Conflicting information
Claims that disagree within a page or across related pages.
Evergreen AI for Confluence · Available on Marketplace
Find the Confluence pages that need a second look.
Help your team resolve conflicting instructions, expired information, and outdated references. Evergreen AI flags potential issues with quoted evidence and a confidence score. You decide what needs to change.
For Confluence admins and the teams responsible for runbooks and policies.
From conflicting pages to an owner’s decision
Check the evidence. Resolve the finding.
These travel-policy excerpts come from our test fixtures. They illustrate a review workflow, not customer data or a recorded result.
Two pages disagree
Expense limits — travel lists meals at
$65/day
. Travel policy FAQ saysMeals are reimbursed up to $90/day per traveler.
A finding brings the evidence together
For a detected contradiction, Evergreen AI shows quotes from both pages, the reason for the finding, and a confidence score. Open the source pages to check their context.
The owner decides what to fix
Confirm and assign a valid finding, or dismiss it with a reason. The owner edits the source content in Confluence, then marks the finding resolved in Evergreen AI.
Watch a scan and owner handoff
This 86-second walkthrough follows an expired rollout approval from its Confluence page through the scan to an owner’s queue. It uses genuine product captures with demonstration data.
Read the walkthrough
Let’s look at a page in Confluence. This is a demo space, and the page is a rollout approval. It looks ready to use, but there’s a date we should check.
Now we’re in Evergreen. Choose the space you want to scan, then start the scan. We’ll skip the waiting time here.
In Findings, Evergreen has flagged that same approval. Let’s take a closer look. The quoted passage says it’s valid until the thirtieth of September. That date had already passed when this scan ran.
The finding includes a reason and a confidence score. Those help you review it, but you still need to check the source page and decide what to change.
If someone else owns the content, you can assign the finding to them. Here, we’re choosing a demo account.
Now we’ve switched to that person’s view. In Assigned to me, they can find the same page and open the evidence.
That’s the handoff. The owner now has the finding, with its evidence, ready to review.
What Evergreen AI flags
Expired information
Instructions or approvals whose stated validity period has ended.
Outdated facts and references
Claims, tools, or processes that other pages describe as replaced.
Ownership to verify
Contact or owner details that need a person to check them.
Findings are suggestions for review. Treat confidence as a review aid, and a scan with no findings does not prove every page is accurate. Read how we evaluate findings.
Start with one space
- Install from Marketplace. Ask your Confluence administrator to review the permissions and install the app.
- Select a space. Open Evergreen AI and choose the content to audit. The onboarding scan covers up to 100 pages using the Standard model.
- Review a finding. Read its quotes and source pages. Confirm, dismiss, or assign it, then choose whether to schedule recurring scans.
No model API key is required. Request a guided first scan if you want help choosing a space and reviewing the results.
Pricing and scan limits
Try it with your own Confluence content
Free for up to 10 users. A 30-day free trial is available for teams of all sizes. Larger teams pay through Atlassian Marketplace. Use the Marketplace pricing calculator for your user count and billing cycle.
- Sample scan: up to 100 pages.
- Regular scan: up to 25,000 planned pages. Pages beyond that cap are excluded from that run; check coverage and scan smaller groups of spaces.
- Monthly analysis budget: 20 million tokens by default. Admins can set a cap from 1 million to 200 million tokens. Tokens measure text processed, so they are not a fixed page allowance.
- When the budget is reached: new analysis pauses; existing findings remain available for review. A paused scan can resume when the monthly budget resets or an admin raises the cap. Calls already running can finish.
Analysis stays inside Atlassian
Evergreen AI carries the Runs on Atlassian badge. Page content is processed by Atlassian-hosted models inside Atlassian infrastructure. This app sends no content to vendor-operated or non-Atlassian services.
Admins select the spaces to scan. Evidence is checked against the viewer’s page permissions. Evergreen AI does not edit page bodies; optional audit-status metadata and assignment comments are off by default.
Permissions and security · Storage, retention, and deletion · Data processing agreement