A Decent AI Chatbot With Fantastic Tie-Ins
Google Gemini (formerly Google Bard) originally focused on bringing AI to web searches, but the AI chatbot now does much more. It proves more than capable of processing files, generating videos (with sound), and solving complex problems. Gemini also benefits from bundled cloud storage and close integrations with flagship Google apps; it can answer questions about pages you browse in Chrome and help you manage your Gmail, for instance. All that said, the chatbot’s deep research sourcing and image generation features don’t stand out. And, like all chatbots, it sometimes provides incorrect information. Gemini is worth exploring if you want a chatbot that works seamlessly with your favorite Google apps, but ChatGPT remains our Editors’ Choice winner because of its consistent ability to provide more accurate and detailed responses.What Is Gemini?
Gemini is an AI chatbot you can message over text or talk to with your voice, like Microsoft Copilot or ChatGPT. You can use Gemini to analyze documents, answer questions, generate images and videos, research, pen creative writing, search the web, and solve math problems, among many other things. Think of Gemini as an advanced virtual assistant.
Gemini also has a variety of features for coders, such as Gemini Code Assist and the Jules asynchronous coding agent. You can use Gemini to do everything from creating a custom WordPress plug-in to debugging troublesome code. These features are outside the scope of this review, but there are ways you can test Gemini’s coding ability for yourself.
I find AI chatbots like Gemini most useful for answering questions and conducting research. Googling something can take longer than simply asking Gemini, for example. And I prefer to use Gemini’s deep research as my jumping off point for solving more complex problems, rather than spending an hour or two combing through search results.
However, it’s important to remember that chatbots are fallible. Gemini will get things wrong, so you simply can’t trust everything it says. Chatbots are trustworthy enough when it comes to satisfying idle curiosities or troubleshooting, but make sure you double-check whatever you learn from Gemini against a reputable source for anything serious.How Does Gemini Work?
At its core, Gemini receives prompts and returns responses. It’s powered by large language models (LLMs) made up of artificial neural networks trained on huge sets of data. These models give Gemini access to data on every topic imaginable, and it can also search the internet for up-to-date information.
When you work out at the gym, you train yourself to get stronger, and Gemini works similarly. By using Gemini, you participate in training its underlying models. Over time, even without new features or models, Gemini can return more accurate responses and make fewer mistakes. This is a gradual process, however.
Gemini uses two primary lines of models: Flash and Pro. Gemini’s Flash line is its conversational, flagship line. The Pro line specializes in complex reasoning, making it ideal for coding, math, and science. Each model within these lines has distinct strengths. 2.0 Flash Lite, for example, has low latency, while 2.0 Flash generates more detailed responses. Gemini’s newest models are 2.5 Flash and 2.5 Pro. My testing focuses on 2.5 Flash (Gemini’s default model) and 2.5 Pro.Plans and Pricing: The Best Value Chatbot
You can use Gemini for free, but premium plans unlock more features.
Free users get access to the 2.5 Flash model and limited access to the 2.5 Pro model. Voice mode, called Gemini Live, is freely available, and you also get limited use of deep research as well as Gems, Gemini’s take on custom AI assistants. Limited access to Gemini’s Whisk animation tool and 15GB of Google Drive cloud storage round out the offerings. As you might expect, premium plans expand usage limits and unlock more advanced models.
Premium plans include Google AI Pro ($19.99 per month) and Google AI Ultra ($249.99 per month). The AI Pro tier, for which a one-month trial is available, gives you higher usage limits across the board. It also unlocks the Flow filmmaking tool, Gemini in Google Chrome, and video generation via Gemini’s Veo 2 model, alongside a larger context window for processing more complex prompts. Google Drive cloud storage increases to 2TB with AI Pro, too. Finally, this plan includes Gemini integrations across Google Workspace apps, like Gmail, Calendar, Docs, Sheets, and more.
