Chatgpt Users Just Got A Huge New Feature – Here's How To Use It
Author: Gammatek ISPL, Published: Sep 2026
ChatGPT can now read directly into your Google Drive, OneDrive, Dropbox, SharePoint, and Box accounts, search across all of them from a single chat, and pull real answers out of your actual files instead of whatever you manually copy and paste. That's a genuinely useful productivity upgrade. It's also, for any business running this on a managed device or a shared workspace, a new question about who can see what, once an AI assistant has standing read access to every cloud storage account your team uses. If you're rolling this out to a team rather than just your own account, the setup steps below matter less than the five minutes of permission-checking that should happen before anyone clicks "connect."
In this article
What ChatGPT's new connectors feature actually is
Step-by-step: how to connect and use it
Google Drive vs. OneDrive vs. Dropbox vs. SharePoint vs. Box, compared
What this means for teams running on QuickBooks or an ERP system
The security and compliance questions to ask before rolling it out
Implementation checklist for IT teams
FAQ
What ChatGPT's new connectors feature actually is
OpenAI has been building this out in stages. The original ChatGPT Connectors launch gave paid users the ability to link Google Drive, Dropbox, Box, SharePoint, and OneDrive directly to ChatGPT, letting the assistant search files, pull live data, and reference content without anyone uploading a single document by hand. As of this month, that capability extended further: ChatGPT's Windows desktop app now works with local "projects," folders on your computer that provide ongoing context for a task, and third-party bridge tools have emerged that let those local projects sync with cloud folders in Google Drive, Dropbox, and OneDrive, so a folder that technically lives in the cloud behaves like it's sitting right there on your desktop for ChatGPT's Codex agent to read and edit.
Here's what actually changed for a typical user. Before connectors, "ask ChatGPT about my files" meant uploading one document at a time into a chat, which worked for a single report but fell apart the moment you needed the assistant to cross-reference three spreadsheets and a slide deck stored in different places. Now, you can ask something like "compare last quarter's sales numbers in my Google Drive spreadsheet against the presentation I built in Dropbox" and get a direct answer, with citations that link back to the original file in its native app. OpenAI has also paired this with a "record mode" for meeting transcription, and Model Context Protocol (MCP) support, an open standard that lets an organization plug ChatGPT into internal tools and databases without a custom integration for each one.
The plan requirements matter here. Cloud storage connectors are available to ChatGPT Plus, Pro, Team, Enterprise, and Edu users, not the free tier, and some connector types are gated further, Team, Enterprise, and Edu users get a "Google Drive Synced Connector" that pre-syncs content for faster responses, a feature individual Plus subscribers don't have. Regional restrictions also apply to some individual-tier connectors, so availability isn't identical everywhere.
Step-by-step: how to connect and use it
Setting this up takes about two minutes per service, and the workflow is nearly identical across Google Drive, OneDrive, Dropbox, SharePoint, and Box.
Confirm you're on a paid plan. Cloud storage connectors require Plus, Pro, Team, Enterprise, or Edu. Free-tier accounts won't see the option at all.
Open the tools menu in a new chat. Look for the small tools icon under the message box, then select "Search connectors" or, for more complex multi-step questions, "Perform in-depth research."
Choose the service you want to connect. Select Google Drive, OneDrive, Dropbox, SharePoint, or Box from the list, then sign in and authorize access through that service's own login screen, not a ChatGPT-managed credential.
Set the scope, if your admin allows it. On Team and Enterprise plans, workspace admins can restrict which folders or drives are eligible for connection at all, which is worth checking before individual users start linking personal or shared drives on their own.
Ask your question in plain language. For example: "Show Q2 goals in my planning doc," or "Pull the top three takeaways from the report in my Dropbox," or "Compare this year's sales page against last year's and flag what changed."
Follow the citation back to the source. Every answer built from a connected file includes a citation. Click it to open the original file in its native app, Google Drive, OneDrive, whichever service it came from, so you can verify context ChatGPT might have missed.
Disconnect a service any time you don't need ongoing access. Access isn't all-or-nothing; you can connect a service for a single research task and remove it afterward from ChatGPT's connected-apps settings.
The feature is genuinely built for exactly the kind of task that used to eat twenty minutes of copy-pasting: pulling numbers out of a messy spreadsheet, comparing two document versions, or summarizing a report you don't have time to read end to end.
Google Drive vs. OneDrive vs. Dropbox vs. SharePoint vs. Box, compared
Not every cloud storage service behaves identically once connected, and the differences matter more for a business account than a personal one.
