ChatGPT · Connectors: bringing your drive and tools in
Connectors let ChatGPT read from your work tools (Google Drive, SharePoint, OneDrive, GitHub, and more) so that answers come from your own documents instead of the model's general training. You ask a question in plain language. ChatGPT searches the connected source, pulls the relevant passages, and answers with citations back to the original files.
A naming note first. In December 2025 OpenAI renamed connectors to apps. They now live in the same place as interactive apps built with the Apps SDK (Canva, Figma, Zillow and others). In ChatGPT you will see Drive, SharePoint and GitHub listed under Settings, then Apps. Many people and older docs still say "connectors", and this lesson uses both terms. The mechanics are the same.
This is RAGRAGA method that lets an AI model answer using your own documents, retrieving relevant passages before generating a response instead of relying only on training data.View full definition →, but you are not building the pipelinepipelineAll active sales opportunities across the stages of the sales process, together with their combined potential value and probability of closing.View full definition →. OpenAI hosts the indexing, retrieval, and permission checks. You only authorize the connection.
What a connector actually does
A connector is an authorized bridge between your ChatGPT account and an external data source. When you connect Google Drive, you grant ChatGPT permission (via OAuth) to search and read files on your behalf.
Two modes matter:
- Synced or indexed search. Some connectors index your content ahead of time so ChatGPT can run fast semantic searchsemantic searchSearch that matches on meaning and intent rather than exact keywords, so a query and a result can connect even when they share no words.View full definition → across many files. This powers "search all my docs" behavior.
- On-demand fetch. ChatGPT pulls a specific file when the conversation calls for it, for example when you mention a particular spreadsheet.
The model does not get your entire Drive dumped into the context windowcontext windowThe context window is the maximum amount of text (measured in tokens) a language model can process at once, including both the input prompt and the generated output.View full definition →. It runs a search, gets back the most relevant chunks, and reasons over those. Your foundations lessons covered why: the context window is finite, so retrieval beats stuffing.
The OpenAI help center keeps the current list of supported sources, which grows regularly.
A concrete "answer from our internal docs" example
Say your team keeps everything in Google Drive: a Pricing Policy 2026.gdoc, a folder of signed customer contracts, and a Security Questionnaire Responses.xlsx.
You connect Google Drive once (Settings, then Apps, pick Google Drive, authorize). Now in a normal chat:
"What's our standard discount cap for multi-year deals, and which contracts have exceptions? Cite the docs."
ChatGPT will:
- Search your connected Drive for relevant files (the pricing doc, the contracts folder).
- Retrieve the matching passages.
- Write an answer like: "Standard cap is 15% for 2 or 3 year terms (Pricing Policy 2026, section 4). Two contracts exceed it: Acme (20%) and Globex (18%)."
- Show citations linking back to the source files.
The citations are the point. You can click through and check. An answer with no traceable source is a liability. A cited answer from your own contract is something you can act on.
What gets accessed (and what does not)
Professionals often get this part wrong, so be precise.
Permissions are inherited, not granted. When you connect Google Drive, ChatGPT acts as you. It can only see files your Google account can already see. If a folder is shared with you, it is reachable. If not, the connector cannot bring it into view. This matters most in Business and Enterprise workspaces where document permissions are tightly scoped.
Scope can be narrowed. Admins can restrict which apps are available in a workspace, and some connectors let you limit a connection to specific folders rather than a whole drive.
Indexed content lives in OpenAI's systems for retrieval. For synced connectors, content is processed and stored so it can be searched. Read your plan's data terms. On Business, Enterprise, and Edu, business data is not used to train OpenAI's models by default. The current commitments are at openai.com/enterprise-privacy.
Disconnecting removes access. Revoke the connection in ChatGPT (or in Google's security settings) and the bridge is gone.
A quick mental rule: a connector never elevates your access. It only exposes what you already have to a model that reads fast.
