Key Takeaways
- Accuracy Potential: Generative AI systems offer the potential of higher recall and precision rates, outperforming traditional TAR accuracy while processing multimodal data, including images and audio files.
- Judicial Standards: As of August 2025, there is no key ruling that approves the use of GenAI for document review, such as Da Silva Moore v. Publicis Groupein 2012, where a judge supported the use of TAR.
- Cost-Benefit Evolution: While some GenAI systems may cost more than traditional TAR software, they have the potential to deliver results faster than manual review and substantially reduce the risk of missing critical documents in complex litigation.
AI-Powered Document Review Marks a Pivotal Opportunity in Legal Practice
After more than a decade of courts increasingly accepting Technology Assisted Review (TAR) for document discovery, a new wave of artificial intelligence is entering review platforms. Generative AI (GenAI) for document coding should deliver a significant leap from traditional predictive coding to sophisticated document analysis that can rival and at times exceed human comprehension.
Fundamentally, GenAI approaches document review as an “instruction-following” or human-mimicking collaborator, capable of handling more complex, nuanced, and multimodal evidence with greater transparency and less up-front labor. TAR, although trusted and proven, is ultimately a classification engine dependent on human-driven feedback loops and is constrained by its architecture and training requirements.
The Evolution from TAR to GenAI: A Technological Leap
Traditional TAR systems have served the legal community well, achieving acceptable recall rates while dramatically reducing document review costs. Recall is the percentage of relevant documents that are retrieved, while another metric called Precision reflects the percentage of retrieved documents that are relevant. These machine learning algorithms learn from human-coded “seed sets” to classify documents as relevant or not relevant.
TAR’s limitations have become increasingly apparent as data volumes have exploded and document types have diversified, as complex documents contain many concepts that can confuse or dilute the training’s intended instructions.
Generative AI represents a new approach that may improve the process of document classification, addressing the very shortcomings that hinder TAR. Built on large language models (LLMs), GenAI offers a level of contextual understanding of language that enables it to identify targeted concepts distinct from extraneous ones. Where TAR might achieve respectable recall rates, proponents believe GenAI will exceed TAR results and, because of its more nuanced identification potential, may reduce the risk of missing critical documents.
Perhaps more importantly, GenAI can process multimodal data, analyzing images, interpreting visual elements (such as pictures and diagrams), and transcribing audio files, providing capabilities that extend beyond traditional machine learning’s text-focused limitations.
The Technology Behind the Transformation
GenAI systems for document review coding operate fundamentally differently from their predecessors in TAR. These systems can:
- Process documents using a reasonable seed set development
- Provide detailed explanations for relevance determinations
- Adapt to changing case strategies without complete retraining
GenAI Example using Relativity’s aiR for Review
In a case study published by EDRM, Tara S. Emory of Covington & Burling LLP outlines a practical framework for evaluating Generative AI (“GenAI”) review for organizing documents in discovery using Relativity’s aiR for Review. The study designed a testing and assessment protocol to identify when GenAI performed well, when it didn’t, and how to build effective hybrid workflows, with insights aimed at guiding legal teams considering GenAI for document organization and review.
The study’s findings provide critical insights into the capabilities and limitations of GenAI in legal document review. Through four rounds of iterative prompt refinement, testing showed that GenAI review performed better for some issues than others. Issues requiring nuanced content analysis generally performed well, while those with rules-based components, such as multiple criteria or date dependencies, performed less well. Perhaps most significantly, the research demonstrated that developing and refining prompts required significant upfront investment from senior attorneys, but aiR then effectively scaled their expertise across the document set.
The study revealed that GenAI review functions most effectively as one tool among others, rather than a replacement for traditional search methods. This approach involves strategic search design, which entails understanding each tool’s capabilities and the different legal and factual characteristics of various search needs. This hybrid approach, emerging organically from the testing insights, provides a practical roadmap for legal teams seeking to integrate GenAI into their document review workflows while maintaining the rigor and defensibility required in litigation contexts.
Judicial Standards
As of August 2025, there is no key ruling that approves the use of GenAI for document review, such as Da Silva Moore v. Publicis Groupe in 2012, where a judge supported the use of TAR. In EEOC v. Tesla, Inc., Case No.: 3:23-cv-04984-JSC, Document 88., a stipulated order for ESI states:
“Tesla has notified plaintiff’s counsel that it may use TAR and/or GEN AI tools to further analyze documents for relevance after search terms are used to narrow the starting document universe to exclude documents not likely to be relevant. If the producing party intends to use TAR, GEN AI, or similar advanced analytics as a substitute for attorney responsiveness review, the parties agree to meet and confer in good faith to attempt to reach agreement about the technology and process that a producing party proposes to use to identify responsive ESI and a statistically sound methodology to determine the recall rate and other measures of the effectiveness of the tool and processes in identifying responsive documents.”
