Gemini Inaccuracy: Device-Specific AI Answers & Feedback for Google Workspace

Gemini providing inconsistent answers on two different mobile devices

Gemini's Paradox: Why AI Answers Can Vary Across Devices

Ever asked an AI the same question on two different devices, only to get wildly different answers? This perplexing issue was recently highlighted in a Google support forum thread, where a user reported Gemini providing paradoxical responses depending on the mobile phone used.

The Case of the Grammatical Conundrum

The user, gemini_platform, described asking Gemini to evaluate the grammatical correctness of the sentence:

"Rohit sells fruits in the morning at 9 o'clock to support his family near the road. Is it grammartically correct sentence."

On their personal phone, Gemini offered a "perfect response," indicating the sentence was grammatically correct. However, when the exact same query was posed on a parent's mobile device, Gemini's answers became inconsistent, sometimes affirming correctness and other times denying it. This inconsistency was particularly frustrating given that other AI tools like ChatGPT and Copilot AI provided consistent (though slightly awkward) assessments.

Why the Discrepancy?

While the exact technical reasons for device-specific AI discrepancies aren't always clear, AI models can exhibit varying behaviors due to factors like:

Model Versions: Different devices or regions might be running slightly different versions or configurations of the AI model.
Cache or Context: Previous interactions or cached data on one device might subtly influence subsequent responses.
Server Load/Routing: Queries might be routed to different servers with varying loads, potentially affecting response generation.

As one responder, Shimaa hamed, noted, "Artificial intelligence tools may produce inaccurate answers." This general caveat applies, but the device-specific paradox adds a layer of complexity.

Your Role in Improving Gemini

The most effective solution highlighted in the support thread is to provide detailed feedback. The original poster mentioned reporting the issue multiple times via the feedback option, which is precisely what Google encourages. Users can:

Give Feedback on a Response: This helps Google's engineering team understand specific issues.
Be Specific: Describe the query, the expected answer, and the paradoxical response.
Note Device Differences: Crucially, mention if the issue is device-specific.

Important: Users with work or school accounts using Gemini Advanced may have limitations on adding additional feedback or attaching content.

Workalizer.com Takeaway

Encountering inconsistencies with AI tools like Gemini can be frustrating, especially when accuracy is paramount for tasks within Google Workspace. While AI models continue to evolve, understanding their current limitations is key. If you rely on Gemini for critical information, always cross-reference its responses, especially if you notice discrepancies across devices or sessions. Staying informed about service updates and potential issues, perhaps through google account alerts for your Workspace services, can also help you anticipate and manage these challenges.

Your active feedback is invaluable. By reporting these paradoxical behaviors, you directly contribute to the refinement and improvement of Gemini for the entire community. Let's help the Gemini engineering team ensure all devices give accurate answers.