Since February, I began using Ask Gemini almost daily because it’s built into Google Chrome.
It has helped me filter content from long-form videos and read through a full essay response to questions I needed answered immediately.
While I refuse to take Ask Gemini’s word for it, I sometimes wonder how accurate it can be.
I decided to do a deeper dive into what Ask Gemini does for YouTube content, simply because I watch a lot of technical videos for both work and personal hobbies, and figured it would be easiest to cross-check its accuracy against its answers when I am already familiar with its sourcing.
I also use this feature on my computer, mainly on my ThinkPad and Chromebook.
Sadly, I don’t love using it on my Android phone, since I find the process clunkier and prefer Gemini Live instead. But you can use this feature through the Gemini app, if you prefer, via the Ask about your screen option.
I almost ignored the Gemini button in Chrome, but now it saves me hours every week
The Gemini Ask button is more useful than it looks
Ask Gemini for YouTube content
The good, the bad, and the ugly
I really liked using Ask Gemini to save time. As a writer, I have to read loads of web pages each day to keep up with the news, updates, and general findings in science and technology.
While I feel guilty that I can’t always watch an entire 15-minute video, I do when I can, but I also have to be choosy about which ones I watch.
That’s where I lean on Ask Gemini. I ask it key questions while I have a YouTube video open and see if a particular sub-topic is being discussed.
I generally do this for videos without timestamps, so I ask for the timestamp when the creator discusses specific topics.
For example, I recently asked about a Samsung OLED screen glitch issue. I wanted to know the cause and the devices it affected, so I used Ask Gemini to pinpoint where in the video the creator discusses the specific triggers and the affected devices.
So far, the timestamps that Ask Gemini has provided have been good. I haven’t had any problems and have been using Ask Gemini on my desktop for three months.
Though I have noticed one detail, and it actually occurred in the same video, it sometimes answers my question using context outside the video, which I disliked.
However, the AI tool provided sources, so when it did, I could tell right away that the answers it provided were not covered in the video. I ended up watching the entire video again to check, and this was mostly the case.
I have also re-evaluated how I use my prompts so that I wouldn’t just reproduce the same problems, as I really wanted to avoid the AI derailing or going off-topic, hallucinating, and not trusting its information.
Another observation I made during some tests was that it relied heavily on built-in subtitles to compute the video’s context.
Gemini can’t actually see the video, so if the chosen video includes creator-provided on-screen text or records a presentation, Gemini relies on the speaker for its data.
While this wasn’t an issue when I opened tech-based videos with common English terms, I discovered I had more trouble when I used this tool for my niche hobbies.
For example, for gaming, I love theory-crafting. Especially when I want to understand what makes a character strong and how it affects the game’s meta, I look to content creators and public testers for more analysis.
However, not all these terms are picked up well when some abilities use non-English terms or references that don’t translate well into readable captions, and I found Ask Gemini inconsistent at finding context-based answers, which led me to stop using it for this practice.
Using Ask Gemini has drawbacks
It can negatively affect your privacy
I like using Ask Gemini to filter out clickbait and prioritize articles about hot news topics.
For example, I am struggling to find reliable information from Valve about the Steam Machine, especially regarding pricing.
There are quite a few videos that are heavily speculative about the launch window and pricing, and some focus on rumors.
After a long day at work, my brain sometimes can’t tell the difference between what’s factual and what’s not.
That’s when I like using Ask Gemini to see if it can analyze the video when I ask it questions like, “Did the content creator say that this Valve’s pricing structure?” and “Are there any leaks being sourced, and what are they?”
I’m fine with using Ask Gemini as I am, but I have major privacy concerns.
For one, Gemini can now read active tabs. So anything I am browsing while using this feature is fair game for training the model, unless I turn it off.
The opt-in option is on by default. At the same time, I turned off Gemini in the app. I don’t know whether this policy extends to the built-in Ask Gemini in my Chrome browser (it should, but it is always best to double-check and manually clean your activity when you can).
That’s where you prefer to turn it off and avoid engaging with it when working with sensitive data, such as your email, or when browsing banking information or reading confidential documents.
It isn’t that hard to do if you are using it on your browser like me. You can turn it off in Chrome by selecting the three-dot icon in the upper-right corner and choosing Settings > AI Innovations.
Then, choose Gemini in Chrome and toggle everything off. As an added precaution, clear your history through the Gemini Apps Activity to remove any traces from your previous sessions.
If the Ask Gemini button shown at the top of your browser becomes too much of an eye-sore, right-click on it and select Unpin.
You can nuke the feature if you want, which is nice, unlike Google’s AI Overviews, which seem to be permanently embedded in Google Search unless you use custom workarounds to make them disappear.
Gemini is about to get even bigger
Google is about to full-send Android into Gemini with its leap toward agentic territory.
This means Google’s move toward Gemini Intelligence closes the gap between passive tools and smart assistance for your Google Workspace apps.
So, you can start using autonomous app automation and more proactive assistance across your apps.
Pretty much, you offload work to the AI so it can do more for you without being locked out of your various apps in the OS.
If you are interested, you can try it out on a new-gen Samsung or Google Pixel phone when that releases with Android 17/One UI 9.
