Although Anthropic's Claude AI chatbot once stood out for its user-friendly privacy policy, it no longer has that distinction. It still offers competent reasoning, creative writing, deep research, and web searching skills, and we like its sleek user interface and interesting app maker. However, its move to train on user data by default, along with the continued absence of image and video generation tools, are both downsides. Claude is worth exploring if you want a chatbot that's both enjoyable and reliable (enough to earn our Readers' Choice award), but ChatGPT is our overall Editors’ Choice winner, thanks to its more comprehensive feature set and peerless response quality.

What Is Claude?

Claude is an AI chatbot, like ChatGPT and Gemini, that you can chat with over text. You can use Claude to answer questions, do research, write creatively, search the web, solve equations, and many other things. Put simply, Claude is an advanced virtual assistant.

You can use Claude to code, and Anthropic even has an agentic coding tool, Claude Code, that lives in your terminal and understands your codebase, learning how to help you code faster over time. That tool is outside the scope of this review, but you can still test Claude’s coding abilities for yourself.

I think AI chatbots work best for answering questions and searching the web. Depending on what you’re looking for, it can be easier and quicker to simply ask a chatbot like Claude instead of Googling something and scrolling through search results. Claude’s deep research might be a better jumping-off point for learning about certain complex topics, but more on that later.

Regardless of which chatbot you use and for what, it’s important to keep in mind that they are fallible. Chatbots can and will confidently provide incorrect information, so you shouldn’t take everything one says at face value. Make sure to check anything you learn from a chatbot against a reputable source if it’s important.

How Does Claude Work?

At the most basic level, Claude takes in prompts and puts out responses. It does this via large language models (LLMs) made up of artificial neural networks trained on massive amounts of data. These models are complicated algorithms that give Claude the ability to access information on everything from Peruvian cuisine to Soviet architecture and ultimately answer your questions. Claude can also search the internet for knowledge on current events.

Claude isn’t like traditional software: Anthropic doesn’t need to update it for it to get better. Much like a person, Claude can learn and improve over time. Simply by talking to Claude, you help to train its underlying models, assuming you don’t opt out of Anthropic training its models on your data. However, don’t expect Claude to change from chat to chat—it’s a gradual process.

Claude uses two primary lines of models: Opus and Sonnet. Sonnet is Claude’s conversational, flagship line, while Opus specializes in complex reasoning, making it a good fit for things related to coding, math, and science. Opus 4.1 and Sonnet 4.5 are the latest versions of Claude’s primary models, though you can also access Opus 4 and Sonnet 4, among others, with a paid subscription. Claude has a third model line, Haiku, that delivers responses faster than Opus or Sonnet at the expense of accuracy and detail. My testing focuses on Sonnet 4.5 (Claude’s default model) and Opus 4.1.

Pricing: Slightly Cheaper Than Competitors

Claude has a free version, but premium plans increase usage limits and unlock more features.

For free, you can use Claude’s Sonnet model to answer questions, generate code, make apps, process files, search the web, or visualize data, among other things. A Pro plan ($17 per month, billed annually) gives you access to Claude in your terminal, deep research, extended thinking for complicated tasks, full integration support, higher usage limits, models beyond Sonnet 4, and unlimited Projects for organizing chats and documents.

The top-tier Max plan gives you priority access to Claude during high-traffic times and even higher usage limits. Max comes in two flavors: 5x more usage than Pro ($100 per month, billed annually) and 20x more usage than Pro ($200 per month, billed annually). You also get early access to new features, but none were available at the time of writing. The vast majority of people interested in premium plans, though, should opt for the Pro plan, which has generous usage limits. For this review, I primarily tested the Free and Max 5x tiers.

For chatbots, $20 per month is the standard fare for a flagship premium plan. Thus, Claude Pro’s $17-per-month rate offers some savings. The top plans from ChatGPT ($200 per month) and Gemini ($250 per month) are in line with Claude’s most expensive Max plan, but Claude gives you the option to sign up for a cheaper version of Max, which I appreciate. However, Claude’s free plan lacks a lot of the functionality you get with ChatGPT and Gemini, including deep research, image and video generation, model switching, and voice chat. In fact, Claude doesn’t support image and video generation at all.

Where Is Claude Available?

Claude is accessible as a desktop app (available for macOS and Windows), a mobile app (available for Android and iOS), and on the web. It has an official Chrome extension, but that’s limited to Max subscribers.

Other services and sites use Claude’s models, such as Perplexity, but Anthropic doesn’t develop, operate, and own these iterations, meaning they aren’t part of the official Claude package. Regardless of where you see Claude’s models pop up, stick with the official apps and site if you want to access its full set of features.

