Key Takeaway: How are courts and commentators evaluating the use of generative AI tools for discovery and review? In a recent ruling, Kevin Schulte v. LinkedIn Corporation, the Court viewed generative AI review as another form of TAR, not as an entirely new category of technology requiring separate rules or extensive disclosures. This decision, and likely others, will enable legal and discovery teams to evaluate AI adoption strategies, disclosure obligations, and defensibility expectations with greater confidence.

Generative AI continues to reshape discovery, and a recent decision from the Northern District of California provides strategic guidance about how courts will evaluate what types of new or additional disclosures may be required when a party is using generative AI to assess responsiveness.

In Schulte v. LinkedIn Corp., No. 22-cv-00237-HSG (LB) (N.D. Cal. June 30, 2026), Magistrate Judge Laurel Beeler addressed challenges to LinkedIn's use of a generative AI document review tool during discovery.

The decision offers a roadmap for legal and discovery teams. It clarifies requirements for AI disclosures and reinforces the principles that will continue to shape defensible review workflows as generative AI adoption grows. 

The Issues Before the Court 

In this case, LinkedIn disclosed its intent to use a generative AI tool to make responsiveness determinations after first narrowing the review population through a set of search terms. Plaintiffs challenged both the pre-culling process and LinkedIn's refusal to provide additional validation metrics.

The Court rejected those challenges. Rather than creating a separate framework for generative AI, the Court analyzed LinkedIn's approach through the same lens that traditionally applies to TAR.

Search-Term Pre-Culling Remains a Viable Strategy

The plaintiffs’ first argument was that the Court should require LinkedIn to run its generative AI review tool across all agreed custodial data instead of first narrowing the collection through search terms.

The Court disagreed. Relying on precedent developed in the TAR context, the Court concluded that search-term pre-culling remains permissible when it satisfies the proportionality requirements of Fed. R. Civ. P. 26(b) and 34(b)(2) and when no specific deficiency in the search methodology has been demonstrated. Notably, the plaintiffs did not establish that LinkedIn's search terms were inadequate. Instead, they argued that pre-culling itself was improper. The Court declined to accept that premise.

For legal teams managing large matters, this portion of the ruling reinforces an operational reality: defensible reduction techniques remain critical to controlling discovery costs. Search terms, analytics, TAR, and now generative AI continue to work together as components of a broader review strategy.

Established TAR Principles Govern the Use of Generative AI Review 

Perhaps the most significant aspect of the Court’s ruling is its treatment of generative AI review itself.

The governing electronically stored information (ESI) protocol required parties to disclose the use of TAR. LinkedIn's disclosure that it would use a generative AI tool to filter out non-responsive documents was deemed sufficient to satisfy that obligation. The Court noted that “LinkedIn [had] provided the plaintiffs with additional information upon request.”

Plaintiffs sought additional information regarding LinkedIn's validation methodology, including elusion estimates, the document error rate, and reviewer metrics. The Court declined to compel those disclosures, noting “[a]s a general matter, discovery of another party’s evidence preservation and collection efforts, or ‘discovery on discovery,’ is disfavored, as such discovery is typically not relevant to the merits of a claim or defense, and is rarely proportional to the needs of the case.”  It continued: “Such discovery may be warranted if the party requesting it demonstrates that there is a specific deficiency in the other party’s production of documents or other information,” but that “mere speculation” is insufficient. The Court cited Taylor v. Google LLC, No. 20-CV-07956-VKD (N.D. Cal. Dec. 3, 2024), in support of its conclusion.

While courts are likely to expect reasonable disclosure regarding review methodologies, Schulte suggests that they may not require detailed examinations of AI workflows, validation statistics, or model performance without concrete concerns about the quality of the production. Thus, the Court’s reasoning reflects a continuation of traditional TAR analysis. 

What Does This Mean for Legal and Discovery Teams? 

This decision provides encouraging guidance for organizations evaluating the use of AI in the discovery process. Courts appear willing to assess generative AI within the established TAR ecosystem rather than treating it as a fundamentally different category of review.

The accelerating adoption of AI exhibits that successful implementations will require a combination of innovative technology with proven discovery practices. In other words, the focus continues to be on achieving defensible, reasonable outcomes, not on the specific predictive engine used to get there.

The Schulte decision is an early but important milestone in that evolution. It signals that courts are prepared to integrate generative AI into existing discovery frameworks rather than reinvent those frameworks from scratch.

Keep an eye out for more insights that will track and report on how courts and commentators are evaluating the use of AI in handling investigations and litigation.  

Learn more about Epiq AI™.
 

Ed Burke, Managing Director, Antitrust and Global Investigations Practice Group

Ed brings a wealth of experience to Epiq clients as both an experienced litigator and a leader of the Epiq Global Investigations Practice Group. He has overseen document review projects for over 100 complex matters in the US, Canada, and Europe, including over 50 antitrust merger reviews. Ed has more than 15 years of litigation experience at major law firms. He has litigated cases for a wide variety of clients, including UnitedHealth Group, Shell, Reuters, H&R Block, and the NBA Players Association.


Allison Dunham, Vice President, Case Insights

Allison Dunham advises corporate legal departments and law firms on data-driven strategies to reduce risk, control costs, and improve decision-making across the litigation lifecycle. Her work enables clients to understand where automation delivers value and where human judgment remains critical.