This privacy-first chatbot is taking off

Mohith S R/iStock / Getty Images Plus Follow ZDNET:  Add us as a preferred source  on Google.

ZDNET’s key takeaways  DuckDuckGo’s privacy-first chatbot is taking off. Users are increasingly concerned about how their data is used. New features could also be driving growth. Privacy concerns around chatbots are nothing new, but as AI adoption spreads, users are becoming increasingly aware of the risks. Duck.ai, the chatbot from privacy-focused browser DuckDuckGo, could be benefiting from that. 

Also: Stop telling AI your secrets – 5 reasons why, and what to do if you already overshared

New data from Similarweb found web traffic to Duck.ai exploded last month. Duck.ai “reached 11.1 million visits in February, up more than 300% from January,” Similarweb told ZDNET.

DuckDuckGo did not immediately respond to a request for comment.

That number is still small compared to the most dominant chatbots: By contrast, Similarweb estimates that chatgpt.com reached 5.4 billion visits in February, while gemini.google.com reached 2.1 billion and claude.ai hit 290.3 million. Still, for having only launched in beta a year ago, that’s a sharp uptick in visits worth keeping an eye on. 

Similarweb Duck.ai extends the same privacy to users that they have come to expect from its browser, anonymizing queries to prevent third parties from accessing chats. The chatbot does not run a bespoke LLM; it uses frontier models from Anthropic, OpenAI, and Meta, among others, but calls those providers on your behalf so as not to expose your IP address or other personal information. 

(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.)

Also: The permissions behind your AI Chrome extensions deserve a closer look – they may be spying on you

“In addition, we have agreements in place with all model providers that further limit how they can use data from these anonymous requests, including not using Prompts and Outputs to develop or improve their models, as well as deleting all information received once it is no longer necessary to provide Outputs (at most within 30 days, with limited exceptions for safety and legal compliance),” Duck.ai’s privacy policy says.

ZDNET’s Jack Wallen tested Duck.ai last year and found he preferred it over Perplexity at the time. 

Duck.ai’s popularity spike   So why the sudden jump in traffic? 

Duck.ai offers two main benefits over individual proprietary chatbots like Gemini and ChatGPT: the option to toggle between many models and increased privacy protections. The latter is likely what sets Duck.ai apart, though, as Perplexity also offers access to multiple models through a single interface, in addition to its own Sonar model family.

While some users on Reddit said they enjoy Duck.ai — one poster said “it’s way better than Google’s,” ostensibly referring to Gemini, and that it’s the reason they use DuckDuckGo.  Many others said it was “not bad,” neutral, comparably disappointing to other options, or just “better than nothing.” Some users dislike that Duck.ai doesn’t support document uploads.

Also: Tired of seeing AI images online? DuckDuckGo lets you hide them from results now

One Reddit thread from a few months before the spike is full of users lamenting that Duck.ai is just as disappointing as other chatbots, though one user said it “is decent, particularly if you are privacy focused.” A newer thread includes complaints about usage limits without mentioning the specifics of what keeps users coming back to the chatbot.

Also: Want to know which sites are selling your data? This free privacy tool gave me answers

Positive reviews are harder to parse because Duck.ai supplies access to several frontier models as opposed to its own bespoke option. Those who do enjoy Duck.ai cite the efficacy of the specific models they chose within it, like OpenAI’s GPT-5 mini. For at least one user, however, Duck.ai seemed to impact the efficacy of those models.

“A friend of mine uses regular chat-gpt and swears he gets better replies than he does with duck.ai,” one poster wrote. “I don’t know if it’s true. Maybe the privacy focused system prompts or whatever else they do in the background does change the answers.” That poster added that they “really enjoy Duck.ai.”

Increased privacy concerns Last month, Anthropic rejected some proposed applications of its tech for weapons and mass surveillance by the Department of Defense (DoD), which retaliated by cutting its contract. OpenAI quickly swooped in, only to encounter the same debates over use. 

Coverage of the incident brought renewed concerns about privacy and AI to the forefront of public conversation. Nathan Calvin, vice president of state affairs and general counsel at advocacy organization Encode AI, told ZDNET that he noticed an uptick in conversations about data brokers, privacy, and how the government obtains data since the contract incident, in both the general public and policy spaces. 

“It’s an issue that’s been around for a while, but I definitely feel like a lot of folks are taking a look at it with fresh eyes and urgency,” he noted, adding that many people “had never heard of Anthropic or Claude” before the DoD story. 

In that light, chatbots that go further to protect user data from both AI companies themselves and the government may look more appealing than before. 

Also: Copilot quietly grabs your data from other Microsoft products now – here’s how to opt out

But according to Similarweb’s graph, Duck.ai started seeing a slight uptick at the end of 2025 that then grew exponentially last month. Duck.ai added image generation to the platform in December; in mid-February, it added real-time, privacy-protected voice chat. Some Reddit users had complained prior to the release that text-to-speech was the only thing missing from their Duck.ai experience, so the release is likely a driver of the spike as well. 

In keeping with the company’s other policies, Duck.ai voice chats are anonymized and not used to train models, and neither DuckDuckGo nor OpenAI (which provides voice support through Duck.ai) stores audio. That said, Duck.ai advises users that their voice “can be a biometric identifier,” which they should consider before trying the feature. 

One user wrote on the /ChatGPTcomplaints Reddit thread this month that they were trying Duck.ai “for no other reason than to hopefully rebuild my connection with 4o,” referring to GPT-4o, the model OpenAI sunset in ChatGPT in February, to the displeasure of many users. Other users noted their frustration with OpenAI’s DoD contract in the same thread, however, which is still a possible driver away from ChatGPT and toward any other chatbot (Anthropic’s Claude overtook OpenAI’s ChatGPT as the most downloaded free app in the US directly after the contract dispute began). 

How to try Duck.ai You can try Duck.ai yourself for free or $10 per month (or $100 per year, if paying annually), which gets you access to more advanced models.

Artificial Intelligence
Read More

LEAVE A REPLY

Please enter your comment!
Please enter your name here