Ease of Use and Interface: An Elegant Design

To start using Claude, you need to create an account, which requires an email address and a password. Signing in brings you to your dashboard. Claude has the same basic design as pretty much every chatbot: A text field for starting chats sits in the center alongside a left-hand sidebar that shows your conversation history. Buttons below the central text field provide ideas on what exactly you can use Claude to do, from vibe coding to writing a case study.

Claude's core design might not be vastly different from other chatbots, but I enjoy interacting with it more. Its color scheme, font, and slightly blockier aesthetic are pleasing and make it feel less sterile than ChatGPT or Gemini. Even across deep research progress screens and settings pages, Claude’s interface is cohesive. It is consistently tidy and readable.

(Credit: Anthropic/PCMag)

From your dashboard, you can talk to Claude about anything. Responses aren’t quite as snappy as with Gemini, but they aren’t as consistently slow as with DeepSeek. Once you get a response, you can copy or regenerate it. An easily accessible share button is at the top of the screen. Sometimes, responses hang or encounter errors during generation. However, this isn't a major problem with Claude, and all chatbots trip up occasionally.

ChatGPT has had issues being overly friendly in the past, but Claude’s tone, by default, is closer to Gemini’s. In other words, you can expect more direct, slightly stiffer responses. However, you can customize Claude’s personality to your liking, like you can with ChatGPT. Claude calls these styles, and a variety are available to choose from, such as concise or formal. You can also create a custom style simply by describing how you want Claude to communicate or providing an example. That’s not the end of the customization options, either. You can tell Claude to keep any manner of personal preferences in mind or to refer to you by a certain nickname, too. 

Claude can reference and search past conversations when you bring up any related topic. It can also generate its own ‘memory’ as you tell it things, allowing it to remember context and detail. I appreciate the granular control Anthropic gives you over Claude’s memory: You can toggle the ability to generate memories or search chats. Claude shows you what it knows about you in its settings page.

(Credit: Anthropic/PCMag)

Like DeepSeek, Claude doesn’t support speech-to-text or voice chat on the web. However, like Gemini, Claude does offer a voice chat mode for its mobile apps. You can choose from a variety of lifelike voices, albeit ones that don't sound quite as natural as Copilot Voice, and converse about all the same topics you can over text, including files you upload.

However, Claude's voice chat mode is uncharacteristically clunky. It isn't free-flowing, like with Gemini, meaning you (typically) need to click send after speaking. Claude sometimes responded to me before I sent my message, however, which made for an inconsistent experience. The microphone is always active in voice mode, too; it doesn't even turn off if you minimize the app.

Web Search: Competent Responses

Chatbots don't have to rely just on their initial training data. Now, every chatbot, including Claude, can search the web. Accordingly, Claude (Sonnet 4.5), ChatGPT (GPT-5.1), and Gemini (2.5 Flash) had little trouble answering questions about current events at the time of testing, such as when Call of Duty: Black Ops 7 went live in my time zone or what the latest news about Jeffrey Epstein was. Claude managed to tell me which Incarnon weapons were available in Warframe’s current rotation, whereas ChatGPT and Gemini could not. Of course, Claude can and does get this question wrong sometimes.

(Credit: Anthropic/PCMag)

It's also possible to trip Claude up. For example, it told me the current pope was Pope Francis, even though I made sure to toggle the live web search option. I had to explicitly tell it to search the web to get the correct answer. With ChatGPT, you just need to be careful to ask for the current pope or toggle the web search option before you ask your question. Otherwise, ChatGPT tells you it's Pope Francis, courtesy of its outdated training knowledge. 

Claude is slightly slower than competitors at searching the web. Its responses aren't quite as detailed as ChatGPT’s, and its sourcing isn’t as robust, either: ChatGPT highlights the relevant portion of its response when you hover over a source, which makes it easy to connect claims to evidence, and its source icons display the name of the source. Claude doesn’t do in-text highlighting, and its source icons show the name of the attached article, which is less useful at a glance.

Gemini can show you an image if you ask for one, while ChatGPT includes images in its responses automatically when relevant. Claude, on the other hand, doesn’t display images, even if you ask. 

You can ask Claude for buying advice, but it doesn’t have a shopping component. ChatGPT and Gemini can provide clickable tiles with links to products and other relevant information.

Deep Research: Competitive With ChatGPT

I love using chatbots' deep research features: Ask a question, and you'll get a comprehensive report complete with dozens of citations. I’ve used chatbots to research everything from how to translate controller settings between games to which egg brands have the lowest incidence rates of salmonella.

