Confluence to Notion migration: what you lose

What actually transfers from Confluence to Notion?#

A Confluence to Notion migration transfers page content well and page context poorly. Notion's importer and third-party migration tools focus on titles, body text, page trees, tables, images and many attachments; the layer that explains how a page came to say what it says, meaning versions, comments, restrictions and authorship, is where most loss happens.

Results vary by tool, by Confluence edition and by how heavily your teams used macros, so test one representative space before planning the rest. The table shows what to look for in that test.

Why version history and comments matter more than they look#

Version history and comments matter because they hold the reasoning that the final page leaves out. An architecture decision record shows the chosen option; its history and inline comments show the alternatives that were argued down, who raised the risk that changed the design and when the team changed its mind.

That context has practical uses after the migration. Engineers revisit old decisions during incidents, auditors ask when a policy page changed, and new leaders need to know why a system works the way it does. The same decision trails, linked to Jira issues and pull requests, are also what AI developers value in a software company's internal documentation.

The archive-first rule#

The archive-first rule says no Confluence space is imported, trimmed or cancelled until a complete export of it sits in storage the company controls. An import is a copy into a different shape; an archive is the record you fall back on when the copy turns out to be missing something.

Confluence Cloud gives you two export routes. Space admins can export a whole space, or selected pages, to PDF, CSV, HTML or XML from Space settings. Site admins can create a site backup in the Backup manager under Data management; Atlassian says it includes attachments only if you select them, skips spaces sitting in the trash, keeps only one backup file at a time and offers the download link for 14 days. Many teams keep a human-readable HTML export for browsing and a structured XML export or site backup for completeness.

• 1Step 1: inventory every space with its owner, last activity date and approximate page count, and restore any space from the trash that you still need.
• 2Step 2: export each space as HTML for reading and as XML for structure, or take a site backup with attachments selected.
• 3Step 3: download every backup within its 14-day link window and copy it out of Atlassian's storage, since a new backup replaces the old one.
• 4Step 4: for decision-heavy spaces, confirm the export includes page history and comments, or capture them separately.
• 5Step 5: record how pages link to Jira projects and keys, since those macros may not survive the import.
• 6Step 6: store the archive with checksums, an owner and an access list, and test that a sample of pages opens without Confluence.
• 7Step 7: only then run the Notion import and, later, end the Confluence subscription.

Which spaces need a live Notion copy?#

Only spaces that people still edit or consult every week need a live Notion copy; most others can live in the archive. Moving everything recreates the clutter that often motivated the switch, and it buries current pages under years of stale ones.

Sort spaces before importing, and treat sensitive and customer-specific spaces as their own category.

Illustrative: an agency software company moves its wiki#

Illustrative: a fictional agency management software company for insurance brokers ran Confluence Cloud with spaces for engineering, product, support and HR. The CTO planned a full move to Notion and asked an engineer to run a test import of the engineering space first.

The test showed Jira issue macros turned into plain links, inline comments missing from architecture decision records and every page attributed to the engineer who ran the import. The CTO changed the plan: archive all spaces with history first, move only active runbooks, handbooks and current design docs, and keep the decision records read-only in the archive. Notion started clean, and the reasoning behind past architecture choices stayed intact and searchable.

Mistakes that make the loss permanent#

The loss becomes permanent when Confluence is cancelled before anyone checks the archive. Import tools report success when pages arrive, not when history, comments and attachments arrive, so a clean import log says little about what was dropped.

Other common mistakes include exporting only PDFs, which keep the look of a page but lose its structure and comments, and forgetting attachments stored at space level rather than on pages. Losing the mapping between old user accounts and real people is another, because pages written by former employees become anonymous once their accounts are gone.

How SourceX looks at internal documentation archives#

SourceX looks at Confluence archives as product decision records: design docs, decision records and postmortems that connect to Jira issues and code. Under the SourceX Enterprise Data Value Framework, drivers such as domain expertise, human-generated signal and uniqueness raise their value, while privacy burden and preparation cost, for example separating HR pages, reduce net value.

A first conversation needs only a list of spaces, their purpose and the years they cover. Spaces then move through the SourceX five-step transaction, Supply, Rights, Preparation, Approval and Delivery, one package at a time, so a company can include its engineering decision records while keeping HR and finance spaces out entirely.