Claude, ChatGPT, Gemini and Copilot: Verified Features for Windows 11 AI Workflows

Sherman's piece frames the four assistants around distinct, everyday strengths: Claude for natural prose and context retention, ChatGPT as a broad all-rounder for data analysis and research, Gemini for users living inside Google's apps, and Microsoft Copilot for Windows 11 and Microsoft 365 environments. He ultimately named Google Gemini as his personal daily driver due to its ties to Gmail, Docs, Drive, Sheets, Maps, and Photos.

That comparison is a useful starting point for how people actually assemble their workflows, though it reflects personal editorial preferences rather than a controlled, repeatable benchmark test with published prompts or scoring criteria. Even more importantly for Windows users and IT administrators, the AI landscape moves fast enough that several broad assumptions about what these tools can and cannot do no longer match current vendor documentation. Before you start paying for multiple subscriptions or pinning new icons to your Windows 11 taskbar, it pays to look at the verified technical boundaries, desktop footprints, and administrative controls that actually govern these assistants today.

Fact-Checking the Big Four: What the Documentation Actually Says​

When evaluating whether you need a second AI assistant to cover a gap in your primary tool, the first step is making sure that gap still exists. Two of the specific limitations cited in The Star's comparison—that Claude lacks built-in web search and that ChatGPT has no native desktop footprint—have been overtaken by product updates documented by Anthropic and OpenAI.

1. Anthropic's Claude and Live Web Access​

In his overview, Sherman praised Claude for writing in a natural, human tone that avoids cliché AI buzzwords, and noted its strong recall across prompts, while pointing to strict usage caps on free and paid tiers and a lack of direct, built-in web search integration.

According to Anthropic's Claude Help Center documentation, however, Claude now includes both built-in web search and a "web fetch" feature across a wide lineup of models (including Haiku 4.5 and 5.5, Sonnet 4.6, 5, and 5.5, Fable 5 and 5.1, and Opus 4.6 through 5.5). When enabled—either via the + menu in the chat window or automatically in the newer Claude interface—Claude queries the live web, returns direct source citations, and can even display image results powered by Microsoft Bing. With web fetch, users can also paste a direct URL for Claude to retrieve and analyze the full page text.

Where The Star's warning hits home is on capacity limits. Anthropic's documentation explicitly notes that free accounts operate on usage limits that reset every five hours, and both web search and web fetch count directly toward those caps. Pulling a 10,000-word article into Claude's context window via a URL consumes substantially more quota than a standard prompt. On Team and Enterprise plans, an organization Owner or Primary Owner must first enable web search at the workspace level under Organization settings > Capabilities before individual users can toggle it on.

2. OpenAI's ChatGPT on the Desktop​

Sherman characterized ChatGPT as the strongest overall generalist for creative writing, research, complex datasets, and charts, while arguing that it relies on templated outputs, lacks deep operating-system integration, and has no native desktop footprint.

Here it is important to separate a native desktop application from operating-system-level integration. ChatGPT is not wired into the Windows 11 shell the way Microsoft Copilot is, and it cannot natively toggle your Windows system settings. However, OpenAI's Help Center confirms that ChatGPT does offer a dedicated desktop app for Windows and macOS.

In fact, the ChatGPT desktop app includes a built-in browser equipped with "site tools" based on WebMCP, a proposed web standard. According to OpenAI's documentation, when you open a supported website inside the ChatGPT desktop app's browser and sign in, an arrow indicator in the address bar shows available site tools that allow ChatGPT to search documentation, edit shared notebooks or documents, explore dashboards, or update shopping carts directly within that live session. OpenAI notes that this built-in browser maintains its own session state separate from Chrome, and users can disable the feature at any time via Browser settings > Permissions > Enable site tools.

3. Google Gemini and Workspace Boundaries​

Sherman's preference for Gemini rests on its connections across Google's ecosystem, alongside strong image and video generation and generous free access, though he cautioned that Gemini can hallucinate facts or dates and often demands precise prompting.

Google's Gemini Apps Help documentation backs up both the ecosystem reach and the accuracy caveat, explicitly warning users that Gemini can hallucinate or pull outdated information—such as referencing an older email thread when a newer one exists—and advising users to verify linked sources.

