Claude Import Memory from ChatGPT: How it Works, Current Limits, and Supported Plans

Claude’s import memory feature gives ChatGPT users a practical path into Claude without rebuilding all their working context from zero.

The feature is real, official, and already available inside Claude.

Its actual structure is narrower than a direct account-to-account transfer, since the process is manual and Claude does not connect live to ChatGPT to copy saved memory automatically.

The handoff works through exported or written-out memory content from the previous AI provider, which is then pasted into Claude’s own import flow.

Claude processes that material, extracts what it identifies as key information, and converts the result into memory edits inside its own system.

This means the route from ChatGPT to Claude is usable, though it is not a one-click sync.

It also means the result depends on both sides of the process.

What ChatGPT makes available, what the user chooses to include, and what Claude ultimately decides to retain all shape the final outcome.

Anthropic also places clear boundaries around reliability and retention, and those limits are central to understanding the feature properly.

Claude may not always fully incorporate imported memory.

Claude also focuses its memory system on work-related usefulness, which means some personal details that exist in ChatGPT memory may not survive the transfer in the same way.

Once those limits are kept in view, the feature becomes much easier to judge in practical terms.

··········

How Claude’s import memory feature works when the source is ChatGPT.

The ChatGPT-to-Claude route is a manual migration flow in which exported or written-out ChatGPT memory context is pasted into Claude and converted into Claude memory edits.

The feature begins on the source side, not inside Claude alone.

ChatGPT already has its own official Memory system, which OpenAI describes through saved memories and chat history.

That means there is real persistent context on the ChatGPT side that can become the basis for migration.

Anthropic’s Claude-side import feature is built to receive memory from another AI provider through pasted exported content.

In practice, this means the ChatGPT-to-Claude route is not a hidden connector running in the background.

It is a user-driven handoff.

The user first obtains memory or memory-like context from ChatGPT, then opens Claude’s import flow and pastes that material into Claude.

Claude then extracts key information and stores it as individual memory edits inside its own memory system.

That design is important.

Claude is not copying ChatGPT’s internal memory architecture.

Claude is receiving text that represents prior memory and transforming it into Claude-native memory records.

This is why the feature is best understood as a migration layer rather than as a live sync layer.

It transfers usable context.

It does not transfer the full hidden structure of ChatGPT’s memory system.

........

· The route from ChatGPT to Claude is manual rather than automatic.

· Claude accepts pasted memory content from another AI provider and converts it into its own memory edits.

· The imported result is a Claude-native memory layer, not a copied ChatGPT memory database.

........

Execution flow from ChatGPT to Claude

Stage	What happens
Source memory exists	ChatGPT stores saved memories and chat-history-based context
User gathers source content	The user obtains memory or memory-like context from ChatGPT
Claude import starts	The user opens Claude’s import tool
Paste and process	The user pastes the source material into Claude
Final result	Claude extracts key information and stores it as memory edits

··········

Which Claude plans and surfaces support importing memory from ChatGPT.

Anthropic has made the import feature broadly available at the plan level, though the confirmed import surfaces are narrower than Claude’s full device footprint.

Anthropic states that memory import is available to all Claude users.

That includes free, Pro, Max, Team, and Enterprise plans.

This wide availability changes the practical value of the feature immediately.

A user does not need to move onto a premium Claude tier just to test whether ChatGPT-derived memory can be brought over.

The same broad availability also makes the feature relevant for team and enterprise migration scenarios, especially when users or organizations are testing Claude after prior use of another assistant.

The supported import surfaces are clearly named by Anthropic.

Memory import is confirmed on the web and Claude Desktop.

That is the cleanest statement of surface availability in the reviewed official material.

The broader Claude memory system exists more widely across the product, and Anthropic’s release notes discuss memory expansion across plans, though the import article itself specifically confirms web and Desktop.

That difference should remain visible.

The feature is broadly available by plan, while the confirmed import entry points remain tied to those two named surfaces.

........

· Claude memory import is available on free, Pro, Max, Team, and Enterprise plans.

· The confirmed import surfaces are the Claude web app and Claude Desktop.

· The reviewed official sources do not clearly confirm the same import flow on mobile.

........

Supported plans and confirmed surfaces

Area	Confirmed status
Free	Supported
Pro	Supported
Max	Supported
Team	Supported
Enterprise	Supported
Web	Supported
Claude Desktop	Supported
Mobile import flow	Not clearly confirmed in reviewed sources

··········

How the source material from ChatGPT is supposed to be prepared before import.

Anthropic’s documented approach relies on extracting memory-like context from the previous AI provider in readable form rather than on using a dedicated structured transfer protocol.

Anthropic provides a recommended prompt for the previous AI provider.

That prompt asks the earlier assistant to list stored memory and learned context across areas such as response instructions, personal details, projects, goals, tools, languages, frameworks, and preferences or corrections, and to return the result in a single code block.

This is the most important practical clue for the ChatGPT side of the workflow.

