A Decent AI Chatbot With Fantastic Tie-Ins
Google Gemini originally focused on integrating AI into web searches, but the AI chatbot now offers far more. It is extremely capable in complex problem solving, deep research, file processing, image editing and generation, and video generation (with sound). Gemini also benefits from bundled cloud storage and close integrations with flagship Google apps. Its free version is capable enough that you likely won’t feel the need to spring for a paid plan, too. Gemini’s robust data collection can feel invasive, and, like all chatbots, it gets things wrong sometimes, but it’s nonetheless an Editors’ Choice winner, thanks to its seamless connections with your favorite Google apps and top-notch response quality. ChatGPT also remains an Editors’ Choice winner.What Is Gemini?
Gemini is an AI chatbot that you can message over text or talk to with your voice, similar to 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 offers a range of features for coders, including 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. I prefer to use Gemini’s deep research as my starting 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) composed of artificial neural networks trained on vast datasets. These models provide Gemini with access to data on every imaginable topic, 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 contribute to the training of 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. Gemini’s newest model is 3 Pro. My testing focuses on 2.5 Flash (Gemini’s default model) and 3 Pro.
For image generation, Gemini has Nano Banana. Nano Banana is a nickname for the 2.5 Flash Image model, while Nano Banana Pro is another name for the 3 Pro Image model. You can think of Nano Banana as an extension of Gemini’s mainline Flash and Pro models. I used Nano Banana Pro for the image generation tests in this review.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 3 Pro model. Voice chat, called Gemini Live, is freely available, and you also get limited use of deep research as well as Flow, Gemini’s filmmaking-focused video generation tool that uses Veo 3.1. Limited access to Gemini’s Whisk animation tool, Gemini’s NotebookLM research and studying tool, and 15GB of Google Drive cloud storage round out the offerings. As you might expect, premium plans expand usage limits.
The AI Pro tier ($19.99 per month), for which a one-month trial is available, gives you higher usage limits across the board. It also unlocks Gemini in Google Chrome and 3 Pro access in Google Search, 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 ($249.99 per month) has everything in the AI Pro plan, but it increases usage limits to the max and adds in a couple of new features: 30TB of Google Drive cloud storage, access to cutting-edge Gemini features (such as AI agents and Deep Think mode), 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.
It’s outside the scope of this review, but premium Gemini plans get you a variety of coding-focused features, such as the AI coding agent, Jules, and higher daily request limits in Gemini Code Assist. Think of Jules like a tool you can use to code for you, whereas Code Assist helps you code yourself. You can use Code Assist for free, just with lower request limits.
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 Search also has a range of Gemini-powered AI functionality, including AI overviews and Search’s AI mode. The Google app for Windows isn’t solely about Gemini, but you can use AI Mode when searching in the app to ask questions and find links. Chrome also has a Gemini integration (which I discuss later). As mentioned, you can use Gemini in Google apps, including 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 can also expect to see Gemini’s models appear in more places over time, such as with Siri, which already has the ability to 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 must sign in to change models, use deep research, save your chats, and more. 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 receive a response, you can choose to copy, listen to, regenerate, or share it using the buttons located underneath the response. Responses sometimes hang, forcing you to ask your question again, but this also happens with ChatGPT and Copilot. During my review, servers went down at one point, but they came back online quickly.
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 somewhat formal. 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 can be annoying at times, but I generally prefer its tone to 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.
Gemini also has Project Mariner, which is exclusive to AI Ultra users. Mariner is an AI agent that performs tasks on your behalf, such as finding jobs or locating an apartment. When you assign a task to Mariner, a window appears that allows you to observe Mariner completing the task by piloting a virtual web browser. 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. Mariner clearly needs to work out some kinks before it fully launches.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, allowing you to discuss something happening on your phone or what is in front of you in real life. This functionality is comparable to Gemini’s image recognition capabilities, which are generally competent, making the feature potentially useful. However, you could simply send a Gemini a picture alongside a question, making it a more time-saving approach 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 concise and to the point, whereas ChatGPT provided more detailed information. Gemini and ChatGPT both have source icons that you can hover your cursor over to display connected articles, highlighting portions of the response. However, I prefer ChatGPT’s interface, which displays the source name in its icons and displays the full article title when clicked.