Service | Sync behavior | Best-suited plan tier | Admin-level access controls | Typical use case |
Google Drive | Optional pre-sync via "Google Drive Synced Connector" for faster responses | Team, Enterprise, Edu (individual Plus supported with restrictions) | Workspace admin can scope which drives are connectable | Cross-referencing shared docs, spreadsheets, and slides in real time |
OneDrive | On-demand search, not pre-synced | Plus and above | Follows Microsoft 365 permission hierarchy | Pulling from personal or team OneDrive folders already governed by existing Microsoft admin policy |
Dropbox | On-demand search, cloud storage read only | Plus and above | Limited native admin scoping compared to Microsoft/Google options | Quick document summarization and comparison for individual or small-team use |
SharePoint | On-demand search, respects existing SharePoint permission structure | Team, Enterprise, Edu | Strongest existing enterprise permission model of the group, inherited directly | Enterprise document libraries where access is already tightly role-based |
Box | On-demand search | Team, Enterprise, Edu | Enterprise-grade access controls, similar tier to SharePoint | Regulated industries already using Box specifically for its compliance certifications |
The practical takeaway: SharePoint and Box carry over the strongest existing permission structures because both were built for enterprise access control from the start, which makes them the safer default for regulated data. Dropbox and personal-tier Google Drive or OneDrive accounts have thinner admin controls, so a connector there is only as safe as whatever permission structure already existed on that specific folder, which for a lot of small teams is looser than anyone realizes until they actually check.
What this means for teams running on QuickBooks or an ERP system
This is the section where I'll be direct about the stretch: connectors today cover cloud storage services, Google Drive, OneDrive, Dropbox, SharePoint, Box, not accounting platforms or ERP systems directly. There's no current "QuickBooks Enterprise connector" or built-in ERP integration shipping with this feature. But the underlying mechanism that makes today's connectors possible, standing, permissioned, read access to an external system, through the Model Context Protocol OpenAI is building out for business customers, is exactly the architecture that would eventually make a QuickBooks or ERP connector technically straightforward to build. MCP already lets organizations link ChatGPT to internal tools and databases without a bespoke integration for each one, which is the same underlying capability, aimed at a different category of system.
That's worth planning around now rather than reacting to later, for two reasons. First, if your business runs QuickBooks Enterprise, Oracle, SAP, or another ERP system as its system of record, the same governance questions this article raises for cloud storage, who can connect what, what gets logged, what a workspace admin can restrict, will apply just as directly the moment financial or ERP data becomes connectable. Second, the search demand is already there: "QuickBooks Enterprise" and "ERP" both carry meaningfully higher advertiser interest than most general software terms, reflecting how much businesses are actively researching how AI tools intersect with the systems that run their operations. If you're building an AI governance policy today, writing it to cover "any connected business system," not just "cloud storage," saves you from rewriting it the day an ERP connector actually ships.
The security and compliance questions to ask before rolling it out
A connector isn't a one-time file upload, it's standing access. That distinction is the entire reason this deserves a conversation with IT or compliance before a team rolls it out broadly, not after someone connects a shared drive they didn't fully think through.
Does the connected drive contain anything outside the requester's normal need-to-know? A personal Google Drive rarely has this problem. A shared team Drive or SharePoint library often does, because access accumulates over time and nobody audits it regularly.
Is "Improve the model for everyone" enabled by default on the accounts connecting? OpenAI has stated it doesn't use connector data from Team, Enterprise, or Edu accounts to train its models, but that setting is enabled by default for Free, Plus, and Pro users, who can disable it manually. That's a one-time setting check worth making standard onboarding for anyone connecting a personal-tier account to work files.
Who can see the connection exists, and who can revoke it? On Team and Enterprise plans, a workspace admin can restrict which folders are eligible for connection. On individual Plus accounts, that decision sits entirely with the person who clicked "connect," which is a gap worth closing with a written policy if you can't close it with an admin setting.
Does your data retention or compliance obligation require logging AI access to specific files? If you're in a regulated industry, "an AI assistant read this file" may need to be a logged event the same way a human accessing that file would be, and that's not something the connector feature logs for you automatically at the individual-user tier.
None of this is a reason to avoid the feature. It's the same category of question you'd ask before turning on any tool with standing access to shared company data, and the answer is almost always "yes, roll it out, with a written scope."
