Skip to main content
NeuralSearch’s AI models improve by learning from anonymized, combined usage patterns across connected apps. This shared learning helps NeuralSearch better interpret language, understand intent, and surface more relevant results over time. You don’t need to build or maintain custom training pipelines. By default, all applications in your organization are automatically included in model training. Application owners can change this setting at any time in the dashboard. Opting out excludes that application’s data from future training cycles. Search continues to function normally, but it won’t receive improvements from shared model updates. You can opt back in at any time. To manage your organization’s training preference:
1

Go to Organization Settings

  1. Go to the Algolia dashboard.
  2. On the left sidebar, select Settings > Organization Settings.
  3. In the General section, click Model Training.
2

Review your applications

A table lists all applications in your organization. Filter the table and review the details to decide which applications should participate.
3

Update participation

You can manage preferences individually or in bulk:
  • Toggle switch. Turn training on or off for an application.
  • Checkboxes. Select multiple applications to apply bulk actions.
  • Select all. Apply changes to every eligible application at once.
  • To confirm your selection, select Opt In Selected or Opt Out Selected.
4

Save changes

To apply your changes, click Save Settings. To undo any changes before saving, click Discard.

​
What happens after opting out

  • Future model training runs won’t use data from excluded applications. Previously trained models remain unchanged.
  • You’ll continue to receive NeuralSearch feature updates, but not incremental improvements from shared model training.
  • You can opt back in at any time to receive the latest shared model improvements.

​
See also

To learn more about how Algolia trains language models, including data usage, privacy safeguards, and retention policies, see:
Last modified on September 21, 2026