Most of the time when AI features are shoehorned into software and services these days, it gets in the way more than being of any real use.

However, sometimes, an AI feature is added to a service we’re already using that actually makes it better. In this case, many YouTube users now have access to an “ask” button powered by Google Gemini.

I’ve been using it recently, and it’s probably one of the best features for users YouTube has added in ages. It doesn’t make up for losing the dislike count, but otherwise I quite like it for use with tutorial content specifically.

Why this YouTube “Ask” feature caught my attention

I had questions

The “ask” feature has been available on YouTube for a while, or at least so I discovered. I literally hadn’t caught wind of it until my wife called me over to show it to me. Basically, if you click on this button, it brings up a chat box with Google Gemini, and Gemini already knows the content of the video.

It allows you to ask questions about the video and have them answered without having to watch the whole thing. I think this is useful for YouTube tutorials in particular, because most of them tend to be pretty inefficient. They have padded intros and lots of rambling or irrelevant information, which can be annoying when you just want to know how to do a particular thing.

How the “Ask” button actually works in practice

It unlocks a lot more than you’d expect

If you have access to the Ask Gemini button (not all regions and languages do) then you’ll see it as a button in the browser player or in the YouTube app, as seen in these two examples.

When you use the button, Gemini will give you a few suggested questions, such as getting a summary, but you are free to ask any questions you want about the content. That includes asking about things that are not actually in the video, since of course the knowledge contained in the Gemini model is present and correct.

It’s a genuine time-saver for tutorials

Or at least a good companion

In tutorial videos, when I ask “give me the steps outlined in the video” I usually get a bulleted list of steps, but most importantly they include clickable time stamps. After all, the whole reason to watch a tutorial video is not only to know the steps, but to see how something is done so you can learn vicariously.

My wife and I want to build a patio pond, so I used the ask button on a patio pond tutorial video and this is what the result looks like.

I don’t want to use this feature to get around watching the video, I want to use it to guide me to the correct places within the video and help me get the most out of the content.

You can try out the Gemini ask button right now in the video below, though you’ll have to open it on YouTube to see it. For now, the button does not appear in the embedded player, as you can see.

Where it still falls short

All the usual AI pitfalls apply

While this is a great feature for me as a YouTube viewer most of the time, just like most AI tools using LLM technology, there are serious caveats.

First, it can only work with the content that’s in the video. So, if the tutorial is inaccurate or incomplete, then a summary of that content will be similarly flawed. I’ve tried asking Gemini whether the content in a video has any factual issues and it tries to give me a coherent answer, but this has the same issues as relying on AI to fact-check things for you in general.

Subscription with ads

No, all ad-free

Live TV

No

For $14 a month, you get ad-free videos, YouTube Music, and offline video downloads.

Which brings up the second problem: hallucinations. LLMs can make up details that don’t exist in the original content. If you never actually watched the video in question, then you have no way of knowing whether this is the case or not. So it makes sense to leverage those timestamps.

I also worry that the chatbot might decide that certain subtle bits of knowledge or warnings in the video aren’t important enough to make it into the summary. These details are small, but if the person who made the tutorial is an expert, they’re usually included for a reason.

Google has the usual disclaimer about how AI can make mistakes at the bottom of the chat box, but in this specific scenario I suggest that you actually take it seriously.

As a YouTuber, I’m a little worried

I’ve been a YouTuber for half a decade at this point, hosting a tech channel and more niche and experimental content. Like all creators, I actually want people to watch my videos. I was already pushed by the evolving YouTube algorithm over the years to become ruthlessly efficient. If you read the comments under my most popular tech videos, you’ll see people thanking me for getting to the point quickly or having as little fluff as possible.

My fear is that, like Google’s AI summaries in its search engine, this Gemini Ask feature will simply stop people from watching the content. It’s not clear yet how this affects monetization for creators, but if it has a negative effect on it then we’ll see less content. With less content, in the end, what will the ask button even have to summarize?