Every report of legal uncertainty must say which kind it is: a gap in the research or a gap in the law.

Lawyers face two uncertainties that look alike and demand opposite responses. Shallow ambiguity is a research gap: the law exists and the researcher has not found it yet. A question of first impression in one jurisdiction that other jurisdictions have resolved is shallowly ambiguous, and more research closes it. Genuine ambiguity is a doctrinal gap: courts have reached incompatible conclusions, or the legal materials support incompatible readings. More research does not close it. More research returns more conflicting authority.

The practical consequences diverge. Shallow ambiguity calls for a better search. Genuine ambiguity calls for counseling the client about risk, briefing both lines of authority, and distinguishing the adverse one. Treating the first as the second wastes a client's money on a question that has an answer. Treating the second as the first produces confident advice on a question that does not.

A system that says only "the law is unclear" has reported nothing, since some part of every question is unclear. The report has to be typed. Our working taxonomy distinguishes five sources of genuine ambiguity: semantic (open texture in a legal term), normative (principles in tension), methodological (interpretive methods that diverge), jurisdictional (splits between courts), and analogical (new facts that existing cases did not anticipate). Each implies a different next step for the lawyer.

The deeper point concerns the role of the machine. Computation cannot eliminate doctrinal indeterminacy. It can locate it, name its type, and show where the lines of authority run. That makes legal AI a cartographer of contested law rather than an oracle, and a cartographer that cannot tell unmapped ground from disputed ground is not doing its job.