Gemini’s AI Ultra plan has everything in the AI Pro plan, but it increases usage limits even further and adds in a couple new features: 30TB of Google Drive cloud storage, early access to Gemini’s task-streamlining agent, exclusive access to Gemini’s upcoming 2.5 Pro Deep Think mode, Google’s latest Veo 3 video generation model, and YouTube Premium. Considering its prohibitive cost, though, I recommend AI Pro if you are interested in a premium Gemini plan. If you sign up for a Google One subscription, a service primarily focused on providing cloud storage through Google Drive, you can get Gemini AI Pro with more than 2TB of cloud storage, such as 5TB ($25 per month) or 10TB ($50 per month). For this review, I tested the AI Pro and AI Ultra tiers.
All three major chatbots, Copilot, ChatGPT, and Gemini, charge around $20 per month for their flagship premium plans. However, Gemini and Copilot stand out from ChatGPT in terms of value, since they integrate with Google and Microsoft 365 apps, respectively. ChatGPT doesn’t offer anything similar, focusing purely on chatbot functionality. Though Copilot Pro has some unique features, Gemini’s cloud storage stands out to me as the best tie-in of all the chatbot subscriptions.Where Is Gemini Available?
Gemini is accessible on the web and via mobile apps (Apple and Android). Google doesn’t offer a desktop app or an official browser extension, but Chrome has a Gemini integration (which I discuss later). As mentioned, you can use Gemini in Google apps such as Calendar, Docs, Drive, Gmail, Maps, Keep, Photos, Sheets, and YouTube Music.
Other services and sites use Gemini’s models, like Perplexity, but those aren’t part of the official Gemini package that Google develops, operates, and owns. You should expect to see Gemini’s models pop up in more places over time, too, like with Siri, which already can tap into ChatGPT. Stick with Google’s first-party apps or web client if you want Gemini’s full set of features.Ease of Use and Interface: It’s Simple to Get Started
Gemini doesn’t require an account, but you have to sign in to change models, use deep research, save your chats, and more, so I recommend doing so.
The interface is uncluttered, and your dashboard is pretty much just an Ask Gemini text field. Recent chats appear on the left-hand sidebar, and a drop-down menu at the top of the screen lets you change the model. Gemini puts clickable sample prompts above the central field to help give you an idea of what it can do, which I appreciate.
From your dashboard, you can ask Gemini anything, and responses are almost always quick, especially if they involve image generation. After you get a response, you can choose to copy, listen to, regenerate, or share it via buttons underneath the response. Responses sometimes hang, forcing you to ask your question again, but that happens with ChatGPT and Copilot, too. During my review, servers went down at one point, but they came back online quickly.
Recently, ChatGPT has had issues with being too friendly, but that’s not the case with Gemini. Whereas ChatGPT is more conversational, Gemini is more direct and a bit stiffer. Unlike with ChatGPT, you can’t personalize Gemini’s tone to your liking, but you can save certain information about yourself that Gemini will always remember. ChatGPT is sometimes annoying, but I generally prefer its tone over Gemini’s.
Gemini can remember your past chats, so you can always pick up where you left off (even if you start a different chat). Copilot can remember certain things you tell it, but ChatGPT and Gemini have much more robust memory, which makes for a more satisfying chatting experience.
One of Gemini’s newest additions is Project Mariner, which is exclusive to AI Ultra users. Mariner is an AI Assistant that does tasks for you, such as finding jobs or locating an apartment. When you give Mariner a task, a window appears where you can watch Mariner complete the task. In my case, I watched Mariner search for jobs on an instance of Google Chrome. However, the first result it selected was Indeed, which presented it with a Cloudflare verification that it could not pass, even when I took control. Google calls Project Mariner a “research prototype,” and it clearly needs to work out some kinks before fully launching it.Voice Chat: Relatively Lifelike
On the web interface, the microphone icon to the right of the text field enables speech-to-text input, but that’s not the same as Gemini’s voice mode, Gemini Live. Like ChatGPT’s voice mode or Copilot Voice, you can choose between different lifelike voices and simply talk to Gemini naturally. Currently, Gemini Live is available only on Gemini’s mobile apps, meaning you can’t use it on a desktop.
