Studying for a License Exam With AI as a Tutor

Why this matters

A chat assistant will write you a hundred practice questions on any chapter of your code book in a minute, explain each answer patiently at eleven at night, and never get tired of being asked the same thing twice. That makes it the best study partner most techs have ever had access to. It will also, on a predictable share of those questions, give an answer key that is wrong: a requirement from a different edition, a rule from a different state, or a section number that does not exist. A tech who studies from its answers memorizes those errors and walks into the exam confidently wrong. Used as a quiz partner whose every answer you check against the book, it is excellent. Used as an answer key, it is a liability.

This card is about that one method. The broader exam path (applications, experience requirements, which exams your state gives, what to bring on the day) is covered elsewhere (see related: Trade License Exam Preparation Reference), and so is fitting study around a working week (see related: How to Take an Exam While Running a Shop).

Why the tool is a quiz partner and not an answer key

Licensing exams for the trades are mostly open-book tests of the specific code edition and reference books your state names. The question is rarely "what is the rule in general"; it is "what does this edition of this book say, as adopted here." A language model has read many editions of many codes, many states' amendments, many forum arguments and many study guides, all blended. When it answers, it produces the most typical answer across that blend, with no reliable way to tell you which edition or jurisdiction the answer came from. Why it does that, and why it sounds equally sure either way, is covered in its own card (see related: What a Language Model Is Actually Doing When It Answers).

So the tool's strengths and weaknesses line up neatly with the exam's demands:

• Strong: generating many questions in the style of the exam, explaining a concept in plain words, rephrasing a dense section until it clicks, quizzing you on vocabulary, and asking you "why" until you can say it back.
• Weak: knowing what your edition says, knowing your state's amendments, getting section numbers right, and knowing your state's business and law material, which varies the most and is represented the least in what it learned from.

The method below uses the strong half and routes every claim through the book.

The method, step by step

1. Get the exam's own content outline and reference list before anything else. The state board or its testing provider publishes a candidate information bulletin listing the exam's subject areas, how many questions come from each, and exactly which books and editions are allowed. This is the one document that decides what you study. Skip it and you will spend evenings on material that carries a handful of questions while the heavy sections go thin. Also note whether the exam allows tabbed and highlighted books, since that changes how you prepare the book itself.

2. Tab and know your physical book. Most of the score on an open-book exam is speed of lookup. Put tabs on the chapters and tables the outline weights heavily. The tool cannot do this for you, and it is the single biggest difference between candidates who finish and candidates who run out of time.

3. Generate questions one chapter at a time, and tell the tool to cite. A request that works: "Write 20 multiple-choice questions in the style of a trade licensing exam on [chapter topic] in the [year] edition of [code name]. For each, give the answer and the section it comes from." Asking for the section matters even though the section may be wrong, because it gives you something specific to check, and a citation that does not exist is itself a strong signal the answer is wrong too.

4. Answer the questions yourself first, using the book, under time. Do not read the tool's answers first. Work each question with the book open, timed to roughly the exam's pace (the bulletin gives question count and time; divide). This is the actual exam skill: reading a question, finding the right place in the book, and reading the rule exactly.

5. Compare against the book, not against the tool. For each question, you now have your answer from the book and the tool's answer. When they agree, confirm the section once more. When they disagree, the book wins, every time, and you mark the question. Most disagreements will be the tool's error. Some will be yours, usually a misread exception, and those are the most valuable questions in the set because they show you exactly how the exam will try to catch you.

6. Log every error the tool made, by kind. Wrong edition, wrong state, invented section, wrong on the plain text. After a couple of sessions you will see which kind dominates in your trade's material, which tells you how much to trust its explanations in that area (see related: The Invention Patterns to Expect in Trade Questions).

7. Put every missed or marked question on a review schedule. This is spaced repetition: reviewing an item at widening intervals, so each review comes just before you would have forgotten it. Questions you got right on the first try get reviewed less; ones you missed come back sooner. A simple paper card or a spreadsheet does this fine. The schedule is below.