Anthropic is not requiring a specialized memory file format from ChatGPT.

It is asking for readable exported or written-out context that Claude can then interpret.

That makes the route from ChatGPT to Claude flexible, though it also removes the precision that a tightly structured memory-transfer protocol would have offered.

OpenAI also officially supports ChatGPT account data export through Settings and Data Controls.

That confirms that users can export account data from ChatGPT, though the reviewed OpenAI sources do not document a dedicated “export saved memories only for Claude import” feature.

This distinction matters.

The Claude-side workflow is clearly documented.

The ChatGPT-side memory extraction path is real, though it is not presented by OpenAI in the reviewed material as a specialized Claude-transfer flow.

So the most defensible framing is that ChatGPT memory can be turned into source material for Claude import, though this happens through manual extraction or export-style preparation, not through a first-class OpenAI-to-Anthropic handoff protocol.

........

· Anthropic provides a recommended prompt for the previous AI provider to write out memory and learned context.

· The source material from ChatGPT is expected to be readable pasted content rather than a formal direct-transfer file.

· OpenAI supports ChatGPT data export, though the reviewed sources do not document a dedicated saved-memory-only export path for Claude import.

........

Source preparation before Claude import

Source-side element	Confirmed position
Anthropic-recommended source method	Ask the previous AI provider to write out stored memory and learned context
Suggested categories	Instructions, personal details, projects, goals, tools, languages, frameworks, preferences, corrections
Required transfer style	Pasted readable content
Official ChatGPT memory existence	Confirmed
Official ChatGPT account data export	Confirmed
Dedicated OpenAI export-to-Claude flow	Not documented in reviewed sources

··········

How Claude converts imported ChatGPT context into its own memory system.

Claude does not preserve the pasted material as one raw imported block and instead extracts key information into separate memory edits that become part of Claude’s own personalization layer.

After the user pastes the ChatGPT-derived memory content into Claude’s import interface and selects Add to memory, Claude extracts what it considers key information.

Anthropic states that the imported result is stored as individual memory edits.

This is one of the most revealing parts of the product design.

Claude is not trying to preserve the external memory source in untouched form.

It is rewriting that source into Claude-native memory objects.

That makes the migration usable inside Claude’s own broader memory system, though it also means the import outcome is interpretive rather than literal.

Anthropic also says users can inspect the imported result through Manage edits.

That review layer is especially important in a feature built around extraction.

The user can see what Claude actually kept, not just what the user attempted to bring over from ChatGPT.

This is where the real quality of the migration becomes visible.

The pasted content may be broad.

The final Claude memory may be narrower, cleaner, more work-oriented, or partially incomplete depending on what Claude chooses to synthesize into edits.

........

· Claude rewrites imported memory into its own internal memory-edit structure.

· The imported result is interpretive, not a raw copy of the pasted source.

· Manage edits is the built-in review layer for checking what Claude actually retained.

........

How Claude stores the imported result

Area	Confirmed behavior
Input form	Pasted exported or written-out memory content
Claude processing	Extraction of key information
Storage format	Individual memory edits
Review mechanism	Manage edits
Product implication	Claude-native memory synthesis rather than raw source preservation

··········

What Claude is more likely to keep from ChatGPT memory and what may be left out.

Anthropic’s documented memory design favors work-related context, which means Claude may keep projects, tools, and collaboration-relevant preferences more reliably than unrelated personal details.

Anthropic states that Claude’s memory is designed to focus on work-related topics.

That statement defines the retention logic more clearly than any abstract product description could.

If ChatGPT memory contains project context, goals, workflow preferences, tools, languages, frameworks, or other collaboration-relevant information, that material is more aligned with how Claude is meant to use memory.

Anthropic also states that Claude may not retain personal details unrelated to work.

This is a direct scope boundary.

It means a user moving from ChatGPT to Claude should not assume that every saved preference, fact, or personal note stored in ChatGPT memory will survive the Claude-side filtering process.

The transfer is selective by design.

Anthropic also provides a fallback route when a specific item matters and does not survive import.

Users are directed to Settings > Capabilities > View and edit your memory if they want to add or refine memory entries manually.

That confirms the intended working model.

The automated import handles the first pass.

Manual editing handles the missing or high-priority details that need to be pinned more deliberately after Claude has finished its own extraction pass.

........

· Work-related context is more aligned with Claude’s memory design than broad personal detail.

· Personal details unrelated to work may not be retained after import.

· Manual memory editing exists as the fallback when important information from ChatGPT does not survive the first pass.

........

Retention logic after ChatGPT-to-Claude import

Type of source information	Likely Claude treatment
Project context	Strong fit
Goals and work preferences	Strong fit
Tools, languages, and frameworks	Strong fit
General collaboration instructions	Strong fit
Personal details unrelated to work	May not be retained
Missing critical information	Can be added manually later

··········

Why importing memory from ChatGPT to Claude is useful without being a perfect migration.

