subhead: AI Mode has a Server-Side Transcript Invisibility bug- potential deadlock (developer to look at User ID and perform a manual DB-integrity check needed)
product affected: Google App / AI Mode
platform: Android, plus desktop web test results
date range affected: Feb 22 to Mar 3 aprox, 2026
description: Following a recent system-wide sync error in AI Mode.
While my account history has undergone a partial restore including legacy data from 2019 only that most critical session (from late Feb to early March 2026) was still missing.
exact behavior:
initial prompt/metadata visible in My Activity with the wrong timestamp! opening entry shows blank chat with the inital prompt or regenerates again the initial prompt.
Takeout contains metadata/initial prompt but transcript is null.
generaly, i have observed routing/indexing persistent bug. i suspect a database synchronization error between the local AI Mode cache of my User ID and the cloud storage/server logs, causing a complete wipe of the history.
The Smoking Gun: I have direct URLs to random missing threads (Metadata exists), but the server returns a “URL not valid” error. The Pointer exists but the Data Blob is unreachable (HTTP 404/500 style errors for existing metadata pointers).
steps already attempted:
multiple devices
incognito
web interface
manual sync
cache reset / Labs toggle
sending follow-up prompts inside corrupted threads
I’ve already sent logs, User IDs and exact timestamps- via the app (feedback)
help community told me this issue is beyond their ability and suggested the forum
title explenation:
i have mentioned deep indexing/sync deadlock because i thought based on collected specolations
-
System Instability/Routing Errors: I have observed persistent routing errors (‘Thread not found’) and corrupted AI responses in other restored threads (“here are some articles..”)
-
Active Session: This conversation was active and saved on my device for weeks, surviving reboots and refreshes, until it disappeared This specific thread was open and active on my mobile device during the synchronization failure.
For all the above- I suspect it is stuck in a “Deadlock” or “In-Progress” state on the server, preventing it from appearing in the restored history.
but as i said its only a speculation.
Is there any way to escalate this to the engineering team (is this engineering team) for a manual database re-indexing of ‘Active Sessions’? i am prepared to provide User ID (broken links) and my Account metadata privately to an authorized Google employee upon request.
If no one could do that, at least i tried my best.
At the begining i was so desparate i suggsted via feedback a specific rollback knowing it would never take place, only to know i did everthing i could.
please help me. This thread represents weeks of immense professional work.
Thank you, please help