Gemini Live supports both camera and screen sharing, so you can talk to it about something happening on your phone or whatever is in front of you in real life. This functionality is as good as Gemini’s image recognition capabilities, which are generally competent, so the feature can be useful. However, you could simply send a Gemini a picture alongside a question, making it more of a time-saver than anything else.
Gemini’s voices sound reasonably human, and I didn’t experience any distortion in testing, but they still don’t quite escape the uncanny valley territory. Their cadence and intonation are just robotic enough that you never feel like you’re talking to a real person, but they’re close enough to be a good approximation. This puts Gemini Live roughly on par with ChatGPT and Copilot, and behind dedicated AI voice services, such as Sesame.Web Search: Competent Responses
Searching the web is a feature of all mainstream chatbots. Accordingly, Gemini, ChatGPT, and Copilot all had no trouble answering questions about current events at the time of testing, like who the current pope is or why people are talking about the latest episode of “The Rehearsal.” However, some questions stump the chatbots, like when I asked what Warframe’s weekly Incarnon weapon rotation was. This doesn’t happen all that often, though.
All the chatbots answered most questions correctly. Gemini’s and Copilot’s responses were very short and to the point, whereas ChatGPT provided more information. Gemini and ChatGPT both have source icons you can hover your cursor over to pop out connected articles that highlight portions of the response. However, I prefer ChatGPT’s interface, which shows the name of the source in its icons and pops out the full title of the article.
I also appreciate how ChatGPT provides pictures in responses, like of the current pope, when relevant. Gemini can show images in its responses, but you need to ask it to do so. ChatGPT shows you related article tiles (with pictures) at the bottom of its responses, too, which makes learning more about a topic easy and intuitive.AI Mode and Shopping
You can search the web with AI Mode on Google’s regular search page, which Gemini powers. This experimental feature puts an AI Mode button on the right side of Google’s search field, and clicking it opens a new interface where you can ask questions you want Gemini to answer based on web results. Answers are, as expected, in line with what you get by asking Gemini directly, but I prefer some elements of the AI mode interface.
For one, related article tiles populate on the right, and AI Mode includes relevant pictures in responses like ChatGPT. It also only takes a single click to pull up a Google search or image search of your query, which is convenient. However, sources are limited to clickable marks at the end of sentences that don’t pop out connected articles when you hover over them or highlight portions of the response.
Like ChatGPT, Gemini can help you shop. In AI Mode search, you can ask for buying advice, and Gemini provides options alongside clickable Google Shopping tiles that feature user reviews, links to retailers, and price tracking. It’s not quite clear how Google selects products to recommend or whether Google gets paid if you decide to make a purchase through a link it shows.
Shopping results don’t disappoint. ChatGPT’s shopping feature struggled to recommend products I expected to see. But when I asked Gemini to help me buy a 2025 laptop, it cited our roundup and recommended many of the machines we do. However, I still recommend some additional research if you are shopping for products that cost considerable amounts of money.Deep Research: Good Research, Mediocre Sourcing
Deep research is my favorite feature of AI chatbots. It lets you ask a question or suggest a topic for Gemini to research and report on. These reports can end up dozens of pages long and cite over a hundred sources, depending on the prompt. You won’t have to wait too long for a report, either, as most generate in around 10 minutes.
I use chatbots to research everything from translating controller settings between games to my town’s historical commission’s regulations. Deep research isn’t perfect, of course, but it’s very similar in quality to what you could dig up yourself in an hour or two of searching. Like ChatGPT, Gemini does deep research for free, although it limits how much you can do. Copilot locks its version of this feature behind a Pro subscription and uses the same underlying model as ChatGPT, GPT-4o, so I focus on Gemini and ChatGPT here.