8. Use the tool to explain after you have the right answer. Once the book has settled what the rule is, ask the tool why the rule exists, or to explain it in plain words, or to give you a field example. Give it the correct rule in your prompt, so it explains that rule rather than its own version. This is where it is most useful and least risky, because you already hold the fact.

The review schedule

Review intervals for one practice question
(days counted from the first attempt)

Day 0    first attempt, checked against book
Day 1    review   -> correct: go to Day 3
                  -> missed:  back to Day 1
Day 3    review   -> correct: go to Day 7
Day 7    review   -> correct: go to Day 14
Day 14   review   -> correct: go to Day 30
Day 30   review   -> correct: retire the card

A miss at any review sends the card back to Day 1.

Reviews take seconds each, and a few weeks in you will be reviewing a few dozen a day, mostly ones you know. That is the point: the effort goes to the ones you keep missing.

A worked week

A plumbing apprentice is preparing for a journeyman exam with five weeks to go. His bulletin weights the drainage, waste and vent section heavily and names one specific code edition.

Monday. He asks the tool for 40 questions on drainage and venting, edition named, with sections. He works all 40 with his tabbed book, timed. It takes him 95 minutes, about 2.4 minutes a question, against an exam pace that works out to under 2 minutes a question in his bulletin. Lookup speed is his first problem, before content.

Tuesday. He compares. His book-derived answers and the tool's answer key agree on 31 of the 40. On the 9 disagreements:

• 6 are the tool's error. Of those, 3 cite a section that, in his edition, either does not exist or covers something else entirely; 2 state a requirement that he finds in the book's previous edition but not his; 1 is simply wrong against the plain text.
• 3 are his error, all the same kind: he read the general rule and missed an exception two lines below it.

So the tool's answer key was wrong on 6 of 40 questions, 15 percent, and he was wrong on 3 of 40, 7.5 percent. The combined 9 of 40 marked questions go onto his review schedule. Had he studied from the tool's key instead of the book, he would have learned 6 wrong answers in one chapter, and the 2 edition errors are the sort an exam is built to catch, since the exam writers know which requirements changed.

Wednesday to Friday. He repeats with new chapters at 20 questions a night, which leaves time for reviews. His Day 1 reviews of Monday's 9 marked questions come back 7 right, 2 missed; the 2 go back to Day 1. By Friday his time per question on new sets is down to about 1.8 minutes, under exam pace, because the tabs are now where his hand expects them.

The pattern he logged. 5 of the tool's 6 errors in drainage were edition or citation problems, not misunderstanding. That tells him the tool's explanations of concepts (why a vent is needed, how a trap seal is lost) are worth reading, and its numbers and section references are worth nothing until checked.

What changes the approach

The business and law portion. Many states give a separate business and law exam, or a section of it, covering lien law, contracts, licensing rules, insurance and tax. This material is state-specific and the tool's training is a blend of fifty states. Treat its answers here as weaker than on the trade code, and study from the state-named reference book directly; use the tool only to explain concepts like what a mechanical lien is, then confirm your state's deadlines in the book.

A closed-book exam. A few exams or sections are closed-book. Then memorization matters more, the review schedule carries more of the load, and you should cut the timed lookup practice in step 4 back to just checking.

Recent code changes. If your state adopted a new edition recently, the tool has seen far more material about the old one. Expect the edition error rate to be higher, and weight the comparison step more.

One line you do not cross

Do not paste actual exam questions into a tool after you have sat an exam, and do not use questions anyone claims are from the real exam. Candidates agree to keep exam content confidential, and boards treat sharing it as grounds to void a score or deny a license. Practice questions the tool writes from the code are fine; reproducing the exam is not.

References

• Your state licensing board's candidate information bulletin, which names the exam's content outline and the allowed reference editions
• The code edition your state has adopted, with its local amendments
• See related: Trade License Exam Preparation Reference
• See related: How to Take an Exam While Running a Shop
• See related: The AI Quoted a Code Section That Does Not Exist