Stay up to date with all the latest updates, improvements, and new features we ship to help you do better research. We’re constantly improving ChatAcademia for academics and researchers.
Some chats with older history left over from interrupted conversations were failing to send a new message with an internal validation error. Those empty placeholder messages are now skipped automatically, so you can pick up where you left off without having to start a new chat.
Improvements
Resuming a chat no longer fails with a 'message must contain at least one part' error caused by gaps in older conversation history
Fewer Chat Timeouts & Clearer File Errors
Fewer Chat Timeouts & Clearer File Errors
Long chats on the pro model now recover automatically when one provider is slow, and any time we can't read your file we'll tell you exactly why — whether it's password-protected, too large, or has too many pages — so you know what to fix.
Improvements
Chats no longer fail with 'The operation timed out' when the pro model takes a long time to respond — they now automatically fall back to a backup model and continue
File reading errors now explain the actual problem (e.g. 'Document is password-protected — remove the password and try again', 'File is too large (47 MB; max 20 MB)') instead of a generic 'Bad Request' message
Academic database and grant searches surface specific, actionable errors when a provider rate-limits us, including when to retry
Chats with files attached from Zotero, Mendeley, Google Drive, or saved Notes now skip redundant downloads on every turn, so follow-up messages feel snappier. Long conversations on the fast model also recover more cleanly when something goes wrong mid-stream, and every paper analyzed by the Research Verdict agent now has a clickable title that opens the source.
Improvements
Files attached from Zotero, Mendeley, Google Drive, or saved Notes are reused across turns instead of being re-fetched from the source on every message — noticeably faster follow-ups in chats with attachments
Long-running conversations no longer fail intermittently with 'Failed after 3 attempts' errors caused by older history that some AI providers can't process
Every paper in the Research Verdict 'Papers Analyzed' tab now has a clickable title that opens the original source — previously some papers appeared as plain text
Send Feedback Instantly & Track Every Chat's Credit Usage
Send Feedback Instantly & Track Every Chat's Credit Usage
A new Feedback button in the page header lets you send feature ideas, bug reports, or questions in seconds — and every chat now tells you exactly how many credits it has consumed, with a friendly warning when a conversation is getting expensive.
Improvements
Click the Feedback button next to What's new on any page to open a quick form
Pick a category — Feature, Bug, Improve, Question, or Other — so the team can route your note to the right person
Optionally attach a PNG, JPEG, WEBP, or GIF screenshot up to 5 MB to give visual context
Live character counter and minimum-length helper make sure your message is detailed enough to act on
Confirmation toast tells you we received it, and the team is notified by email right away
See how many credits each chat has consumed right in the context window popover in the chat header
A friendly warning appears when a chat crosses 2,000 credits, explaining that long conversations cost more per message
One-click Create new chat button in the warning opens a fresh conversation in the same agent, so you can keep working without losing pace
Fixed AI chat tools that search academic databases (Semantic Scholar, ArXiv, PubMed, Google Scholar, ERIC, Europe PMC, OpenAlex, Crossref) failing with an internal error
More Reliable Reading Assistant & Friendlier Chat Navigation
More Reliable Reading Assistant & Friendlier Chat Navigation
Reading Assistant now handles a wider range of PDFs that previously failed during upload, and opening a missing or deleted chat takes you home with a helpful message instead of a broken error screen.
Improvements
Reading Assistant accepts PDFs with unusually dense or unsplittable sections that previously caused creation to fail
Opening a deleted or unavailable /chat/:id link now redirects home with a clear toast instead of showing an error retry prompt
Microphone speech-to-text no longer surfaces noisy error popups for transient network or no-speech events