I also appreciate how ChatGPT provides pictures in responses, such as those of the current pope, when relevant. Gemini can show images in its responses, but you need to ask it to do so. ChatGPT displays related article tiles (with images) at the bottom of its responses, too, making it easy and intuitive to learn more about a topic.AI Mode and Shopping
You can search the web with AI Mode on Google’s regular search page, which Gemini powers. This feature puts an AI Mode button on the right side of Google’s central 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: Generate Reports On Anything
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, but both limit how much you can do.
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 generate a variety of reports on the video game Where Winds Meet, explaining its mechanics, what to look for in a guild, where to find quests, and more. Then, I compared the results. Both were able to produce generally accurate research on a topic with limited documentation and many conflicting reports, but I still noticed meaningful differences between the two.
Gemini cited more sources (and did more searches), worked more quickly, and formatted its report in more engaging ways. To that last point, it allowed me to click through to different sections of the report via a drop-down menu at the top of the page. I also appreciate Gemini’s ‘thought process’ while researching, which felt more focused and less likely to go off on unnecessary tangents than ChatGPT’s. Gemini also offers more quality-of-life features, such as a one-click method for creating audio overviews, flashcards, infographics, quizzes, and web pages based on my reports. It’s also possible to export a report directly to Google Docs.
That said, ChatGPT’s in-text citations make connecting claims to sources more straightforward, and I love how ChatGPT asks follow-up questions about my prompt before it commences research. Furthermore, I didn’t notice a big difference in report quality between Gemini and ChatGPT, even though Gemini cited more sources. Even if the conclusions are similar, however, I’d rather reports cite as many sources as possible.
The difference in tone is noticeable. Gemini’s reports read more like essays, whereas ChatGPT’s read more like forum posts. I generally prefer the latter, but depending on the topic and your preferences, you might not.
Although it’s not a deep research feature, Gemini also offers the NotebookLM tool. Rather than generating research reports for you, this lets you upload your own documents, along with audio files, Google Slides presentations, websites, YouTube videos, and more. Then, NotebookLM connects and summarizes everything. This tool is intuitive to use and can be useful, but its core functionality is largely already possible in a regular Gemini chat.Image Generation: Best-in-Class Image Tech
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 (3 Pro/Nano Banana Pro), ChatGPT (GPT-5.1), and Copilot (GPT-5): “Generate me an image of a cozy suburban home with an open floor plan. I want to see a nice living space with a dining room, kitchen, and living room. Nothing too fancy.” Below are the results from Gemini (first slide), ChatGPT (second slide), and Copilot (third slide):
Gemini generated its photo the fastest, and its result is the most detailed and realistic. Even with all its details, the image limits errors and distortion to small instances in the background that you might not notice without close inspection. ChatGPT’s image has similar imperfections, but it’s far less detailed, looking more like a stock photo. Copilot’s image can’t compete with Gemini’s detail, either, and it features a distracting number of ceiling light fixtures.
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):
Once again, Gemini blows the competition away: Gemini’s comic has the highest-quality art, includes ample legible text, and features a cohesive twist at the end. ChatGPT’s image has generally disappointing art, doesn’t include text beyond a single word, and its twist in the final panel (if you can even call it one) doesn’t make much sense. Copilot’s comic doesn’t impress because of all the visible distortion, but it does manage to include a twist.
My final test was to generate a technical diagram with the following prompt: “I’ve got an Ethernet switch, a modem, and a router. I want to connect my PC, PlayStation, smart light hub, and smart TV via Ethernet. I want the modem to connect to the router, and then the router to connect to the switch. Then, I want the switch to connect to the devices. Generate me a diagram showing this.”
All three chatbots generally do a good job with this prompt, except for ChatGPT’s misspelling of “PlayStation” and Copilot’s repeated “Smart TV” text. However, Gemini still has the best image overall. Apart from avoiding errors and distortion, it includes the most accurate and detailed graphics, as well as some nice touches, such as small icons next to each item.Image Editing: Gemini Goes From Strength to Strength
Beyond image generation, Gemini can also do AI image editing. Simply send Gemini a photo and describe the edits you want. AI-edited images often look distorted or obviously AI-generated, but Gemini generally doesn’t have these problems. You might encounter some blurriness in image edits, however.
