Technology-assisted review (TAR) in Australian litigation

Technology-assisted review (TAR), also called predictive coding, uses machine learning trained on senior reviewers' decisions to rank or classify documents by likely relevance. Instead of reviewing an entire discovery set linearly, the team reviews the highest-ranked material first and validates the model's performance with sampling.

Australia was an early adopter among common law jurisdictions. The Supreme Court of Victoria endorsed the use of technology-assisted review in large matters, and courts have since supported its use where the parties adopt a transparent, well-documented process. The consistent theme in Australian practice is proportionality: where a matter involves hundreds of thousands of documents, courts expect parties to consider technology rather than defaulting to exhaustive manual review.

A defensible TAR workflow in an Australian matter generally involves:

• Senior lawyers training the model through review of sample documents

• Iterative rounds where the model learns and re-ranks the set

• Statistical validation, such as sampling to measure precision and recall

• Documentation of the process, often agreed with the other side in the discovery protocol

Newer AI methods extend this foundation. Language models can now summarise documents, extract facts, and classify for privilege alongside traditional relevance ranking - the same defensibility principles of transparency, validation, and human control apply.

For how courts and practice guidance are treating these newer approaches, see Sedona Canada and Technology-Assisted Review in 2026, which covers themes that apply well beyond Canada.

Claira is an AI eDiscovery platform that combines classification, summarisation, and fact extraction for large-scale review. See how it works.