Editor’s Note: This is Part 1 of a four part-series. Part 2 is available here.

These are chicken nuggets. Do we know what part of the chicken these are made from, breast, wing, thigh or is it made from mechanically separated chicken, which is essentially chicken parts forced through a sieve at high pressure until they form a paste, then shaped and breaded; do we even know if it is chicken. We have moved from farm-to-table to factory farming to ultra-processed foods where we have no idea what we’re eating. Sadly, our information ecosystem has followed the same trajectory as the food ecosystem. This collapsing of the information ecosystem has been decades in the making and is now exacerbated/accelerated by generative AI (GAI). We once encountered primary legal sources directly by physically pulling volumes from library shelves, reading full judicial opinions, understanding law as textual and jurisdictional. Then came digital databases where sources remained visible but mediated through search interfaces. Now we’ve arrived at AI-generated summaries where sources disappear entirely behind synthesized answers. We know (think?) nuggets contain “chicken” without understanding whether that means breast meat or mechanically separated paste, in the same way many of our students know their research contains “law” without understanding whether that means binding Supreme Court precedent or a lawyer’s blog post. Everything looks the same on a screen. Everything is just “content” delivered through identical interfaces. Information literacy, much like food literacy, is vitally important. A student who can’t identify chicken parts might make poor nutritional choices. A lawyer who can’t identify legal sources, much less relevant ones, may commit malpractice.

We’re training a generation of lawyers who rarely engage with the raw materials of their profession, and are increasingly consuming only the processed, pre-digested, AI-synthesized versions. Students are suffering from what we might call source blindness, the inability to distinguish between fundamentally different types of sources, compounded by source erasure, where sources disappear behind AI-generated summaries.  The question facing legal education is whether we can build information literacy in an environment designed to make sources invisible.

How We Got Here

The journey to this crisis happened in stages. Think of it as the ultra-processing of legal knowledge, following the same trajectory as our food system’s evolution from farm to factory to packaged products of mysterious origin.

In the first stage, the farm-to-table era of bound reporters, students experienced direct connection to sources. They would have walked into a law library, confronted rows of bound reporters, and experienced the materiality of legal authority. The weight of the Federal Reporter volumes would have communicated something about the accumulated density of precedent. They understood implicitly that this volume contained cases from specific courts during specific years. They saw how many cases existed on a topic because they physically encountered them. They read full opinions, including the procedural history that explained how the case arrived at this court, the facts that grounded the legal analysis, the reasoning that led to the holding, and often the dissents that revealed doctrinal tensions. The citation format emerged naturally from the physical structure of the reporters themselves—volume number, reporter abbreviation, page number, year. Everything about this experience reinforced that legal authority was textual, specific, jurisdictional, and temporal.

The second stage, factory processing through early digital databases, maintained source visibility while adding distance. Westlaw and Lexis transformed legal research by making vast libraries searchable, but students still encountered sources as distinct objects. When they searched for Fourth Amendment cell phone cases, they received a list of actual cases. They saw that there were two hundred results and learned to refine their search, to evaluate which cases seemed most relevant, to read headnotes before diving into full opinions. The database was an intermediary, but a transparent one. They knew they were looking at cases, not summaries. They developed Boolean logic skills and learned controlled vocabulary because the database required them to think carefully about how to find what they needed. The research process remained iterative and visible, they could see their search evolving, understand why they were getting certain results, recognize gaps in their research.

The third stage, ultra-processing through generative AI (GAI) synthesis. When students ask a GAI tool about Fourth Amendment cell phone protections, they receive synthesized answers that hide all traces of their own construction. The GAI may show them a list of cases to evaluate, but that list is subordinate to the conclusions presented. Students have little to no visibility into which sources the GAI consulted, which it excluded, or how it weighted different authorities. They don’t see the 200 cases that exist on this topic, they see the five or six the GAI selected for reasons it may or may not explain. They don’t encounter doctrinal tensions because the GAI smooths them into apparent consensus. They don’t learn about circuit splits because the GAI synthesizes across jurisdictions as if law were uniform. They don’t understand temporal dynamics because the GAI’s training data has a cutoff date they don’t know and couldn’t interpret if they did. The answer appears authoritative, comprehensive, and final. It includes case names, citations in proper Bluebook format, and a clear explanation of the legal rule. The student copies this into an assignment and moves on. They have no idea that they haven’t done legal research at all. They’ve consumed a summary of unknown provenance, synthesized from sources they’ve never seen, using selection criteria they don’t understand, with potential errors they cannot detect.

Part 2 will look at the cognitive architecture of dependency, the ways GAI amplifies cognitive bias, and issues of source blindness and source erasure.

Editor’s Note – This article is republished with permission of the author as well as the Editor of RIPS Law Librarian, first publication.