Connect HubSpot to ChatGPT: What Marketers Actually Get (and What a CRM Connector Will Never Know)
If you run marketing at a Series A or B company, there is a good chance you have already connected HubSpot to ChatGPT. HubSpot said in June 2025 that over three quarters of its customers were already using ChatGPT, and the connector was the obvious next click. Then you asked it to draft the nurture sequence, and it gave you text to paste.
That is not a failure of the connector. It is what a CRM connector is. This post lays out the three different ways HubSpot reaches ChatGPT as of September 2026, what each one reads and writes, why HubSpot’s own pages describe them differently, and the four things about your marketing that no CRM connector can know. Then it shows the setup that actually removes the tab-switching: two connectors in one chat, with HubSpot answering the who questions and a marketing-native connector answering the what and when.
Three ways HubSpot reaches ChatGPT
People say “the HubSpot ChatGPT connector” as if it were one thing. There are three surfaces, and they do different jobs.
| Surface | Where you use it | What it reads | What it writes | Source |
|---|---|---|---|---|
| HubSpot connector for ChatGPT | ChatGPT Business and Enterprise, added under Connectors | CRM records: contacts, companies, deals, tickets and their properties; ask questions in plain language | Since early 2026, creates and updates CRM records in small batches; drafts copy as text in the chat | HubSpot knowledge base |
| HubSpot as a deep research source | ChatGPT deep research | CRM data pulled into a long-form research report alongside web sources | Nothing | Same knowledge base article |
| HubSpot remote MCP server | Any MCP client: ChatGPT with Developer mode, Claude, Cursor and others | CRM objects plus campaign analytics | Creates, updates and clones marketing email drafts, alongside CRM writes | HubSpot developer docs |
The first row is what most marketing teams have. It is good at exactly what it says: “which contacts from the webinar have not booked a demo,” “summarize the deals that stalled in proposal this month,” “show me companies over 200 employees with no owner.” Those are who questions, and a CRM is the right system to answer them.
The third row is newer and more capable. HubSpot’s remote MCP server went generally available in April 2026, and its tool set includes managing marketing email drafts and reading campaign analytics. It is the surface to watch, and it is still a CRM-first tool: it does not build workflows and it does not send.
Why HubSpot’s own pages disagree
Read HubSpot’s material side by side and you will find two descriptions of the same connector. The knowledge base article says ChatGPT will draft campaign copy for manual implementation rather than automatically building emails or workflows. The marketing page for the OpenAI connector talks about creating campaigns and building landing pages.
Both are true from a certain angle. The knowledge base is describing the tool’s capabilities: it returns text. The marketing page is describing what you, a person, can then do with that text. The gap between the two sentences is your Monday afternoon: the connector produces the words, and someone carries them into the email editor, the workflow builder and the landing page tool by hand.
Practitioners have noticed the edges too. A March 2026 thread on OpenAI’s developer forum reports that the Business-plan connector does not provide usable access to HubSpot custom objects, which matters if your qualification data lives in one. None of this is a knock on HubSpot. It is a description of what a CRM connector is for.
What the connector does well
Before the gap, the credit. Connected to ChatGPT, HubSpot is genuinely useful for the part of marketing that lives in the CRM:
- Segment questions in plain language. Who attended but did not book, which accounts opened three emails and clicked none, which deals in proposal have gone quiet for two weeks. This used to take a saved view and a spreadsheet.
- Handoff hygiene. Summarize a deal’s history before a call, draft a note for the owner, update a property after the conversation. The small-batch record writes added in 2026 make this real work rather than reading only.
- Reporting narratives. Deep research over CRM data plus the web produces a decent first draft of the board slide, with the caveat that the numbers deserve a check.
- Copy from context. Ask for a re-engagement email for the stalled-deals cohort and you get a competent draft that knows who it is for.
If your only question is “who,” the connector already answers it. The problem starts one step later.
What a CRM connector will never know
HubSpot holds contacts, companies, deals and the activity attached to them. It does not hold, in any form its connectors expose, the four things a marketing assistant needs before it can do more than write text.
What your brand sounds like. The voice guide, the ICP as you describe it, the current offer and the words you do not use. HubSpot has a place for some of this, but the connector returns records, not guidance. ChatGPT is left with whatever you pasted into the chat, which decays the way brand voice drifts in long sessions.
What is going out this week. The calendar of scheduled sends, the LinkedIn post on Thursday, the webinar reminder on the 16th. A CRM records that an email was sent; it is not the place that holds the plan for the emails that have not been sent yet.
Which automation a contact is inside, and what step is next. HubSpot workflows exist, but the connector does not expose “this contact is on step two of the pilot-offer follow-up and step three fires Friday.” So the re-engagement email ChatGPT drafts for the stalled cohort may land on top of a follow-up that is already scheduled.
What is waiting for approval. There is no draft queue in a CRM connector. A draft is a block of text in a chat until a person moves it somewhere. Nothing tracks that it exists.
These are not missing features that HubSpot could add next quarter. They are the difference between a system of record for people and a system of record for marketing. The longer version of that argument is in HubSpot will not run your campaigns and in the comparison of HubSpot workflows and AI agents.