The feature reduces switching friction in a practical way, though Anthropic explicitly stops short of promising full-fidelity memory transfer from the source assistant.

There is clear value in the feature even before considering its limits.

A user moving from ChatGPT to Claude can avoid retyping a large amount of working context, preferred response behavior, project history, and recurring operational details.

That alone can shorten the setup phase significantly.

The limits are just as important.

Anthropic explicitly states that memory imports are experimental and still in active development.

Anthropic also states that Claude may not always successfully incorporate imported memories.

This is not vague caution.

It is a direct product boundary.

The imported result should therefore be treated as a useful first transfer, not as a deterministic memory clone.

That distinction becomes even more important in the ChatGPT-to-Claude route, since the source material itself is not described in the reviewed OpenAI sources as a dedicated structured Claude-ready memory export.

There are at least three filters in the chain.

What ChatGPT stores or surfaces, what the user chooses to export or write out, and what Claude chooses to absorb all affect the outcome.

So the feature is valuable for migration.

It is not a guarantee that Claude will behave as if it inherited ChatGPT’s full internal memory state.

........

· The feature is useful as a first-pass migration tool.

· Anthropic explicitly says imported memory may not always be fully incorporated.

· The final result is shaped by the source content, the user’s choices, and Claude’s own extraction logic.

........

What the feature does and does not guarantee

Area	Position
Reduces migration friction	Yes
Avoids full manual rebuild	Partly
Direct provider-to-provider fidelity	Not guaranteed
Full reproduction of ChatGPT memory	Not guaranteed
Development status	Experimental
Reliable first-pass transfer value	Yes

··········

How long the imported ChatGPT memory takes to appear and how it is governed afterward.

Anthropic gives a concrete timing window for imported memory to appear, and once it is stored it falls under Claude’s standard memory controls and, in enterprise environments, under organization-level governance.

Anthropic states that imported memory should appear within 24 hours.

That timing matters because the feature should not be treated as an instant runtime rewrite where Claude immediately reflects every imported detail in the next message.

The post-import result needs to be understood as a memory update inside Claude’s broader personalization system.

Once the imported information has been incorporated, it becomes part of Claude’s larger memory framework.

Memory and past-chat search can be toggled on or off through Settings > Capabilities.

In Enterprise environments, Anthropic adds organization-level controls.

Owners and Primary Owners can manage memory features organization-wide.

Anthropic also states that if memory is disabled at the organization level, existing memory synthesis data is deleted for that organization.

That means imported ChatGPT-derived memory is not living outside normal Claude governance.

It is absorbed into the same control structure that governs the rest of Claude’s memory behavior.

........

· Imported memory should appear within 24 hours.

· Once incorporated, it falls under Claude’s normal memory and past-chat controls.

· Enterprise governance can override user-level behavior and can delete organization memory synthesis data if memory is disabled.

........

Timing and governance after import

Area	Confirmed position
Post-import timing	Within 24 hours
User-level controls	Memory and past-chat search can be toggled on or off
Enterprise controls	Owners and Primary Owners can manage memory org-wide
Org-level disable effect	Existing memory synthesis data is deleted
Imported memory status	Becomes part of Claude’s standard memory system

··········

What the current hard limits are in the ChatGPT-to-Claude memory route.

The current route is broad in availability, though limited in determinism, surface certainty, and technical transparency around the exact import envelope.

The first hard limit is reliability.

Anthropic clearly states that imports are experimental and may not always be fully incorporated.

The second limit is retention scope.

Claude may leave out personal details unrelated to work.

The third limit is transport design.

The reviewed official sources do not document a dedicated one-click OpenAI export path specifically built for moving saved ChatGPT memory into Claude.

The fourth limit is technical specification.

Anthropic does not publish, in the reviewed source set, a formal public import schema, a maximum paste size, or a deterministic parsing specification for imported memory text.

The fifth limit is surface certainty.

Web and Claude Desktop are clearly confirmed for imports, while the same flow is not clearly confirmed on mobile in the reviewed official material.

Taken together, these limits define the product accurately.

The route from ChatGPT to Claude is real, useful, and already available.

It is also manual, selective, and not engineered as a perfectly standardized memory-migration protocol.

........

· The feature is real and broadly available, though not deterministic.

· OpenAI’s reviewed sources do not document a dedicated one-click ChatGPT-memory-to-Claude export flow.

· Anthropic’s reviewed sources do not publish a formal public schema or maximum import envelope for pasted memory text.

· The process remains manual, selective, and experimental.

........

Current limits of importing memory from ChatGPT to Claude

Limit area	Current position
Import reliability	Not guaranteed
Retention completeness	Selective
One-click ChatGPT-to-Claude transfer	Not documented
Formal public import schema	Not clearly published
Confirmed import surfaces	Web and Claude Desktop
Mobile import confirmation	Not clearly confirmed
Development state	Experimental

·····

FOLLOW US FOR MORE.

·····

·····

DATA STUDIOS

·····

[datastudios.org]