Claude calls its deep research feature 'Research.' Unlike ChatGPT's and Gemini's equivalent features, Claude’s requires a paid subscription. Copilot is the same way. ChatGPT and Gemini merely limit how much deep research you can do for free. 

Like ChatGPT, Claude asks follow-up questions on research prompts before it begins its work. Gemini, on the other hand, provides a research plan you can tweak. In general, I prefer ChatGPT’s and Claude’s questions since they’re easy to answer and often provide essential context for the research at hand, which Gemini sometimes struggles without.

Like other chatbots, Claude handles straightforward research queries with ease, such as providing a synopsis of the Iran-Israel conflict or suggesting ways to augment a central air system that struggles in an intensely hot climate. However, chatbots can encounter issues when researching complex topics with conflicting sources. As such, I tasked Claude and ChatGPT with researching strategies for efficient gear and money farming, optimizing graphics settings, and unlocking achievements in the ARC Raiders video game.

Compared with the last time I tested Claude’s deep research ability, its performance has improved. In the past, Claude gathered more sources than ChatGPT would for the same research prompt, but produced far less detailed reports. That made sense, however, since it took less time (a handful of minutes versus 10 or 15). Now, Claude’s reports are more detailed and take longer to generate, putting them more in line with ChatGPT’s. But it still incorporates only a fraction of the sources it gathers, which is disappointing. For example, although Claude gathered 193 sources for one test report, large sections of it relied on just one.

Claude's deep research interface is pleasant to use. Like ChatGPT, it provides in-text links. These are more readable at a glance than Gemini’s source tiles, which it hides behind clickable carets. Also like ChatGPT, Claude displays useful stats, such as the total research time, total searches, and total sources; Gemini doesn’t do the same. Claude has a dedicated screen that breaks down its searches by website into a bar graph, too, which is nicer than the simple lists of searches and sources you get with ChatGPT and Gemini. I did miss Gemini’s one-click button that exports reports to Google Docs and how ChatGPT displays reports in its central chat window, instead of relegating them to a sidebar, though.

In terms of tone, Claude’s reports read like semiformal essays, similar to those from Gemini. ChatGPT’s reports read more like forum posts. I slightly prefer ChatGPT’s tone here, but you might not, depending on your research topic.

File Processing: Inconsistent Analysis

Claude can process files you upload, which means it can analyze and understand documents and images to critique resumes, evaluate fashion choices on video game characters, identify plants and animals, translate text, or do a million other things. Image recognition and complex document processing, though, are tough tasks for chatbots, so I tested them.

Similar Products

Our Current Picks for The Best AI Chatbots for 2026

Google Gemini

4.0 Excellent

Read Our Review

ChatGPT

4.0 Excellent

Read Our Review

Microsoft Copilot

4.0 Excellent

Read Our Review

Perplexity

3.5 Good

Read Our Review

Claude

3.5 Good

Read Our Review

Grok

3.5 Good

Read Our Review

To start, I provided an image of my computer (pictured below), which includes ample RGB lighting and reflections that might trip up a chatbot. I asked Claude (Sonnet 4.5), ChatGPT (GPT-5.1), and Gemini (2.5 Flash) to identify as many components as possible with as much specificity as they could muster. I also asked them not to reference prior conversations in their answers, since I’ve run similar tests before.

(Credit: Ruben Circelli)

All the chatbots made some mistakes and were occasionally vague in their identifications. However, Claude had the least detailed response of the bunch. For example, Claude identified a “cylindrical reservoir tube visible in the center-right area,” whereas ChatGPT and Gemini identified the same component as a D5 pump. Furthermore, ChatGPT provided a much longer answer than Claude, while Gemini thoughtfully included a section on each component’s role in the larger cooling system, in addition to its identifications, which Claude didn't do.

Images aside, how effectively does Claude handle document processing? I uploaded the manuals for my case, CPU waterblock, and fan controller to Claude, ChatGPT, and Gemini. Then, I gave them the following prompt: “Based solely on the provided manuals, what RGB connections do my CPU block and case have, and can they connect directly to my Octo fan controller?”

Claude mistakenly told me that both my case and CPU waterblock could connect directly to my fan controller, while both ChatGPT and Gemini correctly concluded that I could not connect either without an external adapter. Claude’s performance doesn’t impress, but your mileage might vary. When I’ve run this test before, Claude gave more accurate answers, going so far as to provide part numbers for possible adapters I could use.

Nonetheless, I don't recommend uploading your homework for Claude to do for you: It can sometimes, like ChatGPT and Gemini, make up quotes and misunderstand documents. As always, make sure to verify the answers you get about anything important.

