GoKawiilMegan Ellis / Android Authority
When it comes to choosing the best AI chatbot, the winner will depend on who you ask. Some people swear by Gemini and ChatGPT, but others prefer Claude. While I use both Gemini and Claude for everyday tasks, I prefer Claude for its high-quality responses and effectiveness for certain tasks, such as generating recipes.
That’s why I’m usually happy whenever I see Google take inspiration from Anthropic’s chatbot. However, a recent change to Gemini saw the AI tool adopt one of the most frustrating features of Claude — rather than learning from the many improvements it could make.
Do you prefer using Gemini or Claude? 66 votes Gemini. 47 % Claude. 39 % I use another AI tool (let us know in the comments). 5 % I don't use AI. 9 %
Gemini could learn a lot from Claude
Brady Snyder / Android Authority
In theory, Gemini should be the more proficient chatbot. It’s multimodal, meaning it can generate responses in a variety of formats, including images and videos. As a Google product, it also benefits from the company’s dominance in free apps and its status as the primary developer of Android. It also has unique features like Gemini Live, which allows you to hold audio conversations with the chatbot or use your phone’s camera to show it a live video feed and ask it questions about what you’re currently recording.
And yet, when it comes to anything mildly important, I turn to Claude instead. In my experience, Anthropic’s AI tool just delivers higher-quality responses. It also hallucinates a lot less.
For example, when I tried both AI tools for generating visuals of the next four weeks of Pokémon Go, Claude worked better overall. It made one mistake when it placed a date on the wrong day of the week (it said an event was on a Friday, even though the date was a Saturday). But at least the rest of the information was correct.
With Gemini, I asked it to generate an image of the events. On every attempt, it made significant mistakes. The first time around was the most accurate, though there were mistakes about when certain events ended. But every time since, when I asked it to generate a visual, it either completely makes up events or pulls events from last year. Pointing out the mistake doesn’t fix this either; it will just generate another made-up schedule. It eventually gives up and just sends a text list of events, which defeats the point of the entire exercise.
... continue reading