Live
Black Hat USAAI BusinessBlack Hat AsiaAI BusinessUK National Education Union poll: 66% of secondary school teachers in England say pupils using AI are losing their capacity for core skills like writing (Sally Weale/The Guardian)TechmemeAlibaba unveils agentic AI-focused model Qwen3.6-Plus - Seeking AlphaGNews AI AlibabaExperian uncovers fraud paradox in financial services AI adoptionAI NewsAutonomous AI systems depend on data governanceAI NewsCollege students say they are changing their majors because of AIBusiness InsiderRamp and Visa expand partnership to use agentic AI for corporate bill pay - Electronic Payments InternationalGNews AI agenticMen are ditching TV for YouTube as AI usage and social media fatigue growHacker News AI TopIBM strikes Arm deal in AI-focused hardware push - IT ProGNews AI IBMAlibaba releases Qwen3.6-Plus, its third proprietary, closed-source AI model launched within a three-day period, saying it "drastically enhanced" agentic coding (Luz Ding/Bloomberg)TechmemeRestore old photos easily with these 5 Google Gemini AI prompts - The Financial ExpressGoogle News: GeminiInside KPMG's push to turn tax experts into hands-on software buildersBusiness InsiderArm IBM Alliance Targets Mission Critical Role In Enterprise AI - simplywall.stGNews AI IBMBlack Hat USAAI BusinessBlack Hat AsiaAI BusinessUK National Education Union poll: 66% of secondary school teachers in England say pupils using AI are losing their capacity for core skills like writing (Sally Weale/The Guardian)TechmemeAlibaba unveils agentic AI-focused model Qwen3.6-Plus - Seeking AlphaGNews AI AlibabaExperian uncovers fraud paradox in financial services AI adoptionAI NewsAutonomous AI systems depend on data governanceAI NewsCollege students say they are changing their majors because of AIBusiness InsiderRamp and Visa expand partnership to use agentic AI for corporate bill pay - Electronic Payments InternationalGNews AI agenticMen are ditching TV for YouTube as AI usage and social media fatigue growHacker News AI TopIBM strikes Arm deal in AI-focused hardware push - IT ProGNews AI IBMAlibaba releases Qwen3.6-Plus, its third proprietary, closed-source AI model launched within a three-day period, saying it "drastically enhanced" agentic coding (Luz Ding/Bloomberg)TechmemeRestore old photos easily with these 5 Google Gemini AI prompts - The Financial ExpressGoogle News: GeminiInside KPMG's push to turn tax experts into hands-on software buildersBusiness InsiderArm IBM Alliance Targets Mission Critical Role In Enterprise AI - simplywall.stGNews AI IBM
AI NEWS HUBbyEIGENVECTOREigenvector

I tested ChatGPT vs. Claude to see which is better - and if it's worth switching

ZDNet AIby Elyse Betters PicaroApril 1, 202618 min read1 views
Source Quiz

Considering ditching ChatGPT for Claude? I tested both on the same 10 tasks. Here's which came out on top.

Elyse Betters Picaro / ZDNET

Follow ZDNET: Add us as a preferred source on Google.

ZDNET's key takeaways

  • Both excel at different tasks.
  • The free tiers are very limited.
  • Don't use one AI for everything.

I've used ChatGPT for years, but OpenAI's recent missteps, along with the growing number of people flocking to Claude, have had me looking at the competition. I'm not ready to completely abandon the platform I've spent countless hours learning and becoming proficient with because others seem to be jumping ship. But I am curious, so which AI is better?

So, I ran tests comparing ChatGPT and Claude across different tasks people use AI for, myself included. The one exception is coding. I'm not a programmer or developer, so I can't fairly evaluate those results. (ZDNET's David Gewirtz tests AI tools extensively for coding, though, so be sure to check out his recommendations.)

Also: Switching to Claude? How to take your ChatGPT memories with you

I looked at data to help inform which tests I'd conduct, including a 2025 Associated Press-NORC poll of 1,093 US adults on AI use, a Talker Research survey of 2,000 adults -- commissioned by Samsung in late 2025 -- on how people use AI on their phones, and OpenAI's own 'How people use ChatGPT' study, which shows usage shifting in mid-2025 from work to everyday tasks.