Creative Writing: On Par With Other Chatbots

Chatbots can do all manner of creative writing, whether that means preparing a speech for a conference, telling a joke, or writing a personalized message. Of course, chatbots can tell stories, too, but as they get more advanced and their underlying technologies mature, evaluating their creative writing abilities goes beyond whether they can generate something coherent. 

I focused my creative writing test on poetry, which requires careful attention to detail across not just content but also punctuation and style. I gave the following prompt to Claude (Sonnet 4.5), ChatGPT (GPT-5.1), and Gemini (2.5 Flash): “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.”

The results were mixed. Claude included a title for its poem, which is a nice touch, and it had the best punctuation variety, but it also read the most like prose. ChatGPT’s poem read less like prose and did relatively well with punctuation variety, but it didn’t otherwise impress and left out a title. Gemini’s poem stood out among the group with an interesting structure that didn’t read like prose at all, but it didn’t have much in the way of punctuation variety, nor did it feature a title.

All three poems excel in different areas, so it’s hard to pick a clear winner. With Claude, it’s worth noting that sometimes it generates creative writing as a clickable Artifact, allowing formatting to remain consistent across different screen resolutions and window sizes, which I appreciate as a writer. You can ask Claude to specifically generate your poem as an Artifact if you prefer.

Complex Reasoning: Extended Thinking Is a Must

Chatbots aren’t just creative: They can solve complicated problems, too. To test this ability, I gave Claude, ChatGPT, and Gemini exam questions from undergraduate courses in computer science, math, and physics from Harvard, MIT, and Stanford. Afterward, I compared their responses to the solutions. I used each chatbot’s complex reasoning model: Claude’s Opus 4.1, ChatGPT’s GPT-5.1 Thinking, and Gemini’s 2.5 Pro. If you sign up for a premium Claude subscription, you can enable extended thinking, which I did for the following test.

All three chatbots answered the computer science and math questions without issue, while physics posed some problems for Claude and Gemini. Claude answered two-and-a-half physics questions incorrectly, while Gemini answered half a physics problem incorrectly. ChatGPT is the star performer in complex reasoning, answering all questions correctly. However, Gemini’s results are, more or less, the same as ChatGPT’s, considering that the part of the one question it stumbled on came down to it including only one of two correct answers. 

Although Claude lags behind the competition slightly in complex reasoning, it can generally handle complicated prompts well. That said, I don’t recommend using any chatbot for mission-critical complex reasoning tasks, because even the top performer can and will get things wrong from time to time. If you plan to study with a chatbot, make sure to double-check the information it provides.

Integrations: Desktop and Web Offerings

Claude now has a ton of different Connectors (what it calls integrations). Some connect Claude to mainstream apps, such as those within the Google Workspace suite, while more novel ones AI video generation or payment processor apps. Claude's Connectors do everything directly within its interface. Gemini, for comparison, pops up in the apps it integrates with; for example, you get a Gemini button directly within Gmail.

(Credit: Anthropic/PCMag)

On the one hand, this behavior makes some of Claude’s integrations, like within Workspace apps, feel less convenient and robust. I can ask Claude to list my unread emails or find a particular email that a keyword search isn't surfacing, but its analysis of my email often feels superficial and biased toward newer messages. Gemini doesn’t have this issue.

On the other hand, since Claude doesn’t only integrate with apps its parent company owns, like Gemini, you get a wider variety of options. For instance, you can connect Claude to Asana for coordinating tasks and projects; Canva for creating new graphic designs; Notion for taking notes; or Workato for automating workflows, among other options. Keep in mind, however, that the above are just its web integrations.

If you download Claude’s desktop app, you can get another set of integrations via downloadable extensions. These similarly run the gamut, allowing you to control desktop apps such as Apple Notes, Chrome, iMessage, Spotify, your file explorer, and more. No other chatbot I've tested offers anything similar.

However, the actual performance of some of these integrations is mixed, and you need to keep some limitations in mind. For example, when I installed Claude’s Filesystem Connector and gave it access to a folder with creative writing, Claude couldn’t understand some files within it, failed to read through all the files it could understand, and even crashed occasionally. I appreciate Claude’s pop-ups that asked whether I was really sure I wanted it to do something from a privacy perspective, but these can also be a bit overbearing at times.

Other restrictions exist, too. For instance, some integrations depend on AppleScript and thus work only on macOS. This means that Claude can't control Chrome or Spotify on Windows. iMessage and Notes integrations are also exclusive to macOS, too, of course. That said, many of Claude’s desktop integrations aren’t exclusive to macOS.

