Old Rules, New Tools: Why Schulte v. LinkedIn Validates What We’ve Been Saying All Along

A new decision out of the Northern District of California is giving litigants some welcome clarity on generative AI in discovery—and it confirms a position Cimplifi has held since generative AI review options first hit the market.

In *Schulte v. LinkedIn Corp.* (N.D. Cal. June 30, 2026), LinkedIn planned to narrow it’s custodial document population with negotiated search terms, then use Relativity aiR, a generative AI-powered review tool to make responsiveness determinations on what remained, with those AI-driven determinations then quality-checked through sampling and human review. Plaintiffs challenged the approach. The court rejected every challenge.

Two parts of the ruling matter most.

1. Generative AI Review Is Just Another Form of TAR

The court didn’t invent a new legal framework for generative AI. It analysed the technology under the same proportionality, reasonableness, and transparency standards that have governed technology-assisted review for over a decade.

This is exactly the position Cimplifi has taken from the start: generative AI review isn’t a novel, unproven category of discovery requiring special rules—it’s TAR. And it should be evaluated the way TAR has always been evaluated, with recall and precision statistics used to demonstrate reasonableness and defensibility. Parties shouldn’t need a new playbook to justify using it, and they shouldn’t need to approach it with any less confidence than they’ve had in TAR methodologies courts have accepted for years. This ruling gives that position real judicial backing.

2. Combining Workflows Doesn’t Undermine Them

The court also rejected the argument that using search terms to pre-cull the document population before AI-assisted review was itself a problem. Because plaintiffs never challenged the search terms themselves, using them to narrow the population before generative AI review, followed by human sampling for quality control was entirely appropriate.

That’s a meaningful point for anyone building a discovery plan: choosing one workflow tool doesn’t forfeit your ability to use another reasonable one alongside it. Search terms, generative AI review, and human QC aren’t competing options where you have to pick a lane. Used together, they’re how a complete, defensible, and cost-effective discovery plan actually gets built. The right question isn’t “which method should we use?”—it’s how to orchestrate the methods available so the outcome is both efficient and defensible.

The search and information retrieval team at Cimplifi curates and tests search terms to suppress false-positive hits without excluding true positives, tuning precision and recall before a single document moves downstream and our GenAI responsiveness prompts are calibrated against a statistically significant sample, then put through a completeness validation exercise before anything runs to production. Human QC and sampling sit inside both steps, which is what makes the resulting recall and precision statistics defensible rather than merely reported.

The Takeaway

Schulte doesn’t ask parties to rethink discovery from scratch. It confirms that the frameworks already in place—reasonableness, proportionality, and metric-driven defensibility—work just as well for generative AI review as they have for traditional TAR. For parties already comfortable defending TAR workflows with recall and precision data, the path to defending generative AI review runs through the same territory.

The tools are new. The rules, and the confidence required to use them aren’t.

For more regarding Cimplifi specialised expertise regarding AI & machine learning, click here. To learn more about AI-powered review, click here.