ChatGPT vs. Claude: Which is better?

Consider this a casual comparison for everyday users. I'm using only the free versions of ChatGPT (GPT-5.3) and Claude (Sonnet 4.6 and Haiku 4.5). There's nothing complicated here -- just the kinds of prompts you might realistically give either tool. I'm testing both AIs in my web browser and in their mobile apps, and I'm not using other tools, like Atlas or Claude Code.

(Disclosure: Ziff Davis, ZDNET's parent company, filed an April 2025 lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.)

Personally, I don't use AI for writing or editing. I genuinely enjoy the process. It's also my job, and I want to stay sharp and use my own words. But I'm not naive. Plenty of people turn to AI for writing, so let's test it.

Test prompt: "Write a brief email asking my manager Aly if I can get more time on an assignment due Friday. I want to sound confident and not overly apologetic."

Results

Claude on Sonnet 4.6 took several minutes to respond to my prompt, repeatedly saying it was "taking longer than usual," even after I started a new chat. When it worked, its writing was concise and more natural than ChatGPT, but both responses were still too formal for me and needed editing.

Also: Did AI write that? 5 ways to distinguish chatbots from human authors

ChatGPT replied instantly and offered tone customization, though it also showed an ad for email templates, while Claude was ad-free. I liked that Claude's response also included a "Send via Gmail" button.

Ultimately, the delay would push me to use ChatGPT in real life, even if Claude's writing is marginally better. Still, judging purely on writing, Claude edges out the win.

Show more

Elyse Betters Picaro / ZDNET

I have a love-hate relationship with AI images. While I hate AI slop all over my feeds, I use AI to create images, like when I brought my child's art to life or previewed a porch on my house before hiring a contractor to build it. Recently, I've seen a lot of talk about AI headshots, so I decided to test that with ChatGPT and Claude.

Test prompt: "Turn my selfie into a professional headshot suitable for LinkedIn. Do not change my face, and keep my likeness intact."

Results

Well, guess what? Claude doesn't directly generate images, so ChatGPT wins this test, hands down.

Also: The best AI image generators of 2026: There's only one clear winner now

Did I actually like the headshot it created of me? Sure. It didn't change my face or hairstyle, and it looked realistic. Do I plan to use it on LinkedIn? Um, no. Also, I'm just two tests in, and ChatGPT has already told me I have "3 messages remaining" and is pushing me to upgrade to ChatGPT Go to "keep the conversation going."

Show more

Elyse Betters Picaro / ZDNET

Many people speak to their AI chatbot instead of typing prompts, so let's see whether ChatGPT or Claude does a better job of responding quickly and accurately, without unnecessary fluff. I'm using a somewhat difficult word to pronounce, and I have a Canadian-leaning accent, so I'm curious if they will understand.

Test prompt: "I'm making a lemon beurre blanc to go with red snapper tonight, but I don't have white wine vinegar. What's a good substitute?"

Results

Claude thought I said "lemon burb long," then got hung up and froze. I ended the voice chat after about a few minutes of waiting and never got an answer.

Also: 7 surprisingly useful ways to use ChatGPT's voice mode, from a former skeptic

ChatGPT immediately asked me to pick a voice style. After I settled on one called Sol, I spoke my test prompt. Sol heard me the first time and quickly gave me the correct substitute for my recipe: lemon juice. Overall, I thought Sol sounded natural, casual, and friendly.

Show more

Elyse Betters Picaro / ZDNET

Some people use AI for advice on sensitive topics, including deeply personal questions about their mental and physical health. My test here is simple: will either AI tell me to consult a doctor if I ask about something concerning my health? Suggestions are fine, but the AI should ultimately point me toward a medical professional.

Test prompt: "I've been feeling very tired, physically sluggish, and mentally foggy for a few weeks now. What could be causing it?"

Results

ChatGPT listed common causes, including sleep issues, stress, poor nutrition, post-illness effects, possible hormone problems, and lifestyle factors like too much screen time. It told me to consider a "medical check" if my symptoms have lasted more than a few weeks and recommended basic bloodwork.

Also: Asking AI for medical advice? There's a right and wrong way, one doctor explains

ChatGPT also said I had two messages left as a free user.

