Over the past two years, Chatgpt has become a default term for AI chatbots in the United States and Europe despite many viable competitors for a larger piece on the market. This is part of what made the eruption of China Chatbot, Deepseek, based in China, feels so seismic.
Deepseek’s rapid ascent has drawn enormous attention and use, but not without controversy. The large collection of user data for storage on Chinese servers is just a leading example.
I decided to put these two heavyweights from AI, Chatgpt and Deepseek, through their actions in the combination of their conversation capacities with online research, which is a particularly precious arena.
I designed four questions covering everything, sports news and consumer advice in the best local places for cocktails and comedy. I wanted to see how AI assistants would behave, so I mixed the specificity with the imprecision in the details. I used Deepsek R1 and Chatgpt-4o models to answer questions. Although R1 is comparable to the new O1 OPNAI model for chatgpt, this model cannot look online for the answers for the moment. You can see the IA questions and answers below.
One at a time
I also immediately discovered that even if Chatgpt was happy to answer several questions in a single prompt, Deepseek would only seek information about the first question and abandon the last, no matter how I wrote the initial invitation. Immediately, it was a point against. Although the conversational approach to the prompt and the answer is in many cases, you must sometimes ask a lot of questions for the chatbot or include several elements to consider. You can see how Deepseek answered an early attempt from several questions in a single invite below.
Count words
Even when it is divided into individual questions, the prompts to Deepseek required a little additional work in terms of definition of the amount of information that I wanted to receive. Depending on the type of question I have submitted, Deepseek almost always gave me too much information, and she was often foreign. Worse, sometimes the very long answer was just a filling, essentially telling me to look for things by myself. Chatgpt is not immune to similar behaviors, but that did not happen at all during this test.
And it was not only my own preferences, the same self -control was obvious when using chatgpt without connecting. I felt the need to handicap the test with a limit of 65 words to argue. With all these restrictions in place, here are the questions and answers of the AI. Chatgpt’s responses are on the left and Deepseek’s responses are on the right.
1. What were the highlights of the NBA match last night and who won?
2. What is a new trendy place for Brooklyn for cocktails and small plates?
3. Which laptop is the best for the game with a budget of $ 2,000?
Deepseek is lost
With the warnings of what was necessary to make the test possible, it is just to say that the two chatbots worked well enough. Deepseek had solid answers thanks to a much more in -depth research effort, which has drawn more than 30 sources for each question. The question of the cocktail bar, in particular, was great, and the AI was proactive enough to suggest a drink to obtain. The basketball response was also more substantial, but undoubtedly, Chatgpt’s decision to keep the emphasis on a game, as indicated by the singular “game” in the question, meant that it granted more attention.
It was in the responses to the recommendations of the computer and comedy that Deepseek displayed its weaknesses. The two looked less like conversational responses and more like the toplines of their Google summaries. To be fair, Chatgpt was not much better on these two answers, but the defect was less blatant, especially by looking at all the parentheses in the Deepseek computer response.
I understand why Deepseek has his fans. It is free, good to recover the latest information and a solid option for users. I just have the impression that chatgpt cuts in the heart of what I ask, even when it is not stated. And, while no technological business is a paragon of consumer confidentiality, the terms and conditions of Deepseek make other IA chatbots seem squarely polished with regard to the amount of information you must accept to share, until the very rhythm where you type your questions. Deepseek is almost like a joke on the depth of finding information about you.
In addition, Chatgpt was simply faster, that I use the DEEPSEEK R1 model or his less powerful brother. And, while this test was focused on research, I cannot ignore the many other Deepseek limits, as a lack of persistent memory or image generator.
For me, Chatgpt remains the winner when choosing a chatbot ai to do a search. Part of this can simply be the bias of familiarity, but the fact that Chatgpt gave me good answers to a single invite to resist as a killer functionality. This can become particularly true as the O1 model and the next O3 model get internet access. Deepseek can find a lot of information, but if I was stuck with it, I would be lost.
