Adamya Sharma / Android Authority

Google recently made some major changes to its One AI plans. While some were more explicit, like a new, cheaper Ultra tier, others came to the surface only after some users ran into issues. As it turned out, after the Google I/O announcement and the introduction of the visible usage limit tracker, Google AI Pro plan users were affected by increasingly capped limits that would run out with even basic use in many cases.

While Google didn’t deny tightening the usage limits further, it did assure the users most affected by this new limit. So, I decided to test it out for myself and see what crushes Gemini’s usage limit the most and what you can do to circumvent it.

What went into the testing

Karandeep Singh / Android Authority

The world of AI assistants is quite dynamic, which is particularly true for something like Gemini, which looks to get an update every single week. Just so you know, Google now displays your usage limits in the hamburger menu, available both on mobile and the web. There you will find your quota presented in a percentage bar that resets every five hours, as it keeps adding to your weekly limit.

I made sure to account for all such changes introduced in the recent weeks, including the new Gemini 3.5 Flash model. To get a clear picture, I used both the latest Flash and Pro models to see which one hits the limit the fastest. Secondly, I tested Gemini with a variety of usage types, from file analysis to code and video generation — and the findings were the opposite of what I had predicted.

What barely dented the limit

I assumed that media generation would be the biggest culprit, but I was quickly proved wrong when I tried to generate a bunch of images. At first, I thought it was an anomaly, but after multiple trials, I was sure that Gemini’s usage limit (when I mention usage limit, I mean the five-hourly one) was going up by just 1% with each image generation. Even the most complex image I could ask it to create upped the number by just 2% across both Flash and Pro models. That’s negligible if you want to subscribe to the AI Pro plan to generate images for your work, especially considering the limit resets every few hours.

1% per image creation is negligible, especially considering the limit resets every few hours.

Another shocker came in the form of coding. It isn’t something that I know too much about, so I had to take AI’s help to give Gemini proper prompts to test it to its fullest. I asked it to build a complete habit-tracking app for Android from scratch, complete with onboarding, analytics charts, widgets, Material 3 design, and tablet optimization. I then asked it to add Wear OS integration, cloud sync using Firebase, and multi-device support. To push it further, I had Gemini review its own code for performance issues, memory leaks, and vulnerabilities before transitioning the entire project from Firebase to Supabase. I even asked it to create the entire pipeline from planning and design to reviewing and deploying, while explaining what it was doing and why.

To my surprise, all of that work just consumed 4% of the usage limit, and the results were similar across both Gemini models. Of course, I am not the best judge of code quality, but this is, anyway, some solid efficiency on Google’s part.

And to no one’s surprise, regular chat, even with a lot of back and forth that doesn’t involve too much reasoning, would take you ages to hit the ceiling. You don’t have to worry about it at all.

What turned out to be a heavier load than I imagined

I consider PDFs and bare-bones CSV files pretty lightweight, but apparently Gemini doesn’t. When I gave Gemini a 250-page PDF report to analyze, the limit surged by 7% on Flash, while the same file on Pro resulted in a relatively lower 5% bump. The important thing is that each follow-up question led to just a 1% increase on Flash vs 5% on Pro. Seemingly with the Gemini 3.5 Pro model, AI is analyzing the file a lot more deeply, with much more reasoning work in the background to give a better, more comprehensive answer — which was my other observation besides keeping track of the quota filling up.

Similarly, with a CSV file, Gemini consumes 5% of its usage, with a surprising part somewhere else. In a Gemini notebook, I have multiple chats that contain a handful of CSV files with 2,000 to 3,000 rows each. When working under that particular Gemini notebook, Gemini was literally gobbling up its usage quota. I imagined CSV files and a spreadsheet attached from Google Drive to be rather light on the usage limit compared to something like image generation, but querying inside the notebook over 10-15 follow-ups, with each needing querying and refreshing the data, resulted in Gemini consuming 61% of my usage in that very session alone. Which is why I had to wait for the limit to reset before starting to test everything else.