Claude again took longer to respond but eventually gave a nearly identical response to ChatGPT's, although it told me outright to see a doctor now to get a basic blood panel.

I'd say they are tied.

Show more

Elyse Betters Picaro / ZDNET

For this test, I'm trying something I've actually been meaning to do: find my husband a John F Kennedy Jr inspired outfit from the 1990s, and I've been loving the new Love Story TV show. My husband's birthday is in a couple of weeks, and this feels like a fun idea. I have a budget in mind, and I'm interested to see what each AI will curate.

Test prompt: "Build a 90s JFK Jr inspired outfit, including shoes, pants, shirt, jacket, and accessories. Mix new and vintage pieces, and provide shopping links. I have a $350 budget."

Results

ChatGPT searched the web and quickly surfaced leather loafers, affordable chinos, and Oxford shirts, complete with shopping links to new products and inline images. The picks, in my opinion, felt very JFK Jr: classic blazers with soft shoulders and relaxed tailoring. It also suggested aviators and left an $85 buffer for thrifting accessories on my own. Overall, it was a strong starter outfit, but I was hoping for links to retailers that sell used clothing.

Also: How to shop with AI: 6 ways I find deals, price track, and let agents buy for me

Claude, by contrast, stalled. After about 10 minutes, it failed with "Claude's response could not be fully generated." Retrying didn't help, and starting a new chat wiped the previous one, so initially, I thought ChatGPT would win this round. But I retried with Haiku 4.5, a faster Claude model. Finally, it suggested loafers, chinos, a button-down, and sunglasses. Again, all consistent with JFK Jr style. It even recommended lived-in canvas baseball caps, leather-strap watches, and funky scarves. Very him, and something ChatGPT missed.

I really liked that Claude included links to vintage items on Poshmark, eBay, Depop, and Etsy. So, on curation alone, Claude did a better job. I just wish it had inline images.

Show more

Elyse Betters Picaro / ZDNET

Here's another topic I've been meaning to research. My dad wants to gift a property to my brother, but he doesn't want to involve lawyers, and my brother wants to avoid high upfront costs. They're not sure how to proceed. Can AI break it down clearly, with sources and links they can use to verify the information?

Test prompt: "My dad wants to transfer a mortgage-free house in Malone, NY, to my brother without a lawyer. They want to minimize upfront costs. Explain the best option step by step, with sources, including what they need to do, where to go, which forms they need to complete with official links, and the expected costs."

Results

In ChatGPT, I selected Deep Research. It started processing my request, but warned I was out of messages on the "most advanced free model" and would switch to a less powerful one until my limit reset in three hours. After waiting 40 minutes for a report to generate, I ran a normal search instead, which returned results instantly.

Also: I tested ChatGPT's Deep Research against Gemini, Perplexity, and Grok AI to see which is best

Claude also kept spinning and saying it was still working. To finish the test and keep it fair, I switched again from the default Sonnet 4.6 model to the faster Haiku 4.5 model, and that finally worked.

Both said my brother's best option was a quitclaim deed. They said he needed a notarized deed, linked to the required forms, explained where to file them, and estimated his costs. Both said there is no state transfer tax, but neither mentioned he could owe capital gains tax if he later sells, which I know from past experience.

Claude's answer was more scannable, but it had fewer sources. At one point, it cited the New York City Bar Association for the quitclaim deed, even though we are not in New York City. ChatGPT was more detailed and included far more sources, though some were random, like Legal Clarity and Avenue Law Firm. More importantly, it got a key detail wrong. It told my brother to file in Clinton County, but the property is in Franklin County.

Because of that, I have to give Claude the win for accuracy.

Show more

Elyse Betters Picaro / ZDNET

I sometimes use AI to read or find insights in documents, including spreadsheets. But, for this test, I'm evaluating both AIs on their ability to analyze a PDF. I've uploaded a 63-page PDF of OpenAI's "How people use ChatGPT" study, which I've already read and covered, to see how quickly and accurately they interpret the data.

Test prompt: "I uploaded a PDF. Summarize the entire document in one paragraph, and then list the three biggest findings or key takeaways as bullet points."

Results

Unfortunately, ChatGPT made me wait three hours to upload a file and test its capabilities. Claude also hit a rate limit, without any warning, and said I had to wait a couple of hours before I could ask anything else.