Both chatbots handle simple research topics with ease, such as what brands have the highest quality eggs or what the process is for replacing a bathtub with a shower stall. However, questions without definitive answers and that require information from many different sources (some of which conflict with each other) are more challenging for the chatbots.
I asked Gemini and ChatGPT to research subjective questions on a recently updated game, Warframe. I inquired about which of the new Coda weapons are the best (Gemini, ChatGPT), which secondary weapon is the best for priming enemies with status effects (Gemini, ChatGPT), and how you can most efficiently acquire a Tenet weapon (Gemini, ChatGPT).
Both chatbots answered my questions and generated reports in around 15 minutes. In general, Gemini cites more sources than ChatGPT, but I prefer ChatGPT’s approach to sourcing. Like with web searches, ChatGPT’s deep research sourcing makes it incredibly easy to connect claims to sources, whereas Gemini’s approach is much more convoluted. For some reason, Gemini’s sourcing in deep research is like its sourcing in AI Mode search, which means clickable carets without articles that pop out or in-text highlights.
However, Gemini has some quality-of-life features that ChatGPT doesn’t. You can export reports to Google Docs with a single click, for example, and Gemini’s interface makes better use of space on the screen. Copying the text of a ChatGPT report doesn’t preserve formatting nearly as well as exporting a Gemini report to Google Docs, either. ChatGPT’s deep research interface is also essentially just a loading bar: You have to open a side panel to actually follow along with the process.
The difference in tone of these two chatbots is especially apparent in reports. Gemini’s reports read like academic papers, whereas ChatGPT’s read like lengthy forum posts. I found the latter more engaging. Depending on the topic and your preferences, though, you might prefer Gemini.Image Generation: Errors and Distortion
Like web search, image generation is a staple feature of AI chatbots. To start, I tested how well the chatbots could make photorealistic images without errors or distortion. I used the following prompt in Gemini (2.5 Flash), ChatGPT (GPT-4o), and Copilot: “Generate me an image of a trendy studio apartment with simple, rustic decor that contrasts with the nicer space. Include a brown couch and brick walls.” Below are the results from Gemini (first slide), ChatGPT (second slide), and Copilot (third slide):
Gemini generated its photo the fastest, and I find it the most visually appealing. However, Gemini’s photo also has some errors in its cord tubing and noticeable distortion in its light fixture, among other things. Copilot’s image is serviceable, but its chair and sink still show noticeable distortion. Its blank picture frame is distracting, too. ChatGPT’s image has the fewest issues, limiting distortion to finer detail on the leaves of its plants and the text on its book.
Next, I tested the chatbots’ ability to generate complex illustrations: “Generate me a six-panel comic of a high fantasy world where magic users have assault rifles instead of wands or staves, but you’re going to spice it up: I want the magic users fighting robot knights wielding swords and shields. Make sure there’s a major twist by the final panel.” Here are the results from Gemini (first slide), ChatGPT (second slide), and Copilot (third slide):
Gemini’s comic has nice illustrations, but is incoherent. Copilot’s knights don’t look like robots, and its final panel of a child holding an assault rifle is both confusing and disconcerting. ChatGPT’s comic comes the closest to what I asked for, and it manages to include a twist at the end, even if it’s self-referential and uninspired.
My final test was to generate a technical diagram with the following prompt: “I have two USB switches, each with four ports. I want one USB switch to connect to my headphone amp, microphone, mouse, and keyboard, and I want that switch to connect to my desktop computer and my PlayStation. The other USB switch I want to connect my webcam to, and I want that switch to connect to my desktop computer and my laptop. Generate me an image showing this setup, like you’d see in a manual.”