I asked Gemini (3 Pro/Nano Banana Pro), ChatGPT (GPT-5.1), and Copilot (GPT-5) to remove my hand from an image I provided in testing. You can see the original image (first slide) and the results from Gemini (second slide), ChatGPT (third slide), and Copilot (fourth slide) below:ChatGPTGoogle GeminiBuy It NowMicrosoft CopilotBuy It NowPerplexityClaudeGrok
No chatbot handled my prompt perfectly. Gemini stood out for its ability to match the original image’s aspect ratio, which ChatGPT and Copilot couldn’t do, even after specific prompting; however, its image appears a bit blurry. Copilot’s image is particularly disappointing, thanks to its overall graininess. ChatGPT’s image is the sharpest of the bunch, but it didn’t have to generate as much because it essentially just cropped out most of my hand.
Editing goes beyond making simple alterations to images. AI chatbots can also blend images, for example, if you want to add an image of a particular object to the background of another image. To test this feature, I took a picture of my kitchen, and then provided an image of a table and some chairs from Amazon. Next, I prompted the chatbots to add the chairs and table into the photo of my kitchen, making it look as realistic as possible.
Once again, no chatbot did a perfect job. However, Gemini’s image (first slide) shows the least distortion and actually makes an attempt to light the chairs and table appropriately. ChatGPT’s (second slide) and Copilot’s (third slide) chairs and tables don’t look as natural. Copilot’s image, in particular, suffers from a significant amount of distortion.
Unlike ChatGPT, Gemini doesn’t integrate directly with Adobe Express and Photoshop. Still, you can prompt Gemini to adjust the same aspects (brightness, contrast, saturation, and more) of an image with just as good (or better) results.Video Generation: On Par With the Best
AI video generation is a rapidly growing feature of AI chatbots. Gemini has the Flow filmmaker tool, the Veo 3 video generation model, and the Whisk AI animator. Gemini also generates audio with its videos. While Veo 3 used to be exclusive to Gemini’s top-end plan, you can now access it (albeit in a limited fashion) with Gemini’s AI Pro plan ($20 per month).
Veo 3 can produce 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 Gemini (Veo 3.1 Quality) and ChatGPT’s Sora 2 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.”
Veo’s video looks good until you notice that the spoon the subject grabs ends up both in his hand and where he originally grabbed it from. Furthermore, there’s a record player on the kitchen counter for some reason. ChatGPT’s video has similar problems. It treats a levitating cup as if it were a pour-over, and its subject weirdly crouches in front of a table instead of sitting on the couch behind it. The audio doesn’t quite sync up perfectly in either video.
To test how the video generators 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. Neither cube manages to avoid distortion. Veo’s camera zoom is distracting, and ChatGPT’s timer doesn’t line up with what happens on screen. The voice of ChatGPT’s person is also distorted in a way that really makes him sound like AI.
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.” These videos aren’t amazing, either. Veo’s video oddly begins with half the equation on the whiteboard, with the teacher filling in the other half. More importantly, the teacher speaks total nonsense. Meanwhile, ChatGPT’s teacher’s voice sounds distorted.
As mentioned above, Gemini’s Veo 3 model can generate some truly amazing things. However, as evidenced by my testing, achieving this requires careful calibration across multiple generations. This isn’t necessarily a big deal, but you only get so many generations per month. For example, the AI Pro plan gives you 1,000 credits per month, and each video you generate costs 100 credits. That seems quite restrictive for a premium feature.
Flow is another essential component of Gemini’s video generation feature set. 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: Keeps Pace With Competitors
Complex reasoning is another important aspect of chatbots. 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 3 Pro model, ChatGPT’s GPT-5.1 model, and Copilot’s Think Deeper mode.
Gemini and ChatGPT performed well, each getting two questions wrong. For Gemini, that was one computer science and one physics question. ChatGPT got two physics questions wrong. Copilot struggled in comparison, getting four math questions and a physics question wrong.
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.
Gemini doesn’t necessarily have a worse privacy policy than other chatbots, but the fact that it has access to all your web browsing, web search, and Google app data means its collection behavior can feel invasive at times. On the other hand, your data is in Google’s hands either way if you use Google products, so whether Gemini also has access to it might not matter to you.
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.