Optimizing content item selection for user messaging engagement

A non-transitory computer-readable storage medium storing instructions that, when executed by a computing system, cause the computing system to perform operations comprising:

• receiving, at an online system, a request to display a content item to users of the online system, the request to display the content item including targeting criteria for the content item, wherein the content item includes a selection for launching a chat interface via a messaging platform of the online system;
• detecting an impression opportunity to display one or more content items to a first user;
• generating a set of candidate content items for which the first user satisfies targeting criteria, the set of candidate content items including the content item;
• for each candidate content item in the set of candidate content items, applying a trained model to a set of features, the set of features comprising characteristics of a content provider associated with the candidate content item and included in a request from the content provider to display the candidate content item, the trained model predicting a likelihood that the first user will send one or more messages having or lacking specified content to a content provider associated with the candidate content item;
• selecting, based on the predicted likelihood, the content item for display to the first user;
• receiving user input comprising selection of the displayed content item;
• and
• responsive to receiving the user input, launching the chat interface via the messaging platform of the online system between the first user and the content provider.

A system comprising:

• a processor, and a non-transitory computer-readable storage medium comprising computer program instructions that, when executed by the processor of an online system, causes the processor to perform steps comprising:
• receiving, at an online system, a request to display a content item to users of the online system, the request to display the content item including targeting criteria for the content item, wherein the content item includes a selection for launching a chat interface via a messaging platform of the online system;
• detecting an impression opportunity to display one or more content items to a first user;
• generating a set of candidate content items for which the first user satisfies targeting criteria, the set of candidate content items including the content item;
• for each candidate content item in the set of candidate content items, applying a trained model to a set of features, the set of features comprising characteristics of a content provider associated with the candidate content item and historical data including interactions between the first user and content items on the online system, the trained model predicting a likelihood that the first user will send one or more messages having or lacking specified content to a content provider associated with the candidate content item;
• selecting, based on the predicted likelihoods, the content item for display to the first user;
• receiving user input comprising selection of the displayed content item;
• and
• responsive to receiving the user input, launching the chat interface via the messaging platform between the first user and the content provider.