Skip to content
Marketing Working integration

Connecting YouTube to AI: summarising the comments and comparing videos without opening YouTube Studio

YouTube can be connected to AI. The assistant retrieves the videos of a channel, the views, likes and comment counts of a video, and the comments underneath. Uploading, publishing and replying are not part of it: this connector only looks on.

A channel with dozens of videos creates two kinds of work nobody enjoys. The first is comparing: which video did well, which one fell behind, and does that have to do with the subject or with the length. The second is the comments. They hold the most useful things a viewer will ever write, recurring questions, misunderstandings, wishes, but they sit scattered across hundreds of separate lines and nobody reads them all. The YouTube connector takes on both. The assistant retrieves the videos of the channel for a period, requests the figures of the ones that matter and reads the comments underneath. That last part is where the gain sits: it summarises the gist, pulls out the recurring questions and says how often something came up. One thing belongs in the picture: the figures the connector gets back run slightly behind what YouTube Studio shows. For the question of which video worked better that makes no difference, for a report accurate down to the individual view it does.

What the assistant looks up on the channel

The videos of a channel

The assistant retrieves the videos of a channel, narrowed to a period or a search term and ordered by date or popularity. You start with a manageable list instead of an entire archive.

The figures of a video

For a specific video the views, the likes and the number of comments can be requested. That puts two videos side by side and shows what the subject or the format did.

The comments under a video

The comments can be retrieved, sorted and searched on a word. The assistant summarises them, groups the recurring questions and shows which remark comes up more often than you thought.

Nothing posted, nothing answered

Uploading, changing titles and replying to comments are not included. The channel stays with its administrator; the assistant only surfaces what is already there.

Where the channel starts paying off

Healthcare

Situation
A care organisation publishes explanatory videos and suspects viewers keep asking the same things.
What the agent does
The assistant reads the comments under the most viewed videos and groups the questions that recur.
Result
A list of subjects for the next video and for the FAQ on the website, taken straight from the words of viewers.

Technical wholesale

Situation
Assembly instruction videos are online and the service desk still gets the same phone calls.
What the agent does
The assistant searches the comments for words tied to a step in the assembly and picks out the sticking points.
Result
It becomes clear where the instruction falls short, so one extra video saves a pile of phone calls.

Training and education

Situation
A training provider wants to know which subjects are watched most before the new yearly plan is drawn up.
What the agent does
The assistant retrieves the videos of the past year, requests the views and groups them by theme.
Result
A plan that follows where the interest demonstrably is, rather than what the team most enjoyed making.

Tourism and leisure

Situation
After the season there has to be a review of the videos that were made.
What the agent does
The assistant lists the videos of the season with their figures and summarises the comments per video.
Result
A review that says not only how many people watched, but also what they made of it.

Connecting the channel

  1. 1

    Connecting through the platform

    An administrator connects the Google account of the channel in IntraGPT. Access can be withdrawn at any moment, after which the connector retrieves nothing.

  2. 2

    Deciding who gets access

    You set which assistants and which roles may query the channel. Viewer comments are public, but that does not mean everyone in the organisation needs them summarised.

  3. 3

    Using the assistant

    After that someone asks in the chat which video did best or what is being said under a video. The assistant names the videos it looked at, so you can check for yourself.

  4. 4

    Traceable what was requested

    Every question to the channel lands in the log with its result. That makes it possible to trace where a summary or a number in a report came from.

Connected securely

Public comments are still things people wrote

Everything this connector retrieves stands publicly on YouTube, but it was written by viewers sharing an opinion or a complaint. That deserves care, and the connector cannot say anything back in any case.

Where the data lives: Nederland of de EU

No way back to the channel

The connector only looks on. There is no way for the assistant to post a video, change a title or comment under a video, not even on request.

Access stays yours

Access to the channel is not visible to users and is managed by an administrator. It can be withdrawn at Google itself and takes effect immediately.

Roles decide who can ask

You decide which assistants may query the channel. An assistant for customer service can summarise the comments without everyone in the organisation being able to ask the same thing.

Processing on our own server

The comments and figures are processed on our own server in the Netherlands with open-weight models. They do not go to an external model provider and nothing is trained on them.

Questions about the YouTube connector

Can I connect YouTube to ChatGPT?

There is no built in connector in the public chat services, and access to the Google account of the channel does not belong there. In IntraGPT, YouTube sits inside the platform as a connector, with permissions per role and a log, so it is visible who requested what.

Does IntraGPT work with YouTube?

Yes. YouTube is in the connector catalogue. The assistant can retrieve the videos of a channel, request the views, likes and comment counts of a video, and read and summarise the comments under a video.

Can the assistant upload videos or reply to comments?

No. This connector cannot post and cannot answer. For a marketing team that is an advantage: the comments under a video can be summarised daily without anything ever being said on behalf of the channel. Replying stays with whoever runs it.

Why do the figures differ from what I see in YouTube Studio?

The figures the connector gets back run slightly behind the display in YouTube Studio. For the question of which video worked better that is no problem, because the ratio holds. For a report that has to be accurate down to the individual view, take the number from Studio.

Can the assistant look at other channels too?

It can request public videos and their public figures from another channel, a peer in the sector for instance. What it cannot see are the figures only a channel owner holds. So do not expect insight into how long a competitor's viewers kept watching.

Are watch time and audience data included?

No. The connector delivers the public measures: views, likes and the number of comments. Watch time, audience composition and traffic sources are analytics that stay in YouTube Studio. If you want them in a report you add them from there by hand.

What does a YouTube connector cost?

Connecting is part of the project we run together. The work depends on the number of channels, whether summarising comments becomes a fixed recurring report and which roles get access.

How long does setup take?

Connecting itself is a short job and the first question can be asked the same day. The time goes into agreeing which questions recur, for instance a monthly summary of the comments under each new video.

For which sectors

Other connectors

See all connectors

Curious what this connector would deliver for you?

Book a free thirty minute AI call. We look at your systems, the permissions around them and the first use case that saves time or money.

Book a free AI consultation