Although there is no precedential application to a stipulated agreement, and it has no force of law, it will likely exert a force of influence in the future.
Practical Implementation: Lessons from Early Adopters
While GenAI document coding requires smaller “seed” or training sets than TAR, it still requires planning and testing before execution against the full corpus. This effort is minimal compared to the enormous cost and time overruns created by trying to skip or rush this stage of the process.
Legal teams implementing GenAI systems are discovering that success requires careful attention to a few key factors:
Design Tagging Protocols Carefully
Develop clear, concise tag descriptions for GenAI tools. Each tag should be self-contained and not rely on metadata or other tags, which is a shift from some TAR models. Use short, unambiguous instructions, and break up complex concepts into multiple, simpler tags when needed.
Run Pilot Samples and Align Results
Before scaling to your full corpus, select a sample of documents that represent diverse topics and file types, and have both expert human reviewers and GenAI review them independently. Compare the agreements and disagreements between the two. Analyze discrepancies to refine tag protocols and instructions. This iterative QC cycle both calibrates the AI’s performance and builds organizational confidence.
Iterate, Test, and Validate
Continuously adjust tag descriptions, prompts, or review parameters based on the observed results. Use feedback from both successful and erroneous cases to refine instructions for the model, often requiring several rounds of testing to reach optimal recall and precision, much as in TAR’s Continuous Active Learning (CAL) workflows.
Human Oversight Remains Essential
Even with high-performing GenAI, maintain human validation for critical stages:
- Validate outlier results and edge cases
- Review privilege determinations and explainable outputs (e.g., privilege log entries)
- Oversee accuracy, especially for documents that are highly sensitive or ambiguous
Conclusion: A New Era of Legal Discovery
The legal profession’s embrace of GenAI in discovery represents not just an evolution in technology but a revolution in legal practice itself. Those who adapt quickly and thoughtfully to these new capabilities will find themselves better positioned to serve their clients in an increasingly complex and data-driven legal landscape.
For practitioners, this evolution demands new competencies:
- Understanding GenAI system capabilities and limitations
- Developing protocols for GenAI validation and quality control
- Communicating GenAI processes to courts and opposing counsel
- Integrating GenAI tools with traditional legal workflows
GenAI is just another tool in the practitioner’s toolbox. GenAI represents an evolution in the practice of law, but one where the lawyer remains at its center, leveraging technology as a solution to the problem technology has created, i.e., creating a document corpus that is too large for human eyes to evaluate every document.
This analysis is based on current trends in judicial AI adoption and established precedents from Technology Assisted Review case law. As this area of law continues to evolve rapidly, practitioners should monitor new developments and court decisions for the most current guidance.
Written with editorial support by Douglas Forrest, Esq., and Michael Jeffrey Glick, Esq.
ILS Consulting – Gain a Competitive Edge with Strategic ESI Consulting Customized for Your Case
When you work with ILS, our team of eDiscovery experts will assist you in early discovery negotiations designed to prevent these costly delaying tactics, support your motion practice, testify by affidavit and in court, attend discovery-related depositions with you, and help create a strategic discovery plan to litigate your case most effectively. Our consulting team is a well-known, trusted partner that never backs down from advocating for our clients’ positions. We will help ensure that your ESI protocols are designed for Plaintiffs’ interests and that you receive the defense data you need to win your case.
- Passionate, experienced attorneys and technical experts provide battle-tested strategic support and testimony from the outset as eDiscovery challenges and issues arise throughout the case.
- ILS consultants are available for ESI negotiations, meet-and-confers, motion support, testimony, and strategic guidance.
- As exclusive providers to the Plaintiffs’ Bar, our experience, insights, and skillfully utilized toolbox will save you time and money, helping you obtain the evidence that defendants want to deny you.
You might also like:
GenAI Prompting for Legal Professionals
The Modern Attachment Revolution: How Cloud Links Are Reshaping Electronic Discovery
The Hidden Dangers: Understanding Hidden Content Challenges in eDiscovery
The Evolution of Cloud Forensics: Challenges and Solutions in Cloud-Based Investigations
The Digital Deception Dilemma: How Deepfakes Are Transforming eDiscovery
How Microsoft Copilot Tracks Your Every Move and the Potential Legal Risks
When Encrypted Does Not Mean Invisible – Signal, Digital Forensics, and the Illusion of Privacy