According to Trellis, the Hon. Gary Y. Tanaka of the Los Angeles County Superior Court grants 55.5 percent of the demurrers that come across his bench.
Should that make you more likely to file one?
A grant rate is hard to read in isolation. On its own, it’s just a number. The value comes from the questions you ask about it. That’s where the Trellis Connector for Claude comes in — a tool that connects Trellis’s state trial court data directly to an AI model, letting attorneys move from a single data point to a fully developed litigation strategy without switching platforms or starting from scratch.
Most prompting advice focuses on how you talk to the model—include the right persona, the right instructions, the right format. But as models have become more capable, the challenge has changed. It’s no longer the sentence you type into the model that matters. It’s the analytical framework you bring to the problem.
In this article, we’ll take that one grant-rate statistic and show how a sequence of prompts — each one building on the last — can turn it into a concrete litigation strategy.
The structure of effective prompts
The most effective prompts start with a clearer sense of what you want the model to do. In litigation research, that objective usually falls into one of three categories:
- Put information in context. A grant rate percentage is a fact. Whether that number is high, low, or unusual depends on what it’s compared against. Different comparisons answer different questions.
- Look for what’s hidden. Statistics summarize outcomes. They rarely tell you what happened behind the scenes to produce them. The most useful insights are often buried beneath the numbers.
- Apply the insight. Information becomes useful when it changes how you act. Analysis alone rarely tells you what to do next. The value lies in turning what you’ve uncovered into strategy.
Once you know what kind of analysis you want, you can ask for the form that best fits it: a chart for comparison, a timeline for sequence, a decision matrix for strategy, or a narrative explanation for reasoning.
The prompts that follow illustrate each of these principles in practice.
Start with a baseline
Consider the following input-output to the Trellis Connector for Claude.
Prompt 1: How has the Hon. Gary Tanaka of the Los Angeles County Superior Court ruled on demurrers? Use Trellis.
Mechanically, this is a perfectly reasonable prompt. It asks for a fact, and the model delivers one. We learn that Tanaka sustains demurrers 55.5 percent of the time, slightly above county and statewide averages. From this, the model concludes that he’s a defendant-favorable judge who generally falls in line with his peers.
How can we make it better?
Context and comparison
When an attorney reaches for a grant-rate statistic, they’re usually trying to anticipate what might happen to their own motion. 55.5 percent feels like an answer to that question. But whether that number is meaningful depends on the benchmark.
The output for our baseline prompt provides countywide and statewide averages. While those figures tell us how typical Tanaka is when it comes to demurrer rulings, they don’t tell us much about how he compares to judges handling similar cases.
To answer the question we’re actually asking, we need a more relevant benchmark. The easiest way to get that is to ask for it directly.
Prompt 2a: How has the Hon. Gary Tanaka of the Los Angeles County Superior Court ruled on demurrers? Compare his approach to other judges handling similar cases in Los Angeles County. Use Trellis.
Here, the model isn’t simply reporting Tanaka’s grant rate. It’s letting you evaluate that statistic against a benchmark that matters to you—other judges with similar dockets.
Prompting takeaway: An effective prompt doesn’t just ask for a fact. It tells the model what to do with it. The second prompt turns a standalone statistic into a meaningful comparison.
Outcomes and processes
The output from Prompt 2a tells us that Gary Tanaka and Randolph Hammock both sustain demurrers at nearly identical rates. On this metric, they look like the same judge.
But grant rates can only take us so far. They tell us what happened, not why it happened. To answer that question, we need to look beyond the statistics themselves and examine the court materials that produced them.
Prompt 2b: Tanaka and Hammock sustain demurrers at nearly identical rates. Review a sample of rulings from both judges in labor and employment cases. What’s the primary focus of each judge’s analysis? Provide your response in a chart that gives a clear snapshot of their judicial reasoning, supported with concrete examples.
According to this output, the judges in each courtroom do very different kinds of work. Hammock organizes his rulings around the legal obstacles to recovery, focusing on the legal defects that can defeat the claim as pled. Tanaka begins with the procedural framework, asking whether the claim has been presented through the correct legal pathway.
The distinction matters. Two judges can sustain demurrers at the same rate while using the motion to accomplish very different things. Prompt 2b takes tentative rulings and turns them into a reasoning sequence. We can start to see the analytical moves each judge makes—the questions they ask, the issues they prioritize.
Prompting takeaway: An effective prompt names what an answer leaves unresolved. The more clearly you can identify what’s missing from an answer, the easier it becomes to ask for the analysis you actually need.
Signals and strategy
The output from Prompt 2b reveals something a grant rate couldn’t—Tanaka’s emphasis on procedural sufficiency. That insight raises a practical question: how should a defendant adjust their approach to a demurrer in his courtroom?
Prompt 2c asks that question.
Prompt 2c: Assume a defendant is evaluating whether to file a demurrer to a complaint in Tanaka’s courtroom. Based on the grant rate of his demurrers and the reasoning reflected in his tentative rulings, what strategic considerations should inform the defendant’s decision? Present your response in a chart.
This output changes how we think about a demurrer. The question is no longer whether the demurrer will be sustained. The question is what happens if it is.
