Hey @Paige Mulligan Paul here.
When the “Related articles” option is enabled in the Help Center, Intercom uses a machine learning model to determine relevance. It looks at a combination of factors, including:
-
Article content (title, body, keywords)
-
User behavior patterns (what readers clicked next)
-
Article relationships and structure
That said, Intercom does not currently scope related articles by product or category by default. So if you have similar article titles across multiple products, the model may surface unrelated matches.
To improve relevance for product-specific content:
-
Use clear product identifiers in article titles or headings (e.g. “Resetting Password – Product A” vs. “Resetting Password – Product B”)
-
Group articles into separate collections or sections per product, which can help guide structure and behavior signals
-
Consider linking related articles manually within each article if you need stricter control
Let me know if you'd like help reviewing how your Help Center is structured happy to share best practices for scaling with multiple products!
Hi Paul, thanks for the explanation! This gave more clarity as to how the learning model works.
As a follow-up, may you confirm whether the machine learning model takes article tags into account when determining the relevance for related articles? Specifically, can it group or prioritize articles with matching tags to improve contextual relevance (e.g. surfacing only Product A-tagged articles when viewing a Product A-related article)?
We’re trying to understand if using consistent tagging across articles could help refine the model’s suggestions, especially when there are multiple different collections but with similar setups and articles. Thank you!