Run both connectors in one chat
The setup that removes the tab-switching is not “replace HubSpot.” It is two connectors in the same conversation, each answering the questions its system holds. Marqeable syncs audiences from HubSpot, Salesforce and ServiceTitan, and its connector holds the marketing half: business context (brand, ICP, voice, offers), campaigns and their pieces, the content calendar, every automation and its plan, and a draft queue that the assistant can write into but never send from.
Here is a session with both connected. The HubSpot connector is answering the first question; Marqeable’s tools are answering the rest.
You: Which contacts attended the pricing webinar in August, have a deal in proposal, and have not had an email from us in 14 days?
ChatGPT: From HubSpot: 23 contacts match. 19 are at companies over 100 employees; 4 deals are owned by reps who have not logged activity in two weeks. Here is the list.
You: Before we email them, what are they already getting from us?
ChatGPT: From Marqeable’s automation overview: 11 of the 23 are inside the proposal-stage follow-up automation, which sends a check-in on day 7 and a case study on day 14. The other 12 are in no automation. The content calendar shows a nurture email to the full list on the 4th and the webinar reminder on the 16th, so a new send on the 5th or the 17th would collide.
You: Draft a three-step re-engagement automation for the 12 with no coverage. Offer the pilot. Send window inside quiet hours, first step on the 8th.
ChatGPT: I pulled the current offer and voice from your business context and drafted a three-step automation: a short note from the deal owner on the 8th, a customer story on the 12th, and a direct pilot offer on the 19th, all inside the 7am to 9pm window. It is saved as a draft, not live. Review it here: [link to the automation draft in Marqeable].
Three tool calls did the work no CRM connector can do: reading the automation membership, reading the calendar, and writing the result as a draft into the system that will run it. The last step is yours. Nothing in Marqeable’s connector can send, launch or publish, which is the property to insist on before you connect any assistant to anything that emails customers; the reasoning is in the draft-only rule.
Keep HubSpot as the system of record for people and let it answer the who questions. Add a marketing-native connector for what is scheduled, what is running and what is waiting for approval. Neither connector replaces the other, and the same chat can hold both.
How to connect Marqeable to ChatGPT
The server is https://app.marqeable.com/mcp, a remote MCP server over HTTP with OAuth 2.1. You sign in with your normal Marqeable login in a browser window, and ChatGPT sees exactly what your account can see. No API keys are involved.
As of September 2026, in ChatGPT on a plan that supports custom connectors: open Settings > Connectors, and under Advanced turn on Developer mode. Click Create, name the connector “Marqeable”, paste the address, choose OAuth and save. Sign in to Marqeable when prompted and enable the connector from the tools menu in a conversation, next to HubSpot. OpenAI has moved this setting more than once this year, so if the path differs, search Settings for “Developer mode”. On Business plans an admin may need to allowlist the connector, and ChatGPT may ask you to confirm write actions per conversation.
A good first prompt with both connectors on: “Run whoami, then get_business_context, then tell me what is on the calendar this month, then pull the contacts in proposal stage from HubSpot.” If the answers match what you see in each tool, you are set. If you use Claude instead, the equivalent walkthrough is the HubSpot connector for Claude and what to connect alongside it.
Frequently asked questions
Can ChatGPT build HubSpot workflows or send marketing emails through the connector?
Not through the ChatGPT connector. HubSpot’s knowledge base says ChatGPT drafts campaign copy for manual implementation rather than building emails or workflows. HubSpot’s separate MCP server, released in April 2026, can create and update marketing email drafts, but it still does not build workflows or send.
What is the difference between the HubSpot ChatGPT connector and the HubSpot MCP server?
The connector is the one you add inside ChatGPT to chat with CRM data on a Business or Enterprise plan. The MCP server is HubSpot’s developer-facing endpoint that any MCP client can use, with a wider tool set that includes marketing email drafts and campaign analytics. Both are CRM-first; neither holds your brand voice, calendar or approval queue.
What can a CRM connector not tell ChatGPT?
What your brand sounds like, what is scheduled to go out this week, which automation a contact is inside and what step is next, and what drafts are waiting for approval. HubSpot does not hold those in a form its connectors expose, so ChatGPT needs a marketing-native connector alongside the CRM one.
Do I need to disconnect HubSpot to use a marketing connector?
No. Run both in the same chat. HubSpot answers the who questions about contacts, deals and lifecycle stages; a marketing connector like Marqeable answers what is scheduled and what is running, and takes drafts back for approval. Marqeable syncs audiences from HubSpot and complements it rather than replacing it.
The bottom line
Connecting HubSpot to ChatGPT is worth doing, and it does what a CRM connector should: it answers who questions in plain language and, on the newer MCP surface, drafts a marketing email. What it cannot do is know your voice, your calendar, your live automations or your approval queue, because HubSpot does not hold those in a form its connectors expose. The fix is not a better prompt or a different CRM. It is a second connector in the same chat, built for the marketing half, whose every write is a draft you approve.
See the marketing half live: Marqeable’s agents build and run your campaigns, run behavior-triggered automations across text and email, answer and qualify every visitor with AI website chat, and tie revenue to the exact message with attribution.
Marqeable runs your campaigns, answers every visitor, text, and email in seconds, and turns them into booked jobs and meetings - even at 9pm on a Saturday. We’re in private beta with a small early cohort. Get early access