Both Claude’s web integrations and desktop extensions have potential, but I don't yet have strong reasons to use them. I'm not sure why I would use Claude to control Canva or Spotify, for example, when I can just use the apps myself.

If you pay for Claude’s Max plan, which starts at $100 per month, you can join the waitlist to try Claude’s Chrome extension, which unlocks AI web browser features. It’s still in beta, but you can expect a Claude-powered AI agent that can take control of your browser and perform tasks for you, alongside a one-click way to talk to Claude from within Chrome. For now, though, the extension suffers from many of the same drawbacks as other AI web browsers.

Artifacts: Great Concept With Some Bugs

When you use Claude to do deep research or generate creative writing, it presents these responses in the form of Artifacts, which are clickable tiles that open in a sidebar. You can view all your Artifacts from Claude’s Artifacts tab on the left side of the screen while you use the chatbot, which is a great user experience innovation. ChatGPT and Gemini don't give you an easy way to view all your generated content at once.

However, Claude’s Artifacts are much more than simple user interface elements. Artifacts can be apps, which Claude can build for you, even if you don’t know anything about programming. Simply describe your app to Claude, and it will make it. You can browse a library of Artifacts to try out before you create one, too. After you generate an Artifact, you can publish and share it. Artifact apps you can try out include flashcard apps, interactive drum machines, quizzes, simple games, writing editors, and much more.

The apps Claude can create are fairly basic web experiences, so you can’t have Claude develop a new desktop program or iPhone app. However, what it can do is still quite impressive, if you’re willing to put in the necessary time to tweak them. For example, I asked Claude to make me a build calculator for the Warframe character Oraxia. My app allows me to see all her stats and how they change as I set up a build for her. 

(Credit: Anthropic/PCMag)

However, creating apps can be a hassle: I experienced context window limits, which required me to start fresh chats, and occasional crashes. I have occasionally noticed significant memory usage in my browser. This isn’t the situation for every app, especially those that are less complicated. While Artifact app creation can be frustrating at times, it’s still cool to be able to design custom apps at all.

What Can’t Claude Do?

Most importantly, Claude is not conscious. It can’t feel anything, think for itself, or understand like a human being can. In other words, it can’t be your friend, romantic partner, or therapist. Claude is a complicated algorithm designed to accept prompts and return responses.

Adult content, illegal activity, and taboo subjects like hate speech, among other things, are against Anthropic’s policies and Claude’s constitution. With enough prompt calibration, you likely can get around these built-in restrictions, but Claude didn't violate its own policies during testing. I can't say the same for DeepSeek and Gemini.

Claude also has practical limitations, like its context window. Think of a context window as short-term memory. Its size determines how much data it can process in a single conversation. For free users, the context window is entirely dynamic depending on server load. Paid users get up to a 200,000-token context window, but this is also dynamic. I was able to hit a context window limit when uploading documents as a free Claude user, and also when iterating on an Artifact app as a paid user.

Alongside the context window size are usage limits, which determine how many messages you can send to Claude within a five-hour period. For free users, this number is dynamic. Pro users can send up to around 45 messages, and Max tiers support up to 225 (5x Pro) or 900 (20x Pro) messages. Anthropic dynamically caps paid tiers, too, though, so you might be able to send more messages, depending on server load.

Is Your Data Safe With Claude?

One of Claude’s main selling points was its user-friendly approach to data collection and privacy, but Claude recently changed its policy to allow for the training of its models on user data by default. You can opt out, but this shift is disappointing. Claude collects contact data when you sign up for an account, payment information if you choose to purchase something, conversation data when you use Claude, and any feedback you give to Anthropic on its services. Claude also receives certain information automatically, such as cookie data, device and connection data, log and troubleshooting data, and usage information (like time and date data, among other things).

By default, Claude encrypts personal user data, and Anthropic employees can't access it outside special circumstances, such as enforcement of Claude’s usage policy. Claude gives you the ability to delete conversations, and, in general, deletes conversation history automatically from its back end within 30 days. 

Anthropic is no stranger to data-related scandals. In 2024, Anthropic suffered a data leak, accidentally sending customer information to a third party, and Reddit is currently suing Anthropic for using bots to access its site after Anthropic allegedly stopped its bots from doing so. Beyond that, Anthropic supports classified collaboration between AI labs, like itself, and the intelligence agencies. Anthropic has also come under fire for purchasing physical copies of books, scanning them, and then destroying the original copies. 

Like with any online service, I don’t recommend sharing significant amounts of highly personal or sensitive data with Claude. I don't see much reason to trust Claude more than other chatbots due to the policy changes. 

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.