Privacy concerns around chatbots are nothing new, but as AI applications become more widespread, users are increasingly aware of the risks. Duck ai, the chatbot from privacy‑first browser DuckDuckGo, may be benefiting from that trend.
At the same time: Stop revealing your secrets to AI — 5 reasons why, and what to do if you’ve already overshared.
New data from Similarweb shows that Duck ai’s web traffic exploded last month. “Duck ai reached 11.1 million visits in February, an increase of more than 300% compared to January,” Similarweb told.
DuckDuckGo did not immediately respond to a request for comment.
That number is still small compared to the leading chatbots — by comparison, Similarweb estimates that chatgpt reached 5.4 billion visits in February, gemini google 2.1 billion, and claude ai 290.3 million. Still, for a product that launched as a beta only a year ago, it’s a notable surge in interest.
Duck ai offers users the same privacy protection as its browser, anonymizing queries to prevent third parties from accessing chat history. The chatbot does not run its own proprietary large language models; instead, it uses cutting‑edge models from companies like Anthropic, OpenAI, and Meta, but calls those providers on behalf of users without exposing their IP addresses or other personal information.
At the same time: The permissions behind your AI Chrome extension deserve a close look — they could be watching you.
“Additionally, we have agreements with all model providers that further restrict how they can use data from these anonymized requests, including not using prompts and outputs to develop or improve their models, and deleting any information they receive when it is no longer needed to provide the output (within a maximum of 30 days, with limited exceptions for security and legal compliance),” Duck ai’s privacy policy states.
Someone tested Duck ai last year and found at the time that they preferred it over Perplexity.
So why the sudden traffic spike?
Compared to individual proprietary chatbots like Gemini and ChatGPT, Duck ai offers two main advantages: the option to switch between multiple models and enhanced privacy protection. That said, the latter may be what makes Duck ai unique, as Perplexity also provides access to multiple models through a single interface, in addition to its own Sonar model family.
Some Reddit users say they like Duck ai — one poster said “It’s so much better than Google’s,” apparently referring to Gemini, adding that it’s the reason they use DuckDuckGo. Many others call it “not bad,” neutral, just as disappointing as other options, or simply “better than nothing.” Some users dislike that Duck ai doesn’t support document uploads.
At the same time: Tired of seeing AI images online? DuckDuckGo now lets you hide them from search results.
In a Reddit thread from a few months before the traffic surge, users complained that Duck ai is just as disappointing as other chatbots, though one said it’s “not bad, especially if you care about privacy.” A more recent thread included complaints about usage limits, but didn’t mention specific reasons that bring users back to the chatbot.
At the same time: Want to know which sites are selling your data? This free privacy tool gave me answers.
Positive feedback is harder to parse because Duck ai provides access to several cutting‑edge models rather than a proprietary option of its own. Those who do like Duck ai praise the effectiveness of the specific models they choose within it, such as OpenAI’s GPT‑5 mini. However, for at least one user, Duck ai seems to affect those models’ effectiveness.
“A friend of mine uses regular chat‑GPT and swears he gets better responses than when using Duck ai,” one poster wrote. “I don’t know if that’s true. Maybe the privacy‑focused system prompt or something else they’re doing behind the scenes does alter the answers.” The poster added that they “really like Duck ai.”
Last month, Anthropic rejected some requests from the Department of Defense (DoD) to use its technology for weapons and mass surveillance, and the DoD retaliated by canceling the contract. OpenAI quickly stepped in, only to face the same usage controversy.
Coverage of the incident has pushed concerns about privacy and AI back to the forefront of public conversation. Nathan Calvin, Vice President of State Affairs and General Counsel at advocacy group Encode AI, told that he has noticed more discussion about data brokers, privacy, and how governments access data — both in the public sphere and in policy circles — since the contract incident.
“This is a long‑standing issue, but I do feel that many people are looking at it with fresh eyes and a sense of urgency,” he noted, adding that many “had never heard of Anthropic or Claude before the DoD story.”
Against this backdrop, chatbots that further protect user data from both AI companies themselves and government overreach may appear more attractive than before.
At the same time: Copilot is now quietly grabbing your data from other Microsoft products — here’s how to opt out.
But according to Similarweb’s chart, Duck ai began to see modest growth toward the end of 2025, 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 previously complained that text‑to‑speech was the only missing feature in their Duck ai experience, so that launch is likely a driver of the surge as well.
In keeping with the company’s other policies, Duck ai voice chat is anonymous, is not used to train models, and neither DuckDuckGo nor OpenAI (which powers voice support through Duck ai) stores the audio. That said, Duck ai advises users that their voice “could be a biometric identifier” and that they should consider that before trying the feature.
A user posted this month in the /ChatGPTcomplaints Reddit thread that they were trying Duck ai “for no other reason than hoping to rebuild my connection with 4o,” referring to GPT‑4o, a model OpenAI removed from ChatGPT in February, much to users’ displeasure. However, other users in the same thread noted their discontent with OpenAI’s DoD contract, which remains a possible driver of moving away from ChatGPT to any other chatbot (Anthropic’s Claude overtook OpenAI’s ChatGPT as the most downloaded free app in the US right after the contract controversy began).
You can try Duck ai for free, or get access to more advanced models for $10 per month ($100 per year if paid annually).
Q&A
Q1: What is Duck ai, and how is it different from other chatbots?
A: Duck ai is a chatbot from the privacy‑first browser DuckDuckGo. Its main feature is privacy protection — it anonymizes queries to prevent third parties from accessing chat history and does not expose users’ IP addresses or personal information. It also allows users to switch between multiple cutting‑edge models, including those from Anthropic, OpenAI, and Meta.
Q2: Why has Duck ai’s traffic surged?
A: Duck ai reached 11.1 million visits in February, a 300%+ increase from January. The main reasons include: growing user concern over AI privacy issues, especially following the controversy over Anthropic’s contract with the Department of Defense; the addition of image generation in December; and the launch of real‑time voice chat in mid‑February.
Q3: Does Duck ai cost money? Are there any feature limitations?
A: Duck ai offers a free version as well as a paid tier — $10 per month or $100 per year — for access to more advanced models. Some users have mentioned usage limits, though the specifics are not detailed. Currently, Duck ai does not support document uploads.