Since Microsoft launched Copilot to the public in 2023, I have used it regularly both at work and for personal projects. I have written guides to optimize its use, analyzed how to mitigate hallucinations, and integrated it with Microsoft 365. However, after multiple frustrating experiences, I have decided to abandon it for personal tasks. Google Gemini has proven to be remarkably superior in accuracy and efficiency. Below, I detail the specific cases that led me to this decision.

Copilot is ineffective for solving technical problems

Like many people with a technical profile, I often serve as the IT support for my close circle. Copilot has helped me on occasion, but the turning point came when I tried to fix a problem with my wife's new iPhone: she was receiving text messages in her email but not in the Messages app. Copilot confidently stated that there were only “two real explanations” and guided me through several settings. None worked. Then it promised a “final solution” that also didn't solve anything. After more than an hour of failed attempts, I discovered that Copilot was using information from an old version of iOS. When I pointed this out, it apologized and suggested calling the carrier. That's when I turned to Gemini: in 30 seconds it diagnosed the problem and offered a simple solution that required no phone call. It worked immediately.

Copilot fails in personal research

In a project about Parisian neighborhoods in the 1870s, I asked Copilot about the area near the Saint-Lazare station. It claimed it was dangerous and poor, with precarious housing stained by coal smoke. This contradicted my memory of the famous painting “Paris Street; Rainy Day” by Gustave Caillebotte, which shows an elegant and bourgeois area. I asked Gemini and Claude: both agreed it was an expensive and sought-after neighborhood, which I confirmed with my own sources. Copilot was wrong again.

Copilot gives bad advice on schedules

I swim regularly at my gym's pool, but after a temporary closure, I started using a nearby elementary school pool. I asked Copilot what time of day had the least traffic on weekdays. With its characteristic confidence, it answered between 11:30 AM and 1:00 PM. I didn't know the pool doesn't open to the public until 3:00 PM. Gemini got it right: 3:00 PM is the least crowded time. I confirmed this with the lifeguards. Copilot not only gave incorrect information but also demonstrated a lack of updated data.

Goodbye, Copilot

These incidents, along with others like them, have led me to abandon Copilot for research and personal advice tasks. Trust is key in any AI tool, and Copilot has completely lost it. Gemini, on the other hand, has shown accuracy, speed, and superior data freshness. For IT professionals, this experience underscores the importance of verifying sources and not blindly relying on assistants that may offer outdated or erroneous information. As we mentioned in our article on Loops replace prompts: verification becomes your biggest challenge, constant validation is essential. Likewise, in corporate environments, choosing the right AI assistant can directly impact productivity and security, as we saw in Advanced Solutions in Microsoft Azure.

Original source: ComputerWorld. Analysis and adaptation by ForgeNEX.