AI in Law & Legal Services · AI in E-Discovery & Document Review

What is technology-assisted review (TAR) in e-discovery?

Technology-assisted review is a process where attorneys train a machine learning model on sample document decisions, then use that model to help prioritize and classify the rest of a document set for discovery.

Key takeaways

TAR uses machine learning trained on human-reviewed sample documents to predict the relevance of the remaining, much larger document set.
The most common approach has attorneys review a training sample, tag documents as relevant or not, and let the model learn from those decisions.
TAR is typically used to prioritize review order and reduce the number of documents needing full manual review, rather than eliminating human review entirely.
Courts in multiple jurisdictions have accepted TAR as an appropriate method for conducting discovery review when properly validated.

Solving the scale problem in modern discovery

Modern litigation and investigations often involve discovery document sets running into the hundreds of thousands or millions of files — emails, contracts, internal memos, and more — far beyond what any team of human reviewers could read individually within a reasonable time or budget. Technology-assisted review, commonly abbreviated TAR, emerged as a response to this scale problem, using machine learning to help identify which documents in a massive collection are actually likely to be relevant to a case.

How the TAR process typically works

In a common TAR workflow, an attorney or review team starts by reviewing a sample subset of the full document collection and tagging each document as relevant or not relevant to the matter at hand. A machine learning model is trained on these human tagging decisions and then applies what it has learned to predict relevance across the remainder of the much larger document set. Documents predicted to be more likely relevant can be prioritized for full human review, while the model’s confidence scores help the review team understand where to focus and where they can reasonably decrease the intensity of manual review, often with statistical validation to check the model’s accuracy.

Court acceptance and where TAR fits in

TAR is not a fringe or experimental technique in modern litigation — courts across multiple jurisdictions have accepted it as an appropriate method for conducting document review in appropriate cases, provided the process is applied and validated properly. That said, TAR is typically used to prioritize and narrow the review process, not to eliminate human judgment entirely. Attorneys remain responsible for the initial training decisions, for validating the model’s performance against a sample, and for reviewing documents the process identifies as relevant.

Bottom line

Technology-assisted review uses machine learning trained on human-reviewed sample documents to help prioritize and classify much larger discovery collections, reducing the volume of purely manual review while keeping attorneys involved in training, validating, and overseeing the process.

Go deeper

Important caveats

This is a general description of TAR concepts, not legal advice about discovery obligations or protocols for any specific case.

Frequently asked questions

Is TAR the same as predictive coding?

The terms are often used interchangeably or as closely related concepts — predictive coding generally refers to the underlying machine learning method, while TAR describes the broader review workflow that uses it.

Does TAR eliminate the need for human reviewers?

No — TAR is typically used to prioritize and narrow down which documents need full human review, but attorneys remain involved in training the model and reviewing its output.

Do courts require parties to use TAR?

Generally no — courts have accepted TAR as an appropriate discovery method when used, but parties are typically not required to use it unless agreed to or ordered in a specific case.

Sources

[1]E-discovery standards and practice resources — American Bar Association
[2]Federal court rules and discovery guidance — United States Courts

Written by Editorial Team

Last updated July 28, 2026

Get one well-sourced answer a week

No spam. Unsubscribe anytime.