Connectors vs. Actions vs. file uploads
You have three ways to get external data into ChatGPT, and they are not interchangeable.
| Mechanism | Best for | Who controls it |
|---|---|---|
| File upload | One-off analysis of a specific file | You, per chat |
| Connectors (apps) | Standing access to a tool's content (Drive, SharePoint) | You and your admin |
| Actions (in Custom GPTs) | Calling an APIAPIApplication Programming Interface: a standardised interface that lets applications communicate and exchange data without knowing each other's internal workings.View full definition → to do something or fetch live data | The GPT builder |
The distinction that matters: connectors read your knowledge sources; Actions call APIs. A connector answers "what does our policy say?" An Action handles "create the Jira ticket" or "look up this order's live status." We covered Actions in the previous lesson. Connectors are the read-heavy counterpart.
Using Connectors in ChatGPT
Connectors inside Projects, and where Custom GPTs fit
Connectors get more useful when scoped to a workflow.
In a Project, connected sources work alongside project files and custom instructions. Every chat in that Project answers against the same grounded context. Picture a "Q4 RFP Responses" Project with your Drive connected. Each drafted answer pulls from the real, current source material, and nobody re-uploads anything. Shared Projects on Business and Enterprise let a whole team use the same setup.
Custom GPTs work differently. A GPT gets its knowledge from files the builder uploads, plus any Actions the builder configures. It does not use your personal connectors. If you want an internal "Sales Answer Bot" GPT that reads live from SharePoint, the builder wires that up as an Action against the SharePoint API. If uploaded knowledge files are enough, they become a fixed snapshot that someone has to refresh by hand.
The trade-off: a Project with connectors stays current as files change and respects each user's own permissions. A GPT standardizes behavior for everyone, but its knowledge is only as fresh as its uploads or Actions.
The deep research and agent angle
Connectors also feed ChatGPT's more autonomous modes.
When you run deep research or ChatGPT agent, connected sources become part of the toolset. One research run can pull from your internal Drive and the web together. For example: "Compare our current pricing policy against the three competitor pages I'll paste, and flag where we're undercut." It reads your internal doc through the connector and the competitor pages through browsing, then reasons across both.
Pair this with scheduled tasks for a recurring grounded briefing, such as a Monday summary of changes in a tracked Drive folder.
Knowledge check
1. What is the fundamental purpose of a Connector in ChatGPT?
2. The lesson calls Connectors 'RAG, but you are not building the pipeline.' What does this mean in practice?
3. Why doesn't ChatGPT simply dump your entire Drive into the context window when you ask a question?
4. Select ALL correct statements about how a Connector operates.
Select all the correct answers.
5. In the 'answer from internal docs' example, why is the citation step described as 'the point'? Select ALL that apply.
Select all the correct answers.
When you outgrow the ChatGPT UI: building it yourself
Connectors are the no-code path. When you need this behavior inside your own application, you move to the OpenAI API. The equivalent capability there is file search over a vector storevector storeA vector database stores data as high-dimensional numeric vectors (embeddings) and retrieves items by similarity rather than exact matches, powering semantic search and AI applications.View full definition → using the Responses API.
The pattern: upload your documents into a vector store once, then attach that store as a tool. The model retrieves and cites automatically, as a connector does in the UI, but now inside your product.
from openai import OpenAI
client = OpenAI()
# 1. Create a vector store and add your internal docs.
store = client.vector_stores.create(name="internal-policy-docs")
client.vector_stores.files.upload_and_poll(
vector_store_id=store.id,
file=open("Pricing_Policy_2026.pdf", "rb"),
)
# 2. Ask a grounded question; the model searches the store and cites sources.
response = client.responses.create(
model="gpt-5",
input="What's our discount cap for multi-year deals?",
tools=[{"type": "file_search", "vector_store_ids": [store.id]}],
)
print(response.output_text)The file_search tool is the same idea as a connector: retrieval is hosted, you point at the data. The full reference is at platform.openai.com/docs. Check the models page for the current recommended model name before shipping.
The cost is responsibility. In the API, you enforce who can query which vector store. No inherited Google permission protects a document. If a user can hit your endpoint, they can search what you attached. In your own app you must build the access control the ChatGPT connector gave you for free.
