Rayyan’s AI-powered Predictions Classifier helps you prioritize your screening by ranking articles based on how likely they are to be included in your review — based on the patterns it learns from your own decisions.
Whether you're working solo or with a team, this feature can save you hours of screening time by pushing the most relevant articles to the top!
🧠 How It Works
As you make inclusion or exclusion decisions, Rayyan’s AI starts learning from your criteria. Once enough data is available, you can compute ratings to see a prediction score (thumbs up/down) for each undecided article — helping you decide where to focus next.
✅ Steps to Use the Predictions Classifier
1. Start Making Decisions
Before you can use the classifier, make sure to include or exclude at least 50 articles, with at least 5 of each type (Included + Excluded).
If the minimum isn’t met, the Ratings will not be calculated.
💡 You’ll find the Compute Ratings button in the top toolbar of your review screen.
2. Click Compute Ratings
Once eligible, click the Compute Ratings button. You’ll see a message:
“This action requires some time, will let you know when it’s done!”
Rayyan’s AI will now analyze your decisions and assign a rating to the remaining undecided references.
When it's done, you’ll see:
"Ratings computed for articles!"
The Ratings column will show:
-
👍👍 Most likely to include
-
👍 Likely to include
-
❓ No recommendation
-
👎 Likely to exclude
-
👎👎 Most likely to exclude
⚠️ Ratings are not available for articles that were already decided before the rating was computed (even by collaborators).
3. Sort by Ratings
You can now sort the list using the Ratings by clicking the sorting arrow beside the "Filter by Inclusion" facet.
The visible thumbs give a general idea, but the sort order uses the precise internal score for maximum accuracy.
4. Filter by Ratings (Premium Feature)
From the Filters workbench, scroll to the Ratings section and select filters like:
-
Most likely to include
- Likely to include
- No recommendation
- Likely to exclude
-
Most likely to exclude
You’ll also see how many articles fall under each category, which can help you estimate screening progress.
🔐 Note: This filter is available to Premium users only.
5. Recompute Anytime
The more decisions you make, the smarter the predictions get. You can recompute ratings anytime as your review evolves — just click the button again.
🔄 Tip: We recommend recomputing after every major batch of decisions (e.g., every 50–100 decisions).
🧑🤝🧑 Working in Collaborative Reviews
-
Blinding doesn't block predictions — Rayyan computes based on all available decisions, even if members can't see each other's work.
-
If your team is working blindly, ensure each person is making a solid number of decisions to improve the accuracy of predictions.
-
Ratings are never shared as decisions; they are suggestions only.
🧠 Why Use This Feature?
Relevance Ranking helps:
-
Prioritize articles likely to be included
-
Reduce workload by focusing on high-probability papers
-
Surface overlooked articles early in the process
📅 Interested in a webinar?
We're excited to invite you to our upcoming Rayyan webinars! Whether you're new to the platform or looking to deepen your knowledge, our webinars are designed to help you get the most out of Rayyan. Register here to secure your spot!
🆘 Need Assistance?
Start by browsing our FAQs for quick answers. Still stuck? Submit a support request through the Help Center.
When reaching out, please include:
-
The name of your review
-
Steps you took before encountering the issue
-
A screenshot (if available)
Comments
0 comments
Article is closed for comments.