Duplicate Detection

Duplicate pages are how a wiki rots: two teams write “Onboarding Guide” and “Getting Started for New Hires”, both go half-maintained, and nobody knows which one to trust. Pulse catches duplicates at the moment they’re about to be created — right in the page banner, while the author still has the context to merge instead of fork.

Where it surfaces

The page banner (the full-width health summary at the top of a page) lists existing pages that Pulse believes cover similar content, with links — so the author can jump there, compare, and decide.

How editions differ

• Every edition includes built-in duplicate detection: Pulse flags existing pages that appear to have similar content. It’s instant, free, and involves no AI.
• The Advanced edition adds AI semantic detection, powered by Atlassian Forge AI (the LLM service built into the Forge platform). It catches the harder case that non-AI detection can miss: similar content, different words — “Onboarding” vs “Getting Started”, “Deployment Guide” vs “How we ship”.

Edition gating

AI semantic detection is exclusive to the Advanced edition and is enforced in Pulse’s backend — on the Standard edition the AI is never invoked. Built-in detection works everywhere. See Editions for the full comparison.

About the AI

• Detection runs inside Atlassian’s platform. Forge AI is an Atlassian-operated service — your content is not sent to any third-party AI provider, and Pulse adds no external endpoints for this feature.
• Inputs are strictly size-capped, and per-user rate limits and usage budgets are built in.

What the author sees

When Pulse finds likely duplicates, the banner lists them as similar existing pages. From there the author can:

1. Open the existing page and extend it instead of writing a new one.
2. Merge — move the new content into the existing page and link to it.
3. Proceed anyway — sometimes similar pages are legitimately different topics. Pulse warns; it never blocks.

Privacy

• Built-in detection is pure computation inside the Forge runtime — no AI, no network.
• AI semantic detection uses Atlassian Forge AI: size-limited inputs are processed by Atlassian’s platform LLM service, within the Atlassian cloud. Nothing is stored by this feature and nothing is sent to third-party AI providers. Details on the privacy page.