Practical limits and gotchas
Know these before you rely on connectors in production:
- Freshness has a lag. Synced connectors re-index on a schedule, so a file edited two minutes ago may not show up yet. For data that must be live, an Action hitting the source API is a better fit.
- Format matters. Clean text documents (Google Docs, well-structured PDFs) retrieve far better than scanned images or dense slide decks.
- Citations are your QA tool. Ask for them on high-stakes answers and click them. The model can still phrase a summary loosely even when the source is correct.
- On Enterprise and Business, an admin decides which apps exist at all. If Google Drive is missing from your Apps list, that is policy, not a bug. Ask your workspace admin.
- Scope tightly. Connecting all of Drive makes retrieval noisier than connecting one curated folder.
A quick decision guide
- One file, one question, today: upload it.
- A body of knowledge you'll query repeatedly: a connector, ideally inside a Project.
- A team that needs the same grounded setup: a shared Project with connectors, or a Custom GPT with knowledge files and Actions if you need standardized behavior.
- Live data or taking an action in another system: an Action.
- The same grounded behavior inside your own software: the Responses API with
file_search.
Key Takeaways
- Connectors ground ChatGPT in your own tools (Drive, SharePoint, GitHub) and answer with clickable citations you can verify. Since December 2025 the ChatGPT interface calls them apps.
- A connector inherits your permissions and never elevates them. ChatGPT sees only what your account already can, and narrower scope gives sharper retrieval.
- Connectors read knowledge; Actions call APIs.
- Projects are where connectors shine for recurring work. Custom GPTs rely on uploaded knowledge and Actions instead.
- When you outgrow the UI, the API equivalent is `file_search` over a vector store via the Responses API. You gain control and lose inherited permissions, so you build access control yourself.
Frequently asked questions
What does a ChatGPT connector actually do?
A connector is an authorized bridge between your ChatGPT account and an external source like Google Drive, SharePoint, OneDrive or GitHub. You ask a question in plain language, ChatGPT searches the connected source, retrieves the relevant passages and answers with citations back to the original files. It never dumps your whole drive into the context window: it searches, then reasons over the matching chunks.
Why are connectors now called apps in ChatGPT?
OpenAI renamed connectors to apps in December 2025, so Drive, SharePoint and GitHub now appear under Settings then Apps, alongside interactive apps built with the Apps SDK such as Canva or Figma. Many people and older documentation still say connectors. The mechanics are unchanged.
Can a ChatGPT connector see files I'm not allowed to open in Google Drive?
No. A connector inherits your permissions and never elevates them: ChatGPT acts as you and reaches only the files your Google account can already see. A folder shared with you is reachable, a folder that is not stays invisible. Admins can also restrict which apps exist in a workspace, and some connectors let you limit the connection to specific folders.
What's the difference between a connector and an Action in a Custom GPT?
Connectors read your knowledge sources, Actions call APIs. A connector answers "what does our pricing policy say?" by searching Drive or SharePoint; an Action does things like creating a Jira ticket or fetching an order's live status. Custom GPTs also don't use your personal connectors: a GPT relies on files the builder uploads plus the Actions they configure.
When should I upload a file instead of connecting Google Drive to ChatGPT?
Upload the file when it's one document and one question today. Use a connector when you have a body of knowledge you'll query repeatedly, ideally inside a Project so every chat answers against the same grounded context. For live data or taking an action in another system, an Action hitting the source API fits better.
Why doesn't my ChatGPT connector see a document I just edited?
Synced connectors re-index on a schedule, so a file edited minutes ago may not appear in search results yet. If the data must be live, use an Action calling the source API instead. Two other common causes of weak results: scanned images and dense slide decks retrieve far worse than clean text, and connecting an entire drive rather than one curated folder makes retrieval noisier.
What to do, from this lesson
These actions are compiled in the role's Playbook.
- Run one Project per client, scoping files and instructions to it
Related articles
Recent articles from the blog that build on this lesson.