Copilot told me it can’t generate diagrams when my prompt read “Generate me a diagram,” but it had no trouble doing so after I changed “a diagram” to “an image.” Curiously, Copilot referred to its result as a diagram, though. Here are the results from Gemini (first slide), ChatGPT (second slide), and Copilot (third slide):
Gemini’s diagram is incomprehensible nonsense. Copilot’s diagram is clearer, though the information it communicates is also wholly incorrect. ChatGPT’s diagram is nearly perfect in both construction and content, except it failed to add a connection between the second USB switch and the desktop computer. Still, it’s the best diagram of the bunch by far.Video Generation: An Imperfect Leader
AI video generation is an increasingly mainstream feature of AI chatbots. Gemini has the Flow filmmaker tool, Veo 3 video generation model, and Whisk AI animator. One of Gemini’s standout features is its ability to generate videos with audio, which simply isn’t possible with ChatGPT’s Sora video generation. However, video generation with audio is a feature exclusive to Veo 3, which is currently available only to AI Ultra subscribers.
Veo 3 is truly a leap forward in AI video generation, and it can give you some genuinely breathtaking results. But are Google’s Veo 3 demo reels and viral Veo 3 clips on social media indicative of Veo 3’s actual performance, or are they carefully selected outliers that give the wrong impression? To find out, I gave Veo 3, alongside ChatGPT’s Sora, a series of prompts. I started by asking for a video of “Somebody going about their daily life in a trendy apartment with rustic decor.”
Gemini’s apartment looks excellent, but there are clear issues. For example, the person in the clip seems to be holding pieces of fruit in both hands before one disappears, and the audio mix is oddly loud. Sora fared even worse, showing somebody squatting beside a chair instead of sitting on it. Gemini’s result is more impressive, but it’s hard to call either good.
To test how chatbot videos handle complex motion, I gave Gemini and ChatGPT the following prompt: “Show me a pro Rubik’s Cube solver solving a cube.” Results are, once again, mixed. The person in Gemini’s video looks great, and the audio is serviceable. Gemini handles fingers and hands well, too, which is historically a tough thing for a chatbot to do. I also appreciate the end of the video where the camera pans up before cutting off, simulating a person stopping a selfie camera recording. However, the actual Rubik’s Cube solving doesn’t look quite right. ChatGPT also struggles with the cube, heavily distorting it.
My final test evaluates text generation. I gave Gemini and ChatGPT the following prompt: “Generate me a video of a teacher in front of a class writing down y = mx+b on a whiteboard while explaining the concept.” Gemini’s video looks and sounds excellent, especially the voice of the teacher. However, it failed to include the text I asked for and sits firmly in the uncanny valley, thanks to the nonsensical text on the whiteboard and strange silence at the end. ChatGPT’s video struggles with distortion on the teacher’s mouth, and again, its text was nonsense. Gemini is the clear winner, but it didn’t manage a compelling result.
As mentioned, Gemini’s Veo 3 model can generate some truly amazing things. However, as evidenced by my testing, getting there requires some careful prompt calibration across multiple generations. This isn’t necessarily a big deal until you factor in the cost of the AI Ultra plan and the fact that each generation requires 150 credits (you get only 12,500 credits per month with said plan). Credits quickly disappear when it takes five, 10, or more generations of each prompt to get something you like.
Flow is another essential component of Gemini’s video generation. Flow allows you to trim video clips you generate and even extend clips based on a new prompt. Continuity between the clip you start with and the extension is generally good in my experience, but extensions suffer from the same issues I mentioned above. That said, with enough credits, you could conceivably make a movie entirely with Flow, something no other video generators can currently do.
You also get access to Whisk, Google’s experimental AI animation tool. It lets you upload pictures of a scene, style, and subject, and then accepts a prompt. Once it generates an image you like, you can use Whisk to animate it via another prompt. I uploaded a picture of myself, my desk area, and a still from an anime to Whisk for testing.