Also: Is ChatGPT Plus still worth your $20? I compared it to the Free, Go, and Pro plans - here's my advice

When I was finally able to upload the file, both chatbots provided a one-paragraph summary along with the three key findings I requested. Both concluded that nearly half of all messages to ChatGPT, from June 2024 to June 2025, were seeking information or advice rather than task automation. I fact-checked the stats and claims they presented, and everything appeared accurate. I could not determine a clear winner here, so it is a tie.

Show more

Elyse Betters Picaro / ZDNET

You could argue my research test above required multi-step reasoning, but most AI research does, right?

Still, it's worth evaluating, specifically, how well Claude and ChatGPT handle prompts with multiple steps. For this test, I'm asking how to get a boating license online, including a course recommendation, registration steps, and basic details like cost and time. That is at least four steps in one prompt. Let's see how they do.

Test prompt: "I want to get a boating license in New York as an adult. What's the best way to do it online? Recommend a course, and tell me exactly how to enroll. I also want to know the cost, how long it takes, and if there is a test."

Results

Claude on Haiku 4.5 quickly compiled a guide. It found the only free course approved by NY, told me how to enroll, mentioned a boating card and an optional "anchor" add-on to my NY license that would cost extra but would allow me to have just one card. It also included details about the course length and test. Throughout the guide, which was easy to scan and read, I saw plenty of sources.

Also: How to use ChatGPT: A beginner's guide to the most popular AI chatbot

Interestingly, ChatGPT immediately pointed me toward a paid course, with no explanation of why not to choose the free one until the very end, where it finally mentioned the free alternative. I did not specify that I wanted a free option, but you would expect it to prioritize that information, right? Like Claude, it did provide exact steps to enroll and answered my questions about course length and testing, with plenty of official source links.

I found Claude's response easier to read. I liked that there were no emojis, it stated upfront that getting a license can be free, and it surfaced the NY license add-on, which I had not known. It did a solid job of breaking down a complex query into manageable sections and delivering a useful, factual guide. So, Claude wins.

Show more

Elyse Betters Picaro / ZDNET

This one is complicated. While both Anthropic and OpenAI offer tools for experimenting with agentic AI and completing tasks autonomously, I want to test their free AI chatbots. I want free Claude and free ChatGPT to take over and buy event tickets, without me stepping in or requiring me to download a separate app like Atlas or Claude desktop. I want to do this from the mobile apps or from a web browser. Can they do it? Let's find out.

Test prompt: "Find and purchase tickets to the Montreal Biodome for this Saturday for my husband, me, and our 4-year-old daughter."

Results

That was quick.

Neither free ChatGPT nor free Claude can take over my screen and launch an AI agent to autonomously get me tickets. If I subscribe to ChatGPT Plus, Agent mode can handle this task not only in ChatGPT but also in the Atlas browser. In fact, in the past, I've used Agent to purchase Walmart grocery orders and Disney on Ice tickets.

Also: I let Claude AI control my Mac, and it worked flawlessly - with only two minor issues

With the free tiers, however, both chatbots are much more limited and only link out to ticket sites.

Show more

Elyse Betters Picaro / ZDNET

Last but not least, let's test an app integration.

Both AIs have a range of partners, from Instacart to Adobe, that let you get more done directly in the chatbot without switching apps. I want to try building a travel itinerary presentation in Canva using ChatGPT and Claude. I know I can refine it in Canva later, but how well can each AI handle it from the start with just one prompt?

Test prompt: "I've uploaded a PDF of my itinerary for an upcoming camping trip to the Thousand Islands. Create a presentation with slides outlining each day of the trip. Include visuals and a cohesive, location-inspired theme. For each day, cover accommodations, activities, transportation, restaurants, and estimated costs."

Results

Once I connected Canva, Claude on Haiku 4.5 generated a 10-slide presentation with appropriate colors, but the overall design was very rudimentary and lacked visuals. As a creative person, I would not use this deck.

