How we write our practice questions

Every practice question on LSATPathway is original. None of them is a real LSAT question, none is licensed from LSAC, and none is copied or paraphrased from a published test. This page explains how they are made, what they are checked against, and where their limits are.

Questions are drafted by AI writers working to the specification on this page. Every draft then goes through the audits below, and difficulty labels are corrected against real student results. The specification, the audits, and the method are ours; the drafting is automated so the bank can grow by roughly 130 questions a day.

The specification

Questions are written in batches: one full timed section plus one complete practice test a day, about 130 questions. Every batch is written to the same targets, which follow the real test's shape:

In a 26-question Logical Reasoning section: about 4 Necessary Assumption, 4 Flaw, 3 Strengthen, 3 Weaken, 3 Inference, 2 Sufficient Assumption, 2 Method of Reasoning, 1 Main Point, 1 Point at Issue, 1 Parallel Reasoning, and 2 from Principle, Paradox, and Evaluate.
Difficulty rises through an LR section, like the real test: the first ten questions are mostly easier, and the last six are mostly medium to hard.
Reading Comprehension comes in sets of four passages, 450–550 words each, with 7 questions per passage: one main point, one author attitude, two detail, one inference, one passage function, and at most one strengthen or weaken.
Stimuli rotate across six subject domains — law, science, humanities, social science, business, and everyday civic life — so no single domain dominates the bank.
Correct answers are balanced across A through E, roughly five or six per letter in a section, with no patterns to game.

The writing rules

Nothing is reproduced or closely paraphrased from a real LSAT question. Everything is written fresh.
Logical Reasoning stimuli run two to six sentences. No current events, no real politicians, no real companies, no pop culture.
Every wrong answer has to tempt for a specific reason: out of scope, reversed logic, too strong, too weak, half right, or true but not answering the question asked. At least one wrong answer in every question is true — it just does not answer the stem.
No two questions in a batch may share the same argument structure. Vary the skeleton, not just the topic.

The explanation rules

Every question carries an explanation written to fixed rules, in plain language, one or two sentences per point: why the right answer is right, and why the closest wrong answer is wrong. Reading Comprehension explanations must do both halves and tie the right answer to specific passage text.

The method is fixed, and explanations that break it are sent back and rewritten before the batch is imported. Two examples:

A Necessary Assumption explanation states the conclusion, then applies one test: if this conclusion is true, what has to be true? It never uses the negation test.
A Sufficient Assumption explanation shows what the answer adds, so that with it added, the conclusion follows.

The daily audit

Before new questions reach students, and again every morning, the bank is audited:

Duplicates are removed — exact repeats, and questions that make the same argument in different words.
Every question must carry a specific question type, a difficulty from 1 to 5, a subject domain, and, for Flaw questions, a flaw subtype. Anything missing or malformed is flagged, not guessed.
Flaw subtypes are calibrated: causal flaws (mistaking correlation for cause) are capped at a quarter of a day's flaw questions, and the rarer subtypes — ad hominem, equivocation, part-whole, false dilemma, circular reasoning — must each appear.
Answer balance is recomputed across the whole bank. If any letter drifts more than about five points from an even 20%, it is flagged.
Subject balance is recomputed the same way, and thin domains are favored in the next batch.

Difficulty is checked against real students

A difficulty label starts as an estimate. It does not stay one. Once a question has been attempted by at least ten students, we compare the label with how students actually did. If real accuracy says the label is off by two bands or more, the question is flagged and the label is corrected.

That is still not LSAC calibration. LSAC pretests its questions on thousands of test-takers. Our labels are good enough to sort easy from hard and to build sections that rise in difficulty. Our practice scores are estimates, not predictions of your official score — for final calibration, take official PrepTests. Our page on our questions vs. the real LSAT explains what transfers and what does not.

The short version

Original questions. A fixed specification modeled on the real test. An audit every day. Difficulty corrected against real student results. And the limits stated plainly, here and on every page that matters.

Try the free diagnostic — it is built from these questions.

Last updated October 20, 2026