Predictive AV API

  • Version: 1.7.0
  • Host:
  • Protocols: https
  • Accepts: application/json
  • Responds With: application/json


We’ve now deployed the new ML suggest in place of the existing suggest engine as the default suggester. With this, there are a few big changes to acknowledge.

  • Default behaviour when no dataset is selected will now return any address.
  • The Parsed Endpoint has been removed.
  • We’re partway through moving error messages to the new standardised format of messages as an array of strings.
  • There are bugs, please review the release notes linked on the left-hand menu at the bottom.


  • New Machine Learning Suggest Engine
  • Single prediction against all datasets - As of this release we’ve added a new dataset parameter called combined that will return suggestions from all datasets in one call.
  • Address Type filter - You can now limit suggestions by address type using the new parameter. When used you can get a suggestion on just physical address or mailing addresses. this parameter is optional.
  • Improved data consistency - As part of our data work we’ve standardized the terms and format of addresses to be consistent.
  • Single address record from multiple datasets - Now when you get an address back we will identify which datasets that address came from, the response is a comma-separated string example: gnaf,mailAddress

What is The Predictive AV API?

The Predictive AV (Address Verification) API is an address autocomplete service aimed at businesses and developers that want a single national source of government recognised addresses. The API takes in a partially completed address string and returns suggestions of the most likely address the user was trying to enter, to speed up address form completion or other address needs.

We’ve implemented a range of “fuzzy” logic improving the suggestions returned by the API. This includes alias matching, neighbouring suburb matching, vanity suburb correction, misspelling matching and HTTP parameters, giving implementation-specific controls making better matches given a particular use-case.

Why should you use this Predictive API?

PSMA has been and will continue to be the author of GNAF (Geocoded National Address File) the national authority on address. PSMA takes addresses seriously, we believe we must have the latest and most accurate address holdings in the country. We’re undertaking great technical efforts to increase the usability and fuzzy logic to return better and better results.

More control to you.

PSMA is actively developing more controls for you the developer. The first is a parameter to control which address dataset suggestions are made against. We also have state-based filtering to limit the scope of suggestions to a smaller address list giving you better suggestions.

How you use the API

To use the API you’ll need to first get back suggestions of possible address matches by calling the GET Suggested Addressesendpoint. This API requires at a minimum a 4 char string to start giving you suggestions. Once you have a suggestion that matches the users’ target address, the returned suggestion id which is linked to the target address can be fed into the GET an Address Call. This call returns a detailed list of address attributes describing the selected address. The address information can be used to fill in forms and update your records within a database.