How to Connect YouTube Analytics to ChatGPT: Setup and Tested Prompts

To connect YouTube Analytics to ChatGPT, connect your channel in Windsor.ai, then add the Windsor.ai plugin in ChatGPT and ask about your channel in a new chat. You need no code, and the free forever plan covers one YouTube channel.

Below are the setup steps, the questions we asked ChatGPT about our own channel in October 2026 with the answers checked against the connector and YouTube Studio, and the YouTube data ChatGPT cannot get through this connection.

Before you start

• A Windsor.ai account. The free forever plan covers one data source, YouTube included, with no credit card, and reads the last 30 days of data. For longer periods, like the 3 months in our example below, or to keep GA4 or ad accounts next to YouTube, paid plans start at $19 a month for 3 sources, billed annually (pricing).
• The right Google account. Sign in with the channel owner’s account, or as an owner or manager of the Brand Account if the channel still uses one. People invited through YouTube Studio’s channel permissions (Manager, Editor or Viewer) cannot connect the channel, because YouTube does not give them API access.
• A ChatGPT account. We ran every test on this page on ChatGPT’s Free plan. OpenAI says plugin availability depends on your plan, region, workspace and the ChatGPT interface you use.

Step 1: Connect your YouTube channel in Windsor.ai

1. Open onboard.windsor.ai/app/youtube and sign in to Windsor.ai, or create a free account.
2. Sign in with Google and allow Windsor.ai to read your YouTube Analytics.
3. Select the channel you want to analyze. If you select more than one, ChatGPT can query all of them in the same chat.

Step 2: Add the Windsor.ai plugin in ChatGPT

Open the Windsor.ai plugin page in ChatGPT and click Connect (or Install plugin, if that is the button you see). Sign in to Windsor.ai and approve the access when ChatGPT asks. Once it is connected, the page shows Manage and Try in chat.

ChatGPT now lists apps under Plugins, so older guides that say “Apps” mean the same thing. To disconnect Windsor.ai later, open Settings, then Plugins.

Step 3: Use Windsor.ai in a chat

In a new chat, click + and choose Windsor.ai, or type @Windsor and pick Windsor.ai from the Plugins list. A Windsor.ai label appears on your message, and ChatGPT answers from your Windsor.ai data. ChatGPT may ask you to approve a request before it reads the data.

Step 4: Check the connection

Ask “What YouTube channels are connected to my Windsor account? Show total views for 1 to 28 September 2026.” On our account, ChatGPT listed both connected channels, gave 33,406 views for the Windsor.ai channel and said the second channel returned no data for those dates. 33,406 is the same total we got from the Windsor.ai connector directly.

Can ChatGPT watch a YouTube video?

Not through this connection. Windsor.ai gives ChatGPT your channel’s numbers and each video’s title, description, tags and length, not the picture or sound. When we pasted a link to one of our videos into ChatGPT and asked what happens in the first minute, ChatGPT said it could not retrieve the video and asked us to upload it, a screen recording or screenshots. To give ChatGPT what a video says, open it on YouTube, click Show transcript in the description, then copy the text into the chat.

Worked example: our channel, 3 July to 30 September 2026

We asked ChatGPT two questions about the Windsor.ai YouTube channel, then pulled the same numbers from the connector ourselves.

First question, asked on 9 October 2026: “Compare Shorts with long-form videos from 3 July to 30 September 2026: views, engaged views, minutes watched per view, and subscribers gained per 1,000 views.” ChatGPT split the videos by YouTube’s own Shorts label, not by length, and every number matched our pull:

	Shorts	Long-form
Views	23,910	55,599
Engaged views	14,672	39,707
Minutes watched per view	0.43	0.67
Subscribers gained	37	125
Subscribers per 1,000 views	1.55	2.25

Together that is 79,509 views. YouTube Studio’s Advanced mode showed 79,542 for the same dates, 33 more. Engaged views are the fairer comparison for Shorts, because since 31 March 2025 YouTube counts every Shorts play and replay as a view, while engaged views keep the earlier way of counting.

Second question, asked on 11 October 2026: “Which 10 videos gained the most subscribers per 1,000 views from 3 July to 30 September 2026? Show views, subscribers gained and the video title.” ChatGPT ranked every video on the channel. All 10 in its list had fewer than 300 views; the first gained 2 subscribers from 19 views, which is 105 per 1,000. ChatGPT added its own warning that small view counts produce unusually high rates, so the list shows promising videos rather than proven ones. Our own pull from the connector gave the same 10 videos with the same values.

The first answer shows that long-form videos bring our channel more subscribers per 1,000 views than Shorts. The second is worth reading with the view counts next to it: at a few dozen views, one or two new subscribers are enough to top the list.

What ChatGPT cannot get from the YouTube connector

When we asked ChatGPT for videos with a high click-through rate and a below-average view percentage, it said the click-through rate was not in the data and did not make up a list. The connector does not include these (checked 9 October 2026):

• Thumbnail impressions and impressions click-through rate
• The audience retention curve (where viewers stop watching). Average view duration and average percentage viewed are included.
• Device type
• Traffic source detail, such as the search terms people used. The traffic source itself (YouTube search, suggested videos, external and so on) is included.
• Revenue, for monetized channels too

YouTube Studio shows all of these. The full list of what the connector includes is on the YouTube data fields page.

Questions to ask next

• Which 10 videos gained the most subscribers in the last 90 days? Show views and publish date for each.
• Compare views and engaged views for my Shorts in the last 30 days.
• Break down last month’s views by traffic source, with minutes watched per view for each source.
• Show total watch time by week for the last 6 months. Which weeks rose or fell the most, and what did I publish those weeks?
• For each video published in the last 60 days, show subscribers gained, lost and net.
• List my videos published in the last 6 months by average percentage viewed, from lowest to highest.

With GA4 connected as well, you can ask whether weeks with more YouTube views also brought more website sessions from YouTube. GA4 cannot tell which video sent a visit unless the links in your video descriptions carry UTM tags. More examples across sources are in our guide to Windsor MCP use cases.

If something does not work

• The channel is missing in Windsor.ai: check which Google account you signed in with. Accounts invited through YouTube Studio’s channel permissions cannot connect a channel; use the owner’s account or a Brand Account owner or manager.
• ChatGPT answers without your YouTube data: choose Windsor.ai under + or type @Windsor in that chat, then ask again.
• Windsor.ai shows “Disabled by admin”: your ChatGPT workspace administrator has turned it off. Ask them to enable the plugin.
• Numbers differ from YouTube Studio: these are the differences we found on our channel.
  • YouTube says Studio data can take a few days to appear, and totals for recent dates keep growing.
  • Studio’s Average view duration is not watch time divided by views. For our YouTube search traffic from 3 July to 30 September, Studio shows 0:48, while its own watch time divided by its views gives 29 seconds.
  • Studio notes that views are counted differently from 27 August 2026. Compare periods that are both before or both after that date.

The same connection works in Claude. See how to connect YouTube Analytics to Claude. If your company uses Gemini Enterprise, Windsor.ai can be added there as a data store: how to connect YouTube Analytics to Gemini Enterprise.