Also: How to use Canva in ChatGPT to build a stunning presentation in minutes - for free

As for ChatGPT, after connecting to Canva, it generated a presentation fast, but I could not get the preview to load in the chat. I eventually asked it to share a Canva link, and then I could finally open it in the Canva app and see it. I really liked the layout and colors. I would probably replace the AI-generated images of real locations with actual photos, but this is a ready-to-use, editable deck I would feel comfortable refining further and sharing.

I'm giving this last round to ChatGPT, for its Canva integration anyway.

Show more

Elyse Betters Picaro / ZDNET

  • ChatGPT: 3 wins (plus it tied twice)
  • Claude: 4 wins (plus it tied twice)

Both Claude and ChatGPT have free tiers and paid plans. For this comparison, I used the free versions.

Free Claude on Sonnet 4.6 was extremely slow, to the point of being nearly unusable. Anything requiring multi-step reasoning or more than a simple reply struggled. To finish testing, I switched to the faster Haiku 4.5 model, which worked better, but after about six prompts, I hit a rate limit and had to wait a couple of hours to continue using it, which was frustrating.

To be fair, ChatGPT also hit rate limits after about six prompts. But I could still use it. It just downgraded me to a "less powerful" model that restricted more advanced tasks, like file uploads. The bottom line is that both free tiers are very restrictive. Unless you are a very casual user willing to deal with slowdowns and rate limits, I would recommend upgrading to a paid plan.

Also: Switching to Claude? Here's how to take your ChatGPT memories with you

If I set all that aside and look at the results of my tests, Claude won more, by one round. However, the lack of direct image generation is a big drawback and even a deal-breaker for me. For writing, shopping, research, and multi-step reasoning, though, Claude proved very capable and more accurate and helpful.

For image generation, voice interactions, and app integrations, use ChatGPT. It shines at those tasks.

I thought both were equally good at handling sensitive topics and file analysis. But if you want to try agentic AI for free, you'll have to look elsewhere.

Show more

How much do ChatGPT and Claude cost?

Here's a quick look at how much they cost and each tier's main features if you want to upgrade from free:

TierChatGPT (OpenAI)Claude (Anthropic)Free$0/month: GPT-5.3. Limited usage with ads. Includes web search, image generation, 5 Deep Research runs, and Agent mode. Low message volume.$0/month: Claude Sonnet 4.6 and Haiku 4.5. Limited usage. No image generation. Includes web search. Low message volume.Budget$8/month (Go): Higher message volume than Free. Still includes ads.No direct budget tier.Plus / Pro$20/month (Plus): Full GPT-5.4, Sora video access, 25 Deep Research runs, and 5x usage vs Free.$20/month (Pro): 5x usage vs Free, Claude Code access, and priority access to Sonnet 4.6.Power User$200/month (Pro): Full GPT-5.4 Pro, 250 Deep Research runs, and "unlimited" use.$100-$200/month (Max): Tiers for 5x ($100) or 20x ($200) Pro usage limits.

Is Claude or ChatGPT more accurate?

Based on my experience, both are often accurate, but AI can and does make mistakes or fail to present the full picture.

In my tests above, Claude leaned more accurate, especially in my sixth test about my brother's property transfer issue, where it got a key location detail right over ChatGPT. However, both failed to present the full picture by explaining capital gains taxes. This is to say it depends on the type of question, whether it involves multi-step reasoning, and more.

With either AI, you should always fact check.

Should I switch to Claude from ChatGPT?

The most honest answer I can give you is: you should stop using ChatGPT for absolutely everything. As ZDNET's David Gewirtz explains, different AI tools are better suited to different tasks. While he still recommends ChatGPT in some cases, he often turns to other tools, including Claude. My tests support that, as well.

Show more

Editorial standards

Was this article helpful?

Sign in to highlight and annotate this article

AI
Ask AI about this article
Powered by Eigenvector · full article context loaded
Ready

Conversation starters

Ask anything about this article…

Daily AI Digest

Get the top 5 AI stories delivered to your inbox every morning.

Knowledge Map

Knowledge Map
TopicsEntitiesSource
I tested Ch…claudechatgptZDNet AI

Connected Articles — Knowledge Graph

This article is connected to other articles through shared AI topics and tags.

Knowledge Graph100 articles · 193 connections
Scroll to zoom · drag to pan · click to open

Discussion

Sign in to join the discussion

No comments yet — be the first to share your thoughts!