However, "seamless integration" is not automatic or unlimited, especially in workplace environments. According to Google's support documentation for Connected Apps and Workspace integrations:

• Admin and License Gates: Using Connected Apps (which cover Gmail, Google Calendar, Google Chat, Docs, Drive, Keep, and Tasks) with a work or school account requires a qualifying Google Workspace edition, an active Keep Activity setting, and explicit enablement by a Google Workspace administrator.
• Action Limits: Even when connected, Gemini cannot perform every action inside those apps. Google's documentation notes that via Connected Apps, Gemini cannot create email drafts or delete emails in Gmail, cannot read images or comments in Docs or Gmail, cannot access photos or videos stored in Drive, and cannot create folders or move Drive files.
• Third-Party Read-Only Rules: While Gemini in Workspace can connect to third-party platforms like Asana, Atlassian Rovo (Jira and Confluence), HubSpot, Mailchimp, Monday.com, QuickBooks, and Salesforce, Google states that these integrations currently support read actions only—meaning you can query a Jira ticket or check a Salesforce pipeline, but you cannot use the integration to modify records in those third-party systems.

4. Microsoft Copilot on Windows 11 and Microsoft 365​

For Windows users, The Star's observations about Microsoft Copilot touch on a familiar tension: Copilot shines inside Word, Excel, PowerPoint, Outlook, and Teams, and sits deeply embedded inside Windows 11, but unlocking its full workplace capabilities requires additional Microsoft 365 licensing, and its desktop presence can feel intrusive if not tamed.

Microsoft's current support documentation spells out exactly what that Windows 11 footprint looks like today—and how much control users and administrators actually have over it. Notice that Microsoft has also consolidated its branding: what was previously called the "Microsoft 365 Copilot app" is now simply the Microsoft Copilot app across Windows, Mac, web, iOS, and Android.

On new Windows 11 PCs, the Copilot app is installed by default, though it remains a standard app that can be downloaded from the Microsoft Store or uninstalled via Settings > Apps > Installed apps. Its documented Windows-specific features include:

• Hardware and Shortcut Triggers: Launching via the dedicated Copilot keyboard key, the Windows + C shortcut (configurable between full app and a compact "quick view"), a configurable Windows Ink pen shortcut button, or the opt-in "Hey, Copilot" voice wake word.
• Local File Search and Reading: Finding and answering questions about local and OneDrive-synced files on the PC. Microsoft's documentation specifies that file reading supports .docx, .xlsx, .pptx, .txt, .pdf, and .json formats.
• Copilot Vision and Screenshots: Using Copilot Vision to inspect open browser windows or applications for step-by-step guidance, or using the + > Take a Screenshot tool in the Composer to send a selected region of the screen.
• Windows Settings Guidance: Guiding users directly to relevant pages in the Windows Settings app when asked to adjust display readability or focus settings.

Crucially, Microsoft documents a strict boundary between consumer and enterprise accounts inside this single app. You can verify which mode you are running by checking the Work (Microsoft Entra ID) or Personal (Microsoft Account) badge next to your profile icon in the left navigation pane. Data entered in a personal session does not cross into a work or school session, and features like in-app Microsoft 365 file creation, Work IQ context, and Bing web search grounding depend on assigned licenses and administrator policies.

Summary: Claims vs. Documented Reality​

The Multi-Assistant Trade-Off: Versatility vs. Context Fragmentation​

Is pairing two AI assistants actually the smartest strategy? Industry perspectives on that question are split right down the middle.

On one side, workflow guides and power users frequently advocate combining a workspace-native assistant with a specialized reasoning or writing tool—such as pairing Microsoft Copilot for internal company documents with Claude or ChatGPT for long-form drafting or outside research. Because an AI assistant is only as capable as the data it can reach, using Gemini inside a Google-heavy environment or Copilot inside a Microsoft 365 tenant solves the "last mile" problem of accessing your emails, spreadsheets, and calendars without endless copy-pasting. Adding a second tool then lets you pick a preferred writing style or analysis interface when you don't need workspace data.