My observations suggest that analysis, reasoning, and context awareness use up much more resources than text or image generation.

What surprised me most is that this happened much faster than code generation, image generation, or even PDF analysis combined (in a separate session).

While this is not the official word from Google, my observations suggest that analysis, reasoning, and context awareness use up much more resources than text or image generation, though it’s still nuanced. Even though generating code also requires a lot of reasoning, it was still light on usage limits compared to going through the files you supply and responding to you based on that. This is perhaps the most meaningful contrast here because it could mean the difference between staying within your 5-hourly limit and consuming your entire weekly quota in a day.

What’s the biggest culprit

Mishaal Rahman / Android Authority

Video generation.

There are no two ways about it. If you prompt Gemini to create a video from scratch, it will hit your usage limits the hardest. In my experience, the usage scrubber went up by 20 to 25% for each 10-second video generated, either with the Flash or Pro model.

That means you can’t generate more than four short clips every five hours using the AI Pro plan — and that’s when you know exactly what you want the video to look like, and Gemini gets it right on the first attempt. You’ll be able to create even fewer videos if you need to give follow-up prompts to fix anything up. Something to be considered if you are planning to get the AI Pro tier for video production.

What you should really care about

Karandeep Singh / Android Authority

These usage quotas may appear arbitrary, but to end users like you and me, these are perishable commodities that you should consume. Otherwise, they would expire without rolling over. However, with the limit resetting within a reasonable five-hour window, it’s perfectly workable — especially considering it takes a lot of effort to hit 100%. I mean, with all this testing, I could only push my limit to 86% (except for my session inside that Gemini notebook), after accounting for all code, image, and video generation. And within a few minutes, it was back down to zero.

What you actually need to worry about is the weekly limit. This session of mine consumed 5% of my weekly limit. It may not sound huge, but that’s from my work spanning merely 3 to 4 hours. Now, if I were a professional who relies on Gemini for their project management, development, and execution, the weekly limit would become a bigger bottleneck. But there are ways you can avoid it.

What you can do to make your quota last longer

Joe Maring / Android Authority

What emerged from all this testing was a fairly clear pattern. Gemini rarely cared when I asked it to generate text, code, or even images. The usage meter started moving only when the AI had to process large amounts of information or generate entirely new media. PDFs, spreadsheets, and videos consumed more quota than regular chatbot tasks, which wasn’t what I expected when I started testing.

The most straightforward workaround is in plain sight, though. The Google AI Pro plan is by default shared with up to five members in your family, and the good thing you should know is that the usage limits aren’t a shared pool, unlike Drive storage space, so each user gets their own. I found it tough to hit the limit even for the five-hour period, but if you manage to do it, you can always switch to another Google account in your family, and you will have the full quota at your disposal.

The Google AI Pro plan is shared with your family, and the good thing is that each member gets their own Gemini usage limits.

I do understand that it won’t be possible for everyone, especially when you are in the middle of a task or already have all the context and files tied to a particular account. In that case, my suggestion would be to always start with the Flash model, which is not only quick but also quite capable while using much fewer credits, before moving to the Pro model for deeper research work. That way, you can extend your usage within the five-hour window without hitting your weekly limit.

As for video generation, my suggestion would be to be thorough with your scripting and storyboarding process so that the final video generation doesn’t require too much back and forth and saves you on your credits. You can even use image generation before jumping to the final video, since images are much lighter on your quota. And if you plan on creating videos in a larger quantity, it would be wiser to go for the AI Ultra plan — even the new cheaper $100 tier would give you much more headroom than AI Pro.

Even if you are a heavy user, you can make do with the limits by using your Gemini quota smartly

There is a good chance that Google may have tweaked the Pro limits after the widespread backlash, but I still found them difficult to hit. Looking at all the testing, the limits aren’t nearly as restrictive as they first appear — provided you understand what actually consumes them. I do think the AI Pro plan is still pretty generous with its Gemini usage limits, especially when you account for all the other perks that Google has been including lately. Even if you are a heavy user, you can make do with the limits by using your Gemini quota smartly.