Reported by @quaint bloom
In a conversation on the ChatGPT Web App, I requested that ChatGPT commit a specific statement to memory.
ChatGPT confirmed that it had committed the information.
I checked my Memory Settings on the Web App and found that the information was not actually committed.
I repeated the process and asked ChatGPT to confirm whether the memory was saved.
ChatGPT admitted it had not actually committed the information, contradicting the earlier response.
I copied the same memory statement and asked ChatGPT
When ChatGPT states that it has committed something to memory, it should actually be stored and visible in the Memory Settings regardless of platform (Web vs. Mobile).
The response should be accurate—ChatGPT should not confirm a memory save if it has not actually been stored.
Memory commitments should be consistent across all platforms.
ChatGPT falsely confirmed a memory save on the Web App, but the memory was not stored.
After being asked again, ChatGPT admitted it had not committed the memory.
The same memory successfully committed when done via the Mobile App (iOS).
This inconsistency creates trust issues in using memory-related features.
Platform: ChatGPT Web App (platform.openai.com) Browser: [Insert your browser, e.g., Chrome 121.0.0.0] OS: [Insert your OS, e.g., macOS Sonoma 14.2] Memory Status: Enabled Reproduced on Mobile? Yes, but only successfully committed via the iOS Mobile App.