The legal profession knows its sources: statutes, case law, treaties, scholarly literature, and custom. We’re trained to navigate these carefully and critically, always seeking meaning and direction. But a new layer is emerging—one that operates outside formal validation yet increasingly influences our daily legal practice. It’s a source not found on rechtspraak.nl, but in the prompt history of ChatGPT. A quiet, fast-moving development that I, with a touch of irony but equal seriousness, call prompt law.
Strategy as a Use Case
As the co-founder of Uncover—a legal tech company focused on applying AI in legal practice—I see daily how rapidly large language models (LLMs) are becoming embedded in lawyers’ workflows. Not as a gimmick, but as a serious tool for analysis, document drafting, knowledge sharing, and increasingly, for strategy. And it’s no surprise: the power of generative AI is astounding. With the right prompt, you can generate contract clauses, legal summaries, or comparative analyses in seconds. What once took a full afternoon can now be done in minutes.
But that convenience comes with something more fundamental. Anyone who works with LLMs quickly notices patterns. Certain prompts consistently produce good results. They’re saved, shared, reused. Law firms build internal prompt libraries. Legal teams develop their own AI dialects. And increasingly, the same phrasing shows up in AI-generated outputs. If a particular prompt reliably yields a certain outcome—say, for a settlement agreement—that output eventually comes to be seen as "standard." A kind of consensus. A norm.
The Normalizing Role of AI
This gives AI a normative role—even though we never formally granted it one. Prompt law, in other words: not created by courts, but by repetition. Not legitimized by legislation, but by users. An accumulation of assumptions, logic, and linguistic probability that starts to carry legal authority simply because it's well-written and widely used.
For lawyers, this is both an opportunity and a risk. The opportunity lies in AI’s potential as a powerful assistant. Used wisely, it can make legal work more efficient, more accessible, even more creative. But the risk is in mistaking prediction for principle. An LLM doesn’t reason—it predicts. It’s built on patterns, not doctrines. It doesn’t differentiate between a Supreme Court majority opinion and a popular blog post—unless you as the user tell it to.
Generative AI naturally gravitates toward the average. And ironically, the more we reuse AI-generated output, the more powerful that average becomes. Especially in private law, where parties enjoy significant freedom, we risk a scenario where LLM outputs begin to set the tone. “This is how everyone does it now” becomes backed by “this is how the AI usually says it.” But the key question remains: is that sound law?
A New Responsibility for Lawyers
This is where the role of the lawyer becomes crucial. Not to shut AI out—in fact, I believe LLMs will soon be as commonplace as case law databases or internal knowledge banks—but to use it with awareness and expertise. That means understanding how these models function, knowing their blind spots, and recognizing where legal nuance is essential. More importantly, it means realizing that as users, we actively shape what the model returns.
In a sense, we’re witnessing the birth of a new form of legal sourcing. Not a formal source, but one that undeniably impacts the legal landscape. This new layer demands both legal craftsmanship and technological literacy: skill in prompting, awareness of AI ethics, and above all, critical thinking.
An Extra Layer—Not a Replacement
Prompt law doesn’t mean the law is disappearing. It means we’re adding an extra layer—one we must learn to interpret and evaluate as legal professionals. If we do that well, LLMs can enhance our craft—not replace it.
Because the law is not an average outcome. It is a system of choices, interests, and principles. AI can assist in that system, but the responsibility remains ours. That’s why we, as lawyers, must decide which prompts are worth repeating—and which are not.
