The Classify AI won’t make filing recommendations if it doesn’t “see” at least one group that holds other documents that it considers contextually related to the document you are viewing.
For each group that holds documents, Classify tries to identify a pattern of contextual relationships in the text of its documents, that differentiates the group from other groups.
For example, if you populate a group named Astronomy with articles about astronomical topics and another group named Ecology with articles about ecological topics, it is very likely that Classify will be able to see similarities among the documents in each group, based on the terms used and their relative frequencies of use. So if you then add a new document about astronomy to your database, Classify will recommend the group named Astronomy in which to file it.
Generally, you will find that as your database grows larger and if you have consistently populated your groups with topically related items, Classify will continue to improve in the usefulness of its filing recommendations.
Remember that Classify looks only at the Content of a new item in attempting to suggest a filing location. It doesn’t consider the Name, Spotlight Comments or other metadata.
See Also is similar in that it examines only the Content of the document you are currently viewing. It looks at the contextual relationships in that document and compares those relationships to every other document in the database, looking for others that are possibly similar.
These AI assistants work only within a database, not across databases. So don’t expect to select a document that’s in your Global Inbox and have Classify recommend a group in another database. (Personally, I like that behavior. I have a number of databases that I’ve created, and so if new content arrives in my Global Inbox, I want to move it to another database myself, as I understand why I created each database and don’t want to leave that decision to an algorithm. But Classify becomes very useful when I move the new item to a database that contains hundreds of groups.)
We recommend that if a group is hierarchical (contains subgroups), that no documents should be found in a group that also contains subgroups. For example, if you have a group named Canines, which contains subgroups such as Dogs, Wolves, Foxes, etc., you should never file a document into the group named Canines. Instead, that document should be filed into one of the subgroups. That maintains the purpose of the hierarchy, and will improve the performance of Classify. If the document fits into the Canines topic but not into any existing subgroup, create a new subgroup for it.