ChatGPT vs. Gemini: Which Chatbot Is Better at Seeing the World Around You?

A recent comparison test by CNET found that Google’s Gemini chatbot outperformed OpenAI’s ChatGPT in several visual recognition tasks, though ChatGPT won in at least one category. The tests involved uploading identical photos to both chatbots with the same prompts, covering handwriting, chess, plant identification, and pop culture art. CNET noted that Google’s Gemini Flash has been upgraded in multimodality, making this a timely revisit of its camera-based abilities against the more popular ChatGPT.

In the first test, Gemini easily transcribed sloppy handwritten notes taken from a page of Arthur Conan Doyle’s The Hound of the Baskervilles, identifying the source text and noting specific handwriting traits. ChatGPT misread the first line, producing a garbled transcription, and guessed the text concerned a memory about a person named Henry. CNET described the difference as night and day, though it acknowledged that Gemini’s correct identification of the book may have aided its translation.

For a chess position, CNET took a photo of a game from the Esports World Cup YouTube channel and asked both bots for the best next move for black. Gemini gave a long, detailed response, while ChatGPT took over four minutes, revealing it was using a local Stockfish chess engine. ChatGPT did offer a secondary best move, Bxf4, which matched Google’s suggestion. CNET also asked Claude, another AI, to judge the responses, and Claude concluded that Gemini misread the board and ChatGPT gave the more appropriate answer.

The third test involved a photo of a Venus flytrap owned by the reporter. Neither bot named the exact cultivar, but both explained why one trap was turning black. Gemini listed several possible reasons, including the correct one: the trap tried to eat an oversized insect. ChatGPT gave a more detailed answer, including a list of suggested actions, which CNET deemed the winner for this round.

The final test used a photo of three hand-painted artworks depicting the mouths of the Sanderson sisters from the movie Hocus Pocus. ChatGPT could not identify the film, accurately describing the image as custom diamond-shaped artwork and asking for more details. Gemini immediately recognized the iconic smiles and went on to describe which character was in which position and the actresses who played them. CNET’s parent company, Ziff Davis, filed a lawsuit against OpenAI in 2025 over alleged copyright infringement, a disclosure included in the original report.

More AI news from TechManNews.