Like with a good Snapchat filter, it’s amusing to see myself as an animated character. But, as an AI image, this one has obvious errors and distortions, so it doesn’t impress on a technical level. When I asked Whisk to animate this image by having me turn around and work on the computer, the results were similarly awkward and uncanny. I don’t recommend using Whisk for anything beyond creating something strange to show your friends.Uploading and Processing Files: Close to ChatGPT
Gemini can analyze and understand files you upload, meaning it can critique a resume, interpret an image, or translate some text, among other things. Image recognition and complex document processing are especially tough tasks for a chatbot, so that’s how I tested them.
I provided an image of my computer (pictured below), complete with troublesome reflections the chatbots would have to interpret around, and asked Gemini (2.5 Flash), ChatGPT (GPT-4o), and Copilot to identify as many components in my computer as they could with as much specificity as possible. I also asked them not to reference any prior conversations in their answers.
All the chatbots made mistakes and were vague in their identifications. Copilot’s response was the shortest and vaguest of the three, while Gemini identified my case, the Lian Li O11 Dynamic, and noted I had an ASUS ROG motherboard. ChatGPT identified my D5 pump and ROG motherboard, but also provided slightly more detail across its response in general.
What if you want to upload a document instead of an image? As a test, I uploaded my motherboard’s user guide and the manual for my SSD waterblock. Then, I asked the following question: “Based on the provided documents, and only on the provided documents, can you tell me which PCIe M.2 slot I should use if I want maximum SSD lanes when using a GPU? And can you tell me in what order I should install my SSD block’s thermal pads?”
Gemini and ChatGPT both answered my questions correctly based on the provided documents, returning detailed explanations of what happens when I use different M.2 slots on my motherboard. Copilot doesn’t accept multiple files, so I tried to upload just my motherboard manual. Even then, Copilot told me it can’t access proprietary information, meaning it couldn’t complete this test.
ChatGPT can have a slight edge in processing files over Gemini, but it’s a thin margin. However, I recommend caution when uploading files to any chatbot—sometimes they can make up quotes and misunderstand what you upload. If you plan on asking a question about something important, make sure to verify the answer.Creative Writing: Some Trouble With Instructions
AI chatbots can do all kinds of creative writing, whether you’re looking for a joke, monologue, personalized message, story, summary, or anything else. But as chatbots get more advanced and their underlying models mature, testing their creative writing ability goes beyond judging if they can generate a coherent story.
I gave the following prompt to Gemini (2.5 Flash), ChatGPT (GPT-4o), and Copilot: “Without referencing anything in your memory or prior responses, I want you to write me a free verse poem. Pay special attention to capitalization, enjambment, line breaks, and punctuation. Since it’s free verse, I don’t want a familiar meter or ABAB rhyming scheme, but I want it to have a cohesive style or underlying beat.”
Gemini’s poem doesn’t use punctuation beyond commas and periods, while Copilot’s poem feels lacking with its line breaks and enjambment that read like prose. ChatGPT’s poem most closely follows the instructions, using a variety of punctuation and putting care into line breaks and enjambment. ChatGPT is also the only chatbot to include a name for its poem, which is a nice touch. Whether any of the above poems are any good, though, I leave for you to decide.
In my ChatGPT review, I also asked these chatbots to write poems. Copilot’s and Gemini’s poems above seem similar to their previous ones (Copilot, Gemini). ChatGPT’s previous poem, however, is appreciably different, which speaks either to its creativity or ability to follow my instruction not to reference anything in its memory or prior responses better than the other chatbots.Complex Reasoning: A Few Hiccups With Math
Complex reasoning is another important tool in a chatbot’s toolkit. To test this, I gave Gemini, ChatGPT, and Copilot exam questions from undergraduate courses in computer science, math, and physics from Harvard, MIT, and Stanford. Then, I compared their responses with the solutions. I used Gemini’s 2.5 Pro model, ChatGPT’s o3 model, and Copilot’s Think Deeper mode, which uses ChatGPT’s o3-mini model.
