Lead
On 2026-03-26 Google began rolling out a new import tool for Gemini that lets users bring in chat logs and memory data exported from other AI platforms. The feature appears alongside the Gemini 3.1 Flash Live update and offers a web path at gemini.google.com/import to upload exported .zip files. Users can import conversations from named services such as ChatGPT and Claude, with upload limits and daily quotas enforced. The rollout excludes the EEA, Switzerland and the United Kingdom at launch.
Key Takeaways
- Import endpoint: gemini.google.com/import provides two choices — “Import chats” and an “Import memory” route for past assistant data.
- File limits: Gemini accepts .zip export files up to 5 GB each and up to five .zip uploads per day.
- Platform examples: Google explicitly names ChatGPT and Claude as supported export sources for chat imports.
- Visibility: Imported conversations appear in Gemini’s side panel and carry an import-chat icon for identification.
- Deletion and overwrite: Users can delete an entire import entry to remove all chats from that .zip; reuploading the same .zip merges new conversations and overwrites previously imported ones.
- Export steps: The article documents the basic export flows for ChatGPT and Claude, including where to request an export within each app’s settings.
- Geographic limits: Importing is not available in the EEA, Switzerland, or the United Kingdom at the time of the report.
Background
Cross-platform portability of AI chat histories has become a frequent user request as multiple assistant products proliferate. Users who switch between models or run parallel assistants have sought ways to preserve context, preferences and past instructions so that new models can continue ongoing workflows. Industry vendors have gradually added export and import tooling to address data portability and retention concerns while balancing privacy and safety requirements.
Google’s Gemini family has expanded aggressively since the initial releases, adding speed-optimized builds and larger-context capabilities. The import feature follows a broader trend in which providers expose export APIs or UI-driven data-download options; the two named sources in this rollout, ChatGPT and Claude, already offer export workflows that generate downloadable archives. For enterprises and privacy-minded users, the availability of direct import and clear deletion controls matters for lifecycle management of sensitive conversation data.
Main Event
The new import capability is surfaced through Gemini’s settings. If the prompt to import does not appear automatically, users are directed to the “Settings & help” gear in the app’s bottom-left corner where an “Import memory to Gemini” option is listed. Selecting the UI path routes the user to gemini.google.com/import, which displays the two import mechanisms: uploading exported chat .zip files and pasting a structured memory export.
For chat imports, users must first generate an export from the source assistant. The article outlines the ChatGPT flow: open the bottom-left username menu, choose Settings > Data controls, click Export data, and confirm Export. For Claude, the steps are similarly described: username menu > Settings > Privacy, then Export and choose the data range. The external platform then sends a download link to the account email used on that service.
Once a .zip file is uploaded to Gemini, its conversations appear in the side panel with a special import-chat icon. Imported conversations are searchable and can be deleted individually or as a group. To remove all conversations from a particular import, the user finds the import entry and clicks Delete next to it; that action removes all chats associated with that .zip.
Reupload behavior is explicit: reuploading a previously imported .zip will add any new conversations present in the archive and overwrite previously imported items where duplicates occur. The other import path, “Import memory,” requires running a prompt within the source assistant and pasting the resulting payload into Gemini; the article reproduces the instruction pattern Google recommends for that workflow.
Analysis & Implications
Allowing users to import chats and memory reduces friction for switching assistants and strengthens Gemini’s position as a hub for personal AI history. Practical benefits include preserving long-running project context, retaining preference signals, and reducing repeated setup steps across models. For productivity users, seamless continuity of memory can significantly shorten the time required to resume complex tasks after migrating platforms.
Privacy and compliance will be central considerations. Imported archives can contain sensitive personal or business content; the feature therefore requires transparent controls for deletion and clear documentation of retention policies. Google’s decision to exclude the EEA, Switzerland and the UK at launch suggests unresolved regulatory or data-transfer considerations that will need explicit handling before a broader rollout.
From a product-competition perspective, promoting open import paths increases user agency and may pressure other vendors to match portability features. The explicit mention of ChatGPT and Claude signals interoperability intent, but the practical utility will depend on fidelity of exported data, format compatibility, and whether semantic memory elements (preferences, labels, role prompts) map cleanly between systems.
Operationally, the 5 GB per-file cap and five .zip-per-day quota set practical bounds on migration velocity. Users migrating very large histories or enterprise-scale archives will need staged uploads or tailored tooling; Google may expand limits or provide enterprise migration aids if uptake proves high.
Comparison & Data
| Item | Value |
|---|---|
| Maximum .zip size | 5 GB |
| Daily .zip uploads | Up to 5 files per day |
| Unavailable regions at launch | EEA, Switzerland, United Kingdom |
| Named source platforms | ChatGPT, Claude |
The table clarifies the principal numeric constraints and named sources described in the rollout. These bounds matter for planning: a single 50 GB export, for example, would require splitting into multiple archives across several days, and region exclusions will prevent immediate global adoption for some users.
Reactions & Quotes
Google presented the UI affordance and web import route as the supported method for migration. The following short excerpts capture the wording used in the interface and reporting of the feature.
“Import memory to Gemini”
Google (interface label)
The import option is visible under Settings & help in the app UI. That label is the first step for users who do not see an automatic prompt; the article documents how to reach the import page manually.
“users able to upload up to 5 .zip files per day”
9to5Google (news report)
9to5Google reported the per-day upload quota and the 5 GB-per-file cap. These limits are operational details users must consider when planning migrations or bulk transfers.
“ChatGPT and Claude”
9to5Google (news report)
The report names ChatGPT and Claude as examples of supported export sources. The article supplies step-by-step export guidance for those two platforms to help users generate compatible .zip files for upload.
Unconfirmed
- No public confirmation that imported archives are processed with end-to-end encryption while in transit and at rest on Google’s servers.
- Unclear whether all metadata fields (attachments, timestamps, system prompts) from every source export are preserved and mapped perfectly into Gemini.
- No timeline provided for lifting the EEA, Switzerland, and UK restrictions or for expanding supported source platforms beyond those named.
Bottom Line
Google’s import capability for Gemini addresses a common user need: preserving conversational context when moving between AI assistants. The design balances usability—simple web uploads and a memory-paste flow—with conservative operational limits (5 GB files, five .zips/day) that reduce abuse risk and encourage staged migrations.
Adoption and trust will hinge on data-handling transparency and regulatory clearance in restricted regions. For power users and teams managing multi-model workflows, the feature removes a major friction point; for privacy-conscious users and regulated organizations, further documentation and enterprise-grade migration tools will be needed before large-scale use.