Prompt 2c helps us build a decision framework by connecting pleading defects to their likely consequences. We see that some pleading defects can be cured through routine amendments. Others force the plaintiff to identify a new statutory basis, satisfy a procedural prerequisite, narrow a claim, or restructure the case altogether.
For a defendant, each of those outcomes carry very different strategic values. A sustained demurrer is not always a meaningful victory. Some defects simply teach the plaintiff how to revise the complaint. Others can materially alter the course of the litigation.
Prompting takeaway: An effective prompt attaches analysis to a concrete decision. The more precisely you can frame the decision in front of you, the more effectively the model can evaluate the tradeoffs.
Concluding thoughts
Using the Trellis Connector for Claude, we started with a single statistic. Then we put it in context, asked what it might be concealing, and connected it to a real litigation decision.
None of our prompts required a special persona, a proprietary framework, or a carefully engineered instruction set. The value came from knowing what we wanted to learn and asking the kinds of questions that could move the analysis forward.
The prompt isn’t the words. It’s the analytical framework.
FAQ
The most effective judicial research goes beyond reviewing a judge’s grant rates. Attorneys should examine motion outcomes, tentative rulings, case types, and the reasoning judges use when deciding similar matters. Looking at both outcomes and analysis provides a more complete picture of how a judge approaches specific legal issues.
Trellis helps attorneys conduct this research by combining judicial analytics with access to state trial court records, rulings, and motion history. Attorneys can use Trellis data within AI tools like Claude and ChatGPT to analyze judicial behavior, compare rulings, and uncover patterns that may inform litigation strategy.
A demurrer grant rate can provide useful context, but it should not be viewed in isolation. A judge who sustains demurrers 55% of the time may be more favorable to defendants than average, or may simply be handling a different mix of cases.
The most valuable insights come from pairing grant-rate statistics with the underlying rulings. Trellis allows attorneys to move from a single metric to the court records and judicial reasoning behind it, helping them better evaluate whether a demurrer is likely to advance their objectives.
Comparing judges requires more than looking at outcomes alone. Attorneys should evaluate grant rates, the types of cases each judge hears, and the legal or procedural issues emphasized in their rulings.
Using Trellis, attorneys can compare judges handling similar matters, review representative rulings, and identify meaningful differences in judicial approach. Looking beyond the numbers often reveals distinctions that may affect litigation strategy.
Yes. AI can help identify recurring themes, analytical frameworks, and decision-making patterns across large volumes of court records that would otherwise take significant time to review manually.
Attorneys can use Trellis data within AI tools like Claude and ChatGPT to analyze rulings, compare judicial approaches, and better understand how judges evaluate arguments in similar cases. The combination of AI and state trial court data helps turn individual rulings into actionable litigation insights.
A grant rate tells you how often a judge reaches a particular outcome. Judicial reasoning explains how they arrived there. Two judges may sustain motions at similar rates while focusing on entirely different issues in their analysis.
Trellis helps attorneys move beyond surface-level statistics by providing access to the rulings and court records behind the numbers. This allows lawyers to understand not just what happened, but why.
Court data becomes valuable when it helps attorneys make better decisions about motions, case evaluation, settlement, and overall litigation strategy. Historical outcomes, judicial tendencies, and case-specific patterns can all inform how a case is approached.
Trellis provides access to billions of trial court records and judicial analytics that help attorneys evaluate risks, identify opportunities, and develop stronger case strategies. Trellis’s integrations with leading AI platforms make it easier to analyze that information as part of an attorney’s existing workflow.
Attorneys should consider more than whether demurrers are frequently granted. Important factors include the judge’s prior rulings, the types of pleading defects that lead to sustained demurrers, whether leave to amend is commonly granted, and the strategic value of the likely outcome.
Using Trellis, attorneys can review a judge’s motion history, analyze past rulings, and better understand how similar arguments have been received in that courtroom before deciding whether to file.
Trial court data can help attorneys evaluate claims, assess litigation risk, develop motion strategy, identify favorable arguments, and understand how judges have approached similar disputes in the past.
Trellis provides access to one of the largest collections of state trial court data in the country, helping attorneys uncover insights that may not be available through traditional legal research tools alone.
AI can quickly synthesize large volumes of court records and rulings to identify trends that may be difficult to spot through manual review. This includes recurring legal issues, procedural concerns, and analytical approaches that appear across multiple decisions.
Attorneys can use Trellis data within Claude, ChatGPT, and other leading AI platforms to analyze judicial analytics, review court records, and uncover patterns that may help inform litigation strategy.
A grant rate summarizes outcomes, but it does not reveal the reasoning behind those outcomes. Understanding what a judge focuses on, whether procedural requirements, pleading standards, evidentiary issues, or substantive law, often provides more strategic value than the statistic itself.
Trellis allows attorneys to explore both the analytics and the underlying court records, helping them build a more complete picture of judicial behavior and decision-making.
ChatGPT can help analyze and summarize information, but the quality of the analysis depends on the quality of the underlying data. To understand a judge’s motion history, attorneys need access to reliable court records, rulings, and judicial analytics.
By using Trellis data within ChatGPT and other leading AI platforms, attorneys can combine AI-powered analysis with state trial court data, making it easier to evaluate judicial tendencies, compare rulings, and develop litigation strategies based on real court activity.