Confluence space exports omit blog posts from PDF without raising an error

Only content visible to the person clicking export is included in the file ^[8]. A restricted page that a space admin cannot read does not appear in the export that admin produces. Mikhail Klimenko, writing on dev.to, wrote: "The same button in two pairs of hands produces two different files, which is a problem when the export is supposed to be a complete record for legal or compliance reasons" ^[9][20].

Drafts are not in the export at all. An export takes published content, so an edit somebody has been carrying for three weeks does not exist as far as the export is concerned, and nothing on screen points that out ^[6].

The post maps destinations to formats: XML into another Confluence site, HTML for an archive a human can open later, PDF for sending outside, and it costs nothing ^[15]. Word never appears in the space-level list, because in Cloud it is a page-level action ^[2]. Klimenko wrote that anyone who tells you to just export the space to Word is describing Data Center or wishful thinking ^[18]. The 50-image ceiling people repeat for Word export is documented for Confluence Data Center, where a system property can raise it; Cloud has no such property, and the Cloud documentation does not state that number. He advises testing your own pages before planning around it either way ^[16].

The CSV option wastes the most time, because the name promises table data. It dumps the space's content, pages, attachments and comments, and the post says Atlassian's documentation does not describe its column layout ^[10]. Klimenko wrote that "the native export does not do this, and the CSV name is a trap" ^[11]. Copying tables by hand holds up until a merged cell appears, and then the values after it shift into the wrong columns while still looking like perfectly good numbers ^[12].

His cost figure is a measurement of his own reporting job. Eleven tables spread across a space, about four minutes each when nothing is merged, closer to ten when something is, which he rounds to an hour and a half, four times a year ^[13]. Eleven clean tables at four minutes is 44 minutes; eleven merged ones at ten is 110 ^[1]. The ninety-minute estimate therefore assumes merged cells in most of them, and it transfers to your team only if a report pulls from a similar number of tables and somebody spends the extra time reconciling the wiki figure against finance. Klimenko discloses that he builds Confluence apps, including one that exports tables to Excel or CSV with merged cells expanded, and writes that it does not replace the space export ^[17].