YouTube Analytics MCP: Ask an AI About Your Channel's Numbers

YouTube Studio shows you a lot of numbers. Making sense of them takes longer. Why did watch time fall last month? What do your best videos have in common? To answer that, you click through many reports and compare them yourself.

This guide shows how to ask an AI assistant those questions in plain language, using a standard called MCP. You'll learn what MCP is, what the assistant can and can't see, which questions are worth asking, and how to connect your channel with HubTub Pulse's free connector.

Why ask an AI about your YouTube Analytics?

YouTube Studio is good at showing numbers, one report at a time. Connecting the dots is the slow part. Finding out why watch time dropped, what your strongest videos share, or how this month compares with the same month last year means opening several reports and comparing them by hand.

An assistant that can read those reports does the legwork for you. It pulls the right figures, compares periods, and explains what it sees in plain words. You can ask follow-up questions, and you can ask in your own language. It doesn't replace your judgement, but it saves you the digging.

What MCP is, in plain words

MCP stands for Model Context Protocol. It is an open standard that lets an AI assistant connect to an outside service and read data from it through "tools". A tool is one specific action, such as "get the last 28 days of views for this channel". When you ask a question, the assistant picks the tools it needs, reads the data, and answers you.

A connector is the link between an assistant and one particular service. When you add a connector, you allow the assistant to use that service's tools. So a "YouTube Analytics MCP" is simply a connector that lets your assistant read your YouTube Analytics.

What the assistant can read through HubTub Pulse

Everything the assistant reads is YouTube's own figures, as YouTube Analytics reports them to the channel owner. Every figure belongs to one channel.

• Time periods: totals for the last 7 and 28 days beside the previous period, the last 90 days by day, and the last 12 months by month.
• Core numbers: views, watch time, subscribers gained and lost. Estimated revenue only where the owner granted the revenue permission.
• Audience and traffic: audience by country, age and gender; traffic sources; and a split between Shorts, long videos and live.
• Top videos: the 50 most-viewed videos of the last 28 days, with average view duration, percentage viewed, likes, comments, shares and subscribers gained.
• Audience retention: retention curves for the 10 most-viewed videos, fetched overnight once an assistant reads the channel.
• Thumbnails: impressions and click-through rate (the share of impressions that turned into views) by video and day, from YouTube's daily reports. These start about two days after the assistant first reads the channel, with the 30 days before.
• Channel standing: Community Guidelines, copyright and Content ID status. If the difference between a claim and a strike is unclear, see Content ID claim vs copyright strike.

The assistant can't see anything outside this list. It doesn't read comment texts, it can't see channels you haven't connected, and it doesn't get live data.

Questions worth asking your assistant

Ask the way you would ask a colleague. Naming the time period helps the assistant pick the right report. Some examples to start with:

• "Why did watch time drop on my channel? Look at the last 90 days by day."
• "What do my best videos of the last 28 days have in common? Suggest ten title ideas in the same spirit."
• "Compare my last 12 months: which months were strongest, and is there a seasonal pattern?"
• "Where do viewers stop watching my top video?"
• "Which videos get the most thumbnail impressions, and how does their click-through rate change day by day?"
• "Most of my views come from subscribers. What could help more of them come from recommendations?"
• For agencies: "Compare all channels in my team by revenue and watch time for the last 28 days."

If you manage several channels, the last question works across everything you've connected. For more on keeping many channels in view, read our guide to a YouTube Analytics dashboard for multiple channels.

How to set up the HubTub Pulse connector

1. Create a free HubTub account by signing in with Google, then add your YouTube channels to HubTub Pulse. Pulse asks Google for read-only access to YouTube Analytics.
2. In Claude, find "HubTub Pulse" in the connector directory and connect it. Or open Settings → Connectors, choose "Add custom connector" and enter the address app.hubtub.io/api/mcp. Other assistants that support MCP connectors with sign-in connect the same way, with the same address.
3. Sign in with Google when asked, check what the assistant will be able to read, and allow access.
4. Start a new conversation and ask about your channels.

Privacy and control: what happens to your data

• Read-only: the connector can't change anything on YouTube or in your account. It never asks YouTube for data itself; it reads the reports Pulse already keeps, which refresh once a day. YouTube's figures settle about two days late.
• You stay in control: you approve each assistant, and you can disconnect it at any time in Pulse. An assistant not used for 90 days is disconnected automatically.
• Shared channels: channels other people share with you are readable by your assistant only if their owner allowed AI assistants.

What to do next

Connect one channel first and start with a simple question, such as how the last 28 days compare with the period before. Once the answers make sense to you, move on to retention and thumbnail questions, and check anything important in YouTube Studio before you change your plans.