Research

Build the authority map for an issue

Turns a legal question into a research plan: sub-questions, the authority hierarchy governing each, pasteable search strings, and a stopping rule, without naming a single case it cannot verify.

About 12 minintermediateLitigation

Your prompt5,372 characters

Still to fill in: Legal question, Jurisdiction / controlling court

RoleYou are a research lawyer who plans the search before running it, because the expensive mistake is never a bad case: it is three hours in the wrong body of law. You think in sub-questions, authority hierarchies, and the words a court actually uses. You will not tell me what a case holds; you would rather hand me an empty slot than a plausible one.What I needA research plan for Legal question in Jurisdiction / controlling court, searched in Westlaw, inside these limits: Time and scope limits. The map, not the answers. I pull the authority myself.InputsQuestion: Legal question Facts: Facts that make this issue arise Authority already in hand: Authority you already have Jurisdiction / controlling court: Jurisdiction / controlling court Searching in: Westlaw · Limits: Time and scope limitsHow to work this1. Break Legal question into numbered sub-questions, each narrow enough that one authority could answer it. Name the one that ends the analysis if it goes against me. 2. For each, give the authority hierarchy in Jurisdiction / controlling court from binding to persuasive, by category and court level ("the intermediate appellate districts, which may conflict"), never by name. 3. Name where that authority lives: code title and chapter, agency register, reporter, the treatise or deskbook by its actual name. Mark any title you are unsure exists [UNVERIFIED - confirm it]. 4. Write two or three search strings per sub-question in the syntax Westlaw takes: connectors, field restrictors, court and date filters. Strings I can paste, not search advice. 5. Give me the vocabulary problem: the other names courts use, the word that buries me in false hits, and the term of art that finds the line if it exists. 6. List what could moot all of it: preemption, choice of law, limitations, exhaustion, standard of review, whether the provision was amended after the dates in Facts that make this issue arise. 7. Close each sub-question with a stopping rule: what a complete answer looks like, and the signal to stop.Ask me firstYou have the question, the facts, and the authority I already hold above. Ask me these three questions (what none of that reveals), then stop and wait: 1. Is this governed by statute or regulation in Jurisdiction / controlling court, or is it common law, and do you know, or are you guessing? The two paths share no starting point. 2. What have I already searched, and what came back empty? An empty result usually means the court calls this something other than what I called it. 3. Is my forum bound by its own appellate law, or predicting another sovereign's high court? That changes what counts as authority, not just what persuades. Do not begin until I answer. If I tell you to proceed anyway, state each assumption at the top of your output and mark it [ASSUMPTION - verify].Output formatA numbered plan, one block per sub-question: question, authority hierarchy, where to look, search strings, vocabulary notes, stopping rule. Then the threshold questions. Then a blank citation ledger: a row per sub-question: Authority, Court and year, Holding, Binding or persuasive, Still good law. End with one line naming the two of my answers that changed this plan the most, and which sub-question or search string exists only because of them. If an answer changed nothing, say so. That question comes off the list.Never do this- Do not name a case. Not as an illustration, not as "you will likely find something like," not in passing. You have no reliable recall of case names, citations, holdings, years, or subsequent history, and a plausible case name here is worse than a blank: it sends me hunting for something that may not exist, and it makes me trust the rest of the map. Where a case would go, write the slot: "the controlling Jurisdiction / controlling court appellate decision on this element, if one exists." - Same rule for statutes and rules. If a citation did not come from Authority you already have or my answers, describe the provision by subject, say where in the code to look, and mark it [UNVERIFIED - locate before citing]. Never invent a section number. Authority I gave you, repeat by name only. - If this plan would work unchanged in any state, it is too generic. Name Jurisdiction / controlling court's courts, codes, and sources specifically enough to be wrong elsewhere. - No hedging filler. Cut "arguably," "it should be noted," and "there is likely case law on point." I am the attorney to consult. - Where you do not know how Jurisdiction / controlling court organizes this area, say you do not know and name the secondary source that resolves it fastest. Do not smooth over the gap with fluent prose. - Do not pad. A two-sub-question issue gets a two-block plan. Length is not value.Before you answer- Did a case name, citation, or holding get in anywhere? Cut it. - Is every source a real named place (a code title, a reporter, a treatise), not "the relevant case law"? - Are the search strings pasteable into Westlaw as written? - Would this plan work in another state? It should not. - Does every sub-question carry a stopping rule?

Adds driver's-seat tunes: options instead of answers, questions before work, every citation flagged. Your values come with it.

Pressure-test it

Makes the AI switch hats and attack its own answer.

The brief on the other side will cite three authorities this plan never turns up, and the associate who found them started somewhere I did not. Look at my plan as the associate who wrote that brief. Where did my plan send me that they would not have gone? Name the body of law I am not searching at all: a regulation, an older codification, a federal analogue, a secondary source with a subject index I would have hit in ten minutes. Then tell me which single sub-question they would attack first because my search strings were built on the wrong vocabulary.

Go deeper

Pushes the work further once the basics are right.

I have run the plan and filled the ledger; here is what I found. Now build the skeleton of the research memo from it: question presented, short answer with a stated confidence level, and an argument order that leads with binding authority. Include a paragraph recording what I searched and found nothing, with the strings and databases, so the file shows the negative research.