A realistic rollout scenario, and where it usually goes wrong
The pattern we've seen with clients adopting cloud-connected AI tools generally follows a predictable arc, and it's worth walking through because the failure point is rarely the technology itself. A team lead, often in marketing, sales, or finance, discovers the connector feature, connects their own Google Drive or OneDrive in about ninety seconds, and immediately finds it genuinely useful: faster report summaries, quicker cross-referencing between spreadsheets, less time hunting for the right version of a document. Word spreads within the team. Within a week or two, four or five more people have connected their own accounts, entirely reasonably, because nothing stopped them and the feature is easy to find.
The problem shows up about a month later, usually during an unrelated security review or a client's vendor questionnaire, when someone asks a simple question nobody planned for: which of our shared drives are now accessible through a third-party AI tool, and who authorized that? At that point, reconstructing the answer means checking each individual user's connected-apps settings one at a time, because the connections happened at the individual level rather than through a central admin decision. That's not a failure of the feature, connectors work exactly as documented, it's a failure of sequencing: the convenience arrived before the governance did, which is the same sequencing problem that shows up with almost every fast-adopted productivity tool, not just this one.
The fix is straightforward and doesn't require slowing anyone down for long: a same-week policy decision on which connector types are approved by default, which require a request, and a single place, ideally a workspace-admin-level setting rather than an honor system, where that's enforced instead of assumed. Teams that make that decision in week one instead of month two spend a fraction of the time on the eventual audit, because there's an answer on file instead of a reconstruction project.
Implementation checklist for IT teams
Decide connector scope by plan tier before individual users start connecting accounts. Team and Enterprise admins have real controls here; make the decision centrally rather than letting it happen ad hoc.
Audit shared drive and SharePoint permissions before connecting them, not after. A connector doesn't create a new access problem, it just makes an existing over-permissioned folder searchable in a new interface.
Set the "Improve the model for everyone" toggle explicitly as part of onboarding, rather than assuming individual users will find and adjust it themselves.
Write your AI governance policy to cover "any connected system," including ERP and accounting platforms, even though those aren't connectable yet, so you're not rewriting policy the day that changes.
Pilot with a single team and a single connector type first. Google Drive or SharePoint, whichever already has the tighter permission structure, is the safer starting point before expanding to Dropbox or personal-tier accounts.
Revisit connector access quarterly, the same way you'd revisit any standing third-party access to company data, since permissions and personnel both change faster than most access reviews account for.
Frequently asked questions
Is the ChatGPT connectors feature free? No. Cloud storage connectors require a paid plan, Plus, Pro, Team, Enterprise, or Edu. Free-tier ChatGPT accounts don't have access to this feature.
Which cloud storage services can ChatGPT connect to right now? Google Drive, OneDrive, Dropbox, SharePoint, and Box are all supported. Google Drive also has an additional "Synced Connector" option for Team, Enterprise, and Edu users on a Google Workspace plan, which pre-syncs content for faster responses.
Does ChatGPT use my files to train its models once I connect a cloud storage account? For Team, Enterprise, and Edu accounts, OpenAI states connector data isn't used to train its models. For Free, Plus, and Pro accounts, the "Improve the model for everyone" setting is enabled by default but can be turned off manually in account settings.
Can ChatGPT connect to QuickBooks or an ERP system? Not yet, as a built-in connector. The current feature covers cloud storage services specifically. The Model Context Protocol OpenAI is building for business customers is the same underlying technology that would make an ERP or accounting-platform connector feasible in the future, but no such connector currently ships with ChatGPT.
What's the biggest risk of connecting a shared team drive instead of a personal one? Permission sprawl. Shared drives and SharePoint libraries accumulate access over time, often beyond what any single admin has recently audited, so connecting one makes everything in it searchable through a new interface, which is a good reason to audit the folder's existing permissions before connecting rather than after.
Do I need my company's IT approval to use this? If you're on a personal ChatGPT account, technically no, but if you're connecting work files or accounts governed by your employer's data policy, checking first is the safer move, especially since Team and Enterprise admins may already have restrictions in place you're not aware of.
Not sure whether your team's cloud storage permissions would hold up if someone connected them to ChatGPT tomorrow?
Our AI governance and compliance review audits your current cloud storage and file-sharing permissions, flags where access has drifted beyond what's needed, and builds a written connector policy before your team rolls this feature out company-wide.
Book an AI governance and compliance review: https://www.gammateksolutions.com/post/amazon-prime-members-just-got-a-huge-new-perk-here-s-how-to-use-it https://www.gammateksolutions.com/post/america-must-learn-ai-lessons-from-astro-boy https://www.gammateksolutions.com/post/epicenter-tech-security-breach-2024-2026-exposed-data-enterprise-risk
Comments