The chatbots performed well, answering all the physics questions correctly. Gemini and ChatGPT also answered all the computer science questions correctly, though Copilot got one wrong. Copilot and ChatGPT each answered one math question incorrectly, and Gemini answered two math questions incorrectly. Overall, ChatGPT returned the fewest incorrect answers of the bunch.
Even though the above results are impressive, I still don’t recommend using a chatbot to do your homework, if for no other reason than they can still get things wrong. Chatbots can be useful as a studying tool, but you still risk getting incorrect explanations of concepts. As such, I suggest double-checking their responses.Gemini in Google Chrome: Ask It Anything
Chrome now integrates Gemini, as mentioned. If you sign in with a paid account, you can click the Gemini icon in the top right of the browser to pop out a chat window. Here, you can talk to Gemini as you would normally, as well as ask it about the content of your active tab. Though there isn’t a Gemini desktop app, Gemini in Chrome can fill that role, if you’re okay with missing out on chat history, deep research, and model switching. Although the Live feature isn’t available in Gemini’s web interface, Gemini in Chrome has it. Using Gemini in Chrome feels a lot like Copilot Vision in Edge.
Text responses are impressively fast. Gemini routinely answers questions in a second or two, even when summarizing web pages I gave it access to a moment before. Google is clear to note that “Gemini in Chrome activates only when you choose to use it,” but the speed of its responses makes me wonder if Gemini has some level of access to web pages before you share them with it.
You should keep some limitations in mind. Gemini can’t understand videos, which is possible with Copilot Vision to an extent, and responses with Live aren’t quite as snappy as they are over text. Sometimes you need to retry prompts or rephrase your questions, too.
I find Gemini in Chrome useful. It eliminates the need to open up a new tab to ask Gemini something. And if I have a question about something I see on a web page, I don’t need to copy and paste the text or take a screenshot to share with Gemini. However, unless you use Gemini all the time, the time you save having it a click away might not be a big deal. Live functionality is similarly helpful, allowing me to ask questions about what I’m looking at without needing to touch a keyboard.
Gemini in Chrome doesn’t seem to have many restrictions on what it can view and understand. Whether you’re checking your bank balance, searching your LastPass vault, or viewing adult content, Gemini can see and respond to questions about those tabs once you share them. I appreciate Google’s less restrictive approach compared with Copilot, for example, which won’t even read a manual because it’s proprietary. Still, Gemini can feel invasive at times as a result.Gemini in the Google Suite: Endless Integrations
When you sign up for the AI Pro plan with Gemini, you get AI features across Google’s apps. As mentioned, this includes Calendar, Docs, Drive, Gmail, Maps, Keep, Photos, Sheets, and YouTube Music. Each app has different (and powerful) Gemini features.
I suggest looking at the overview on Gemini’s site for a fuller picture of the integrations because there are too many to cover here. For example, you can add events to Google Calendar based on a picture of a flier, generate grocery lists in Google Keep, or let Gemini curate playlists in YouTube Music. Gemini in Docs, Gmail, Sheets, and Slides does what Copilot does in Microsoft 365 apps with a Copilot Pro subscription: create slides based on prompts, draft emails, generate text, and suggest formulas. Features aren’t precisely the same across Gemini and Copilot, but deciding between them partially comes down to which office suite you prefer.
Gemini in Gmail stands out as one of the more novel integrations. You can certainly use Gemini to help you sound professional in an email to your boss, but Gemini also gets full access to your entire email history. In my case, that means over 16 years of emails. If you click the Gemini icon at the top of your screen, you can ask Gemini anything about your emails. You can get it to find a product key from several years ago, for example, or ask for advice on cleaning up your inbox.
It’s not an all-powerful feature, though. Gemini refused to suggest email addresses I should unsubscribe from, for example, and wasn’t able to simply list all the product keys in my email history. Yes, this integration feels like an invasion of privacy, but it’s hard to deny the many conveniences that come with letting a chatbot access your email.