Before you run it

What to gather first

The question stated the way the court would have to answer it
The facts that make the issue arise, including dates
The controlling court, and whether it is bound or predicting another court's law
Anything you have already searched, including the searches that came back empty
Your time and billing limit; it sets the depth

Watch for

This prompt deliberately supplies no case law. If the output names a case, a reporter volume, or a pin cite you did not provide, treat the whole map as unreliable and re-run it.
Statute and rule numbers drift between codifications and get renumbered after amendments. Confirm every citation in the code itself before it reaches a brief.
A plan built on the wrong vocabulary produces confident empty results. If three well-formed searches return nothing, the term is wrong, not the law.
The model will miss regulatory, administrative, and local-rule sources almost every time. Check independently for agency guidance, standing orders, and local rules.
Do not paste client-confidential facts or privileged assessments unless your firm's AI policy and the engagement terms permit it.

What comes back

A numbered plan, one block per sub-question: the question, the authority hierarchy from binding to persuasive, the named sources where it lives, two or three pasteable search strings, the vocabulary problem, and a stopping rule. Then the threshold questions that could moot the research. Then a blank citation ledger for the lawyer to fill as the searches run.

See an example of what you’ll get

*(After I answered: the Freedom to Work Act may govern part of it, my search for "non-compete consideration Illinois" returned 400 hits, and this is Cook County state court.)* Sub-question 1. Does the Illinois Freedom to Work Act apply to a covenant signed on this date, and does it displace the common-law consideration rule or sit on top of it? - *This one ends the analysis if it comes out wrong.* If the Act applies and sets its own consideration floor, the common-law line becomes a fallback argument, not the case. - *Authority hierarchy:* the statute's own text and applicability section; then the Illinois Supreme Court construing it, if it has; then the Appellate Court districts, which may conflict; then federal decisions applying Illinois law: persuasive only, and a Cook County judge will say so. - *Where to look:* 820 ILCS 90 [UNVERIFIED - confirm the chapter number and the effective date in the code, and read the applicability section before anything else]. Then the annotations in West's Smith-Hurd Illinois Compiled Statutes Annotated; then the IICLE employment law deskbook chapter on restrictive covenants, which will tell you the current state of the district split faster than any search will. - *Search strings (Westlaw):* `adv: "820 ILCS 90" & consideration`, then `adv: co(IL) & "Freedom to Work Act" /p (consideration "adequate consideration")`, then `adv: co(IL) & da(aft 01-01-2022) & ("restrictive covenant" "covenant not to compete") /s consideration` - *Vocabulary:* courts here say "restrictive covenant," "covenant not to compete," and "post-employment restriction" interchangeably. Searching `consideration` alone drowns you in contract-formation cases. Pair it with a duration term or with "continued employment." The term of art that finds the line, if the line exists, is "adequate consideration" within the same sentence as a period of months or years. - *Stopping rule:* stop when you have (a) the applicability section read word for word against the signing date in my facts, (b) either a supreme court decision or a secondary source that names which districts are on which side, and (c) one source published in the last eighteen months confirming nothing has moved. Threshold questions that could moot all of this 1. Effective date and applicability: does the Act reach agreements signed before it took effect? 2. Earnings threshold: the Act keys some restrictions to compensation levels; confirm where my employee falls. 3. Choice of law and forum in the agreement itself. If it points at Delaware, sub-question 1 may be irrelevant. 4. Severability and judicial reformation: even a covenant that fails on consideration may be reformed rather than voided, which changes the relief I can ask for. Citation ledger: fill as you go | Sub-Q | Authority found | Court / year | Holding in my words | Binding or persuasive | Still good law | |---|---|---|---|---|---| | 1 | | | | | | | 2 | | | | | | Note on what is missing here, deliberately. No case names appear above. Where you see "the Appellate Court districts, which may conflict," that is a slot for you to fill. I cannot tell you which districts, which decisions, or what they held without inventing them, and an invented case in a research plan costs more time than it saves. What your answers changed. Telling me the Act *may* govern and that you are not sure is why sub-question 1 is a statutory-applicability question, why it sits first, and why it carries the ends-the-analysis flag. On a straight common-law answer the plan opens on the consideration line, the applicability section is not the first thing you read, and threshold question 1 does not exist. The 400-hit search is the reason there is a vocabulary paragraph here at all: pairing `consideration` with a duration term or with “continued employment,” and the third Westlaw string with its date filter and its /s connector, exist only because a bare term-search already drowned you once.

Why this prompt is built the way it is

## Framework 1. **Decompose first.** Break the question into sub-questions narrow enough that one authority could answer each, and name the one that ends the analysis if it comes out wrong. 2. **Hierarchy before search.** For each sub-question, state what binds and what merely persuades in this jurisdiction, by category and court level, never by case name. 3. **Send me to real places.** Code title and chapter, agency register, reporter, named treatise or CLE deskbook, pattern instruction set. Sources by their actual titles. 4. **Search strings, not search advice.** Terms and connectors with field restrictors, in the syntax the lawyer's database actually takes. 5. **Solve the vocabulary problem.** The alternate names courts use, the word that buries you in false hits, the term of art that finds the line if it exists. 6. **Name what moots it.** Preemption, choice of law, limitations, exhaustion, standard of review, amendment dates. 7. **Stopping rule per sub-question.** What a complete answer looks like and the signal that says stop. 8. **No cases from memory.** Where a case would go, leave a labeled slot for the lawyer to fill.