication quality, reasoning under follow-up pressure, and how candidates engage with unexpected questions.
Frequently Asked Questions
Can Zoom detect GirGit AI running during an interview?
In standard Zoom configurations with window filtering enabled, GirGit AI is not visible in the screen-share stream. Zoom's standard capture API respects the OS-level exclusion flags that GirGit AI sets. Kernel-level capture tools, which most standard Zoom interviews don't use, operate differently and bypass these exclusions.
What is the safest screen-sharing mode for using an AI overlay tool?
Tab sharing on Google Meet is the most isolated — only the browser tab is captured, and a desktop application like GirGit AI is completely outside that surface. Window capture on Zoom with window filtering enabled is the next safest. Virtual display setups provide architectural separation that does not rely on any API behavior.
Does display capture share everything on my screen?
Yes. Display capture shares the full contents of the selected monitor — every visible window, notification, taskbar item, and application. If GirGit AI is visible on that monitor and window filtering is not active, it will appear in the share stream. This is why window capture or tab sharing is always preferable in interview settings.
What happens if the interviewer asks me to share my full screen?
Minimize GirGit AI before switching to full-screen sharing. Say 'give me one moment to arrange my windows' — this is a normal, professional thing to say. Once the overlay is minimized and off the visible desktop, switch to full-screen share. GirGit AI's audio capture continues in the background even when the visual overlay is hidden.
How is GirGit AI different from browser-based AI overlay tools for screen sharing?
Browser-based tools live inside the browser — their interface appears within the browser window, and the toolbar icon is visible when you share a browser window. GirGit AI is a desktop application that exists outside the browser stack entirely. This means window and tab sharing modes on any platform do not include GirGit AI interface in the capture at all.