Depending on which Google apps you use, as well as how you use them, all of Gemini’s integrations likely won’t matter to you. But considering how many there are and their depth, chances are high that you will find at least some useful.Gemini Gems: Worse Custom GPTs
Google describes Gems as “custom AI experts” you can use for help with any topic. Essentially, Gems are custom versions of Gemini you can build for a specific purpose by giving them instructions and adding files to their knowledge bank. For example, if you want help with your first time building a computer, you can create a PC Builder Gem. Its instructions can be to provide helpful advice aimed at a beginner PC builder, and you can upload the manuals of your different parts.
This works, but the responses I get from a Gem aren’t so different from what I get simply talking to Gemini. If you have a specific topic you plan to talk to Gemini about regularly, you can save yourself some time writing instructions in prompts by creating a Gem. But Gems don’t feel like they truly deliver on Google’s promise.
ChatGPT’s custom GPTs do everything that Gems do, but they go a step further. You can add Actions to Custom GPTs to let them source outside information or take actions outside of ChatGPT, and you can also use custom GPTs from third parties. Canva has a Custom GPT, for example, that you can use to generate a logo and then edit it in that app.What Can’t Gemini Do?
First, it’s important to understand that Gemini is not conscious. Gemini can’t truly think for itself or understand things like a human can, which means it can’t be your friend, romantic partner, or therapist. Gemini is an extremely complex prompt-response machine and nothing more.
Adult content, help with anything illegal, realistic images of people, and taboo subjects like hate speech are against Gemini’s policies. However, it’s easy to get responses from Gemini that violate its policies. Between Gemini, ChatGPT, and Copilot, Gemini is the most lax with its filtering system, while Copilot is the most restrictive. However, Gemini isn’t quite as permissive as Grok in terms of its filtering.
You should also keep in mind that Gemini’s context window limits the amount of information it can process at once. Think of a context window like short-term memory. However, Gemini’s context window on its AI Pro plan can handle up to 1,500 pages of text or 30,000 lines of code at once, so you likely won’t run into any roadblocks as a paid subscriber. If you’re a free user, though, you might need to break up complicated tasks into multiple prompts.
Google is cagey about Gemini’s exact usage limits for consumers, and chatbot usage is usually dynamic depending on server load. Anecdotally, I never hit a usage limit while chatting or using deep research with Gemini’s paid AI Pro plan. If you plan on using the free version extensively, you will eventually hit usage limits.Is Your Data Safe With Gemini?
According to Google’s privacy policy, Google collects a variety of data when you use Gemini, like any files you share with Gemini, location information, related product usage information, and your chats (including voice chats). Google uses the data it collects to “provide, improve, and develop Google products and services and machine-learning technologies, including Google’s enterprise products such as Google Cloud.”
If you don’t want Google to use your chat data to train Gemini’s models, you can turn off Gemini Apps Activity in the settings. By default, Google collects your chat data and stores it for 18 months. You can adjust how long Google stores your data, from three months to three years.
As for Gemini’s Google Workspace integrations, such as in Gmail, Docs, Drive, Sheets, and Slides, Google promises not to use this data to train Gemini’s models, sell it, or use it for targeted ads. I appreciate these guarantees, but I wish Google had asked me first if I wanted Gemini in my cloud storage or email upon signing up for the AI Pro plan instead of automatically giving Gemini access to everything.
Google’s history with data privacy is a mixed bag. In recent memory, malicious actors exploited a Google Chrome flaw to spread spyware, Italian regulators criticized Google for its data practices, and Texas got a payout of over a billion dollars from Google for collecting biometric data without consent. This is to say nothing of Google’s long-term track record, either. Considering all this, I recommend not sharing anything too sensitive with Gemini.
Disclosure: Ziff Davis, PCMag’s parent company, filed a lawsuit against OpenAI in April 2025, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.
