(Credit: René Ramos; Krot_Studio, AntonKhrupinArt / Adobe Stock)
The conversation around AI has reached a fever pitch, so it only makes sense that we should ask you, our readers, your feelings on perhaps the hottest topic in the world of tech. Recently, we ran our first-ever survey on generative AI, and when it comes to chatbots in particular, the results may surprise you. The overall Readers' Choice winner is probably not the brand you're expecting.
We also asked for your thoughts on AI in general, including the question “Will AI replace human jobs?” 68% of you agreed it will. On a scale of zero to 10, anxiety over that issue rated a firm 6.5. Yet when asked if their own job was at risk, readers rate their personal worry at only 4.2. We see much higher concern about the overall ethics of generative AI. A massive majority of readers—91%—feel there should be regulations around the development and use of AI tools.
We also asked if the emergence of generative AI has reduced respondents’ use of traditional search engines such as Google or Bing. 62% claim the answer is "no." But when asked if they use the AI results atop search engine results pages, only 20% say “never.” Many more say “sometimes” (44%), while the “always” contingent remains small, at 6%.
PCMag senior reporter Emily Forlini has been covering generative AI for the last two and a half years. "AI anxiety is high, but at the same time, more and more people seem to be interested in finding ways to use it," she says. "There's a feeling of 'If you can't beat 'em, join 'em,' and also genuine excitement when people find ways to make AI work."
When asked about the privacy issues posed by AI, 45% of you claim to actually read the privacy policy of an AI service. A quarter of respondents admit to sharing their personal info with a bot.
"It's hard to know how AI will change things in the future, but in the here and now, whether you read the privacy policy or not, AI companies are collecting copious amounts of user data," says analyst Ruben Circelli, whose job is to review generative AI tools for PCMag. "Almost all chatbots do this, and almost all of them use that data for training by default, so definitely be careful not to share anything too sensitive."
Read on to see which of the burgeoning AI chatbots and image generators deliver the most satisfying results to users, at home and at work.
The Top Generative AI Chatbots for 2025
The only constant is change, and in AI, evolution is maddeningly ceaseless. Almost every publicly available AI model has seen a recent update, some during our survey’s time in the field. So keep that in mind when reading the results.
In our inaugural AI Readers' Choice survey, the top pick for a chatbot overall comes from Anthropic, whose founders previously worked for OpenAI. Claude, its AI assistant (aka chatbot) named for an AI research pioneer, came on the scene in 2023 with a focus on reliability and safety. Just as our survey ended, the company announced updates to create Claude 4 (version 3.7 was only released in February, approximately the same time that our survey launched).
It should be noted that Anthropic’s overall satisfaction ratings are a C+ at best; our readers obviously don’t consider it, or any of the other major chatbots, a panacea. OpenAI and Google’s efforts are a half-point below Anthropic in terms of satisfaction; Microsoft and DeepSeek are more than a full point behind Anthropic.
Anthropic scores better for ease of use, stability, and understanding prompts. It wins in almost all the sub-categories we ask about. The exception is Google’s Gemini, which earns the best scores for data sourcing, both in terms of timeliness and citations.
(Note: Click the arrows in our interactive charts to view various elements of our survey results.)
Above, you can view the results for both free generative AI chatbots and those limited to personal use. (The majority of our readers specifically employ a free AI for personal use.) In the end, Anthropic earns a Readers’ Choice for both Best AI Chatbot for Personal Use and Best Free AI Chatbot. In fact, Claude’s grades are even higher than in the overall category when we limit the view.
“Claude AI is my go-to for when I have quite complex interactive questions to ask," one reader tells us. "It's pretty good about giving links to its source material. Most important, when I have fact-checked [Claude's] information, it has always proven to be accurate.”
Circelli sees both Claude's pros and cons. “In a world where chatbots envelop themselves in your personal information, Anthropic’s focus on privacy—making a decision not to train its models on user data by default—is refreshing,” he says. “However, Claude is also missing key features, like deep research and image generation, and doesn’t always search the web quite as effectively as ChatGPT.”
Forlini has followed Anthropic since it launched two years ago. “Anthropic has done a good job straddling the line of being a general-purpose tool," she says, "but also great at a specific task, which is software engineering and coding."
The only qualifying brand in our results for a paid chatbot is OpenAI, so there's no award for this year. But note that people paying for ChatGPT are much happier with it (giving it an 8.7 in overall satisfaction) than those using the free version (which gets a 7.0 in overall satisfaction). “The free system is way too slow,” complains one reader.
It’s worth noting that Microsoft, which has integrated Copilot AI into as many of its products as possible, consistently occupies fourth place in all of our Readers’ Choice charts. OpenAI and Google switch between second and third place, depending on the use case.
In all the charts, the last spot goes to DeepSeek, the Chinese open-source AI model that made a big splash earlier this year due to its free/open-source access model that was created for a relative pittance ($5.5 million to train it, compared with "a few 10s of millions of dollars" for Claude). Based on our results, it does not live up to the early hype.
“The initial hysteria around DeepSeek has certainly died down, but it lives on as a great option for software developers that need a cheap, powerful AI accessed through an API and who aren’t worried about the model spewing CCP propaganda,” Forlini says.
The Top AI Image Generators for 2025
OpenAI, the most used brand in our survey (by 61% of respondents), narrowly wins for image creation tools against Microsoft. This may have been helped by the fact that, during the survey, the OpenAI image creation model Dall-E 3 was replaced by image capabilities integrated directly into the GPT-4o model—a huge step up that greatly improves text rendering in images and maintains consistency across iterations of images, a longtime failing in AI image creation.
“People went wild for the new image generator,” Forlini says. “It was probably the biggest launch OpenAI has had since ChatGPT’s original debut."
Circelli concurs: “ChatGPT’s image generation is peerless among mainstream chatbots right now, managing markedly fewer errors and less distortion.”
Respondents generally view ChatGPT favorably. Some call it good but forgetful, while others say it is “as good (or better) than a therapist” for helping them navigate social situations. One respondent used the text and image generator to create an entire Dungeons & Dragons campaign, with full pictures of characters and maps.
The Top Generative AI Chatbots for Work in 2025
OpenAI also snags an award for office use of AI.
Very few of the people rating generative AI chatbots in our survey claim to use them for work. Only two brands, OpenAI and Microsoft, have enough work-related responses to make it on this chart at all, and it's OpenAI that takes our Business Choice award. OpenAI's ChatGPT stands a full point ahead of Microsoft Copilot (which itself is powered by OpenAI’s GPT-4, for now) in overall satisfaction.
As a work-based chatbot, ChatGPT is ahead of Copilot on most measures, but Copilot scores better for factual accuracy and data source citation. The two tie for the timeliness of the data sourced.
“Microsoft takes care to position Copilot as a tool for professionals, avoiding issues with generating content that violates policy,” Circelli says. “But in the areas that matter most, like accuracy, detail, and sourcing, ChatGPT generally performs meaningfully better, even though the core models powering Copilot are designed by OpenAI.”
To see which generative AI currently leads in our lab testing, read The Best AI Chatbots for 2025.
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.
Full Results
The PCMag Readers’ Choice survey for generative AI was in the field from Feb. 22 to March 12, 2025. For more information on how we conduct surveys, read our methodology.