Atlassian: Guidelines for writing with AI

In June 2026, Atlassian published writing guidelines developed by its Brand team and authored by Liz Fosslien, who leads communications there. The document emerged from a specific concern: AI writing tools, used without structure, produce content that erodes reader trust—not only through recognizable AI phrasing patterns but through a more fundamental flattening of voice, perspective, and accountability.

The guidelines apply to Atlassian’s marketing and content teams rather than to all company writing. They are explicitly framed as an “ongoing journey” rather than a fixed policy, with a commitment to quarterly review as AI capabilities change.

The six principles

Humans lead; AI assists. People must write the first and final drafts. AI is available for validation, editing, and refinement between those stages, but the core argument must originate with a person.

Every published piece carries a human byline who is publicly accountable for the content. This treats authorship as a professional commitment, not a technical formality.

Context beats generic prompts. When using AI to edit or refine, Atlassian requires specifying the text’s audience, purpose, and tone. Generic prompts produce generic output; specific context produces output that can be shaped into something worth publishing.

Verify all factual claims before publishing. AI generates plausible-sounding statements that may not be accurate. Verification remains a human responsibility regardless of how the draft was produced.

Publish nothing you wouldn’t read yourself. A useful internal check that addresses one of the more common AI content failure modes: text that is formally correct but empty or padded.

Guard against AI erasure. AI systems trained on large text corpora can reduce or smooth over perspectives from underrepresented groups when generating text. The guidelines ask writers to actively check for this.

Prohibited uses

The document explicitly lists what AI should not do in Atlassian’s content process: generating statistics (which the guidelines treat as requiring verification from primary sources), inventing claims, or determining the core argument of a piece. These prohibitions address the most common ways AI use introduces factual errors or intellectual shortcuts.

Who this is useful for

The guidelines are most relevant for content teams at companies with established brand voices who are navigating the same internal conversation: how to allow AI into the writing workflow without losing the qualities that built reader trust. The framing is practical rather than ideological, which makes the document easier to adapt and introduce internally than materials that treat AI adoption as either a mandate or a threat.