On the other side, recent productivity analyses warn against "tool sprawl." As commentary in inkl and workplace testing from firms like Lifestack point out, juggling multiple AI apps can increase subscription costs, fragment your conversation history across disconnected silos, and add yet another round of context-switching to a workday where knowledge workers already lose hours moving between tabs. When your project notes live in Copilot, your brainstorming lives in ChatGPT, and your email summaries live in Gemini, neither assistant has the full picture unless you manually bridge the gap.

In practice, the most sensible rule for Windows users is ecosystem first, specialist second:

1. Start with the assistant wired into your primary identity and storage stack (Microsoft Copilot if your day revolves around Windows 11, OneDrive, and Microsoft 365; Google Gemini if your organization runs on Google Workspace).
2. Audit its actual permissions and settings before assuming it can or cannot do a task.
3. Add a second assistant only when a specific workflow hits a wall—for instance, if your corporate administrator has disabled Bing web grounding in Copilot Chat and you need an approved external research tool, or if you strongly prefer another model's prose style for drafting public-facing text.

Practical Guide: Configuring and Controlling Copilot on Windows 11​

If you decide to use Microsoft Copilot as one half of your AI pairing on a Windows 11 PC—or if you want to stop it from popping up uninvited while you use another assistant—Microsoft's support documentation provides direct controls for its shortcuts, local file access, and web search behavior.

Step 1: Verify Your Account Context (Work vs. Personal)​

1. Open the Microsoft Copilot app on your Windows PC.
2. Look at the bottom of the left navigation pane next to your profile avatar.
3. Check the badge label: Work indicates a Microsoft Entra ID work or school login (governed by your organization's licensing and enterprise data protection), while Personal indicates a consumer Microsoft Account.

Step 2: Tame Keyboard, Voice, and Pen Triggers​

If Copilot feels like an unprompted distraction on Windows 11, you do not have to live with accidental key presses:

1. Adjust Keyboard Shortcuts: Inside the Copilot app, go to Account > Settings > Copilot Keyboard Shortcuts to configure how the Copilot key or Windows + C behaves (such as switching between the full app and a smaller quick view).
2. Manage the Wake Word: The "Hey, Copilot" voice trigger is an opt-in feature. Ensure it is toggled off in your Copilot account settings if you do not want hands-free listening active.
3. Reassign Your Stylus Button: If a Windows pen keeps launching Copilot, open Windows Settings > Bluetooth & Devices > Pen & Windows Ink, expand Choose what your shortcut button does, and change it away from Same as Copilot key.
4. Full Removal (If Needed): If you rely entirely on third-party assistants, you can uninstall the client on Windows 11 via Settings > Apps > Installed apps, selecting the three-dot More options (...) menu next to Copilot, and choosing Uninstall.

Step 3: Lock Down Local File Search and Web Grounding​

Before pairing Copilot with local files or external research, review what data the app is allowed to read:

• Local File Access: In the Copilot app on Windows, navigate to Account > Settings > File Search and File Read. Here you can toggle whether Copilot is allowed to index and read local and synced OneDrive files (.docx, .xlsx, .pptx, .txt, .pdf, and .json).
• Bing Web Search in Copilot Chat: When using Copilot Chat, web search generates short Bing queries from your prompt or uploaded file. To toggle this on or off, click the ... (Settings and more) menu in the top-right corner of Copilot Chat, select Chat settings > Personalization, scroll to the bottom, expand Advanced, and switch the Web search toggle. (Note: On work or school accounts, if your IT administrator has disabled web search at the tenant level, this toggle will be locked and display an organizational notice.)

Step 4: Check Capacity and Source Links on Your Secondary Tool​

When handing off tasks to a secondary assistant, keep these documented guardrails in mind:

• Conserving Claude Quota: If you use a free Claude account alongside Windows Copilot, toggle Web search off via the + button in the lower-left corner of the chat box whenever you do not need live internet data, and avoid pasting direct links to massive articles unless necessary, so you don't burn through your five-hour reset window.
• Invoking Gemini Connected Apps: If Gemini is your secondary tool, remember that in every new conversation you must explicitly type @ (for example, @Google Drive) to pull workspace context, and always click the listed source links beneath a response to verify dates and figures against hallucinations.

Pairing two AI assistants can indeed give you the best of both worlds—combining deep operating-system and productivity-suite hooks with a second opinion for drafting or analysis. Just make sure you are building that workflow around what these platforms actually support today, rather than last year's spec sheet.