There are no essays on the NextGen bar exam. The MEE is gone with the legacy UBE, and nothing with its shape replaced it. What the exam does instead is test writing in three smaller, more specific ways — short answers inside counseling question sets, one medium-length drafting answer inside drafting question sets, and one performance task per section — and grades all of it against fixed scoring guides rather than against the other people in the room. Written work is 51% of the score, so how it's graded is worth knowing precisely.
Everything below comes from NCBE's NextGen UBE Constructed Response Guide (published February 2, 2026), its Official Examinees' Guide for July 2026–February 2027, and NCBE's own explanation of the grading model in The Bar Examiner, all read on September 20, 2026. Where a number appears, it is NCBE's number.
What replaced the essay
| Written format | Where it appears | Points | Expected time | Partial credit |
|---|---|---|---|---|
| Short answer — a stated number of answer fields, about a sentence each | Counseling question sets (with multiple-choice components); the research questions of a legal research performance task | 0–2 per question | Inside a 24-minute set | Yes |
| Medium answer — one drafting or editing response | Drafting question sets; the writing assignment of a legal research performance task | 0–8 | Inside a 24-minute set / a 60-minute task | Yes |
| Extended response — a memo section, brief section, letter or similar | Standard performance tasks, one per section | 0–24 | ~60 minutes | Yes |
Two shapes are worth picturing. A counseling set gives you a client scenario, embedded materials — statute excerpts, an opinion, a police report, an interview transcript, emails — and six components mixing select-one and select-two multiple-choice questions with short-answer questions on counseling, advising, negotiation and dispute resolution. A drafting set gives you one medium-length job: either a draft document with errors to identify and explain or fix, or a document (or part of one, NCBE's example is an affidavit) to write from the materials. Each set is designed to take 24 minutes, and NCBE's guide says the components are meant to be answered in order — nothing later in a set is needed for an earlier component.
The performance task is the closest thing to a familiar format: a closed universe of a case file and a library, and one extended writing assignment. Its research variant front-loads multiple-choice and short-answer questions about the authorities, then asks for a medium-length piece "similar to a standard PT but smaller in scale." Partial credit is available everywhere, in full-point rather than half-point steps. The format guide places each of these in the three-hour section; the sample questions post shows one of each.
What the graders are holding
For every constructed-response question NCBE and its subject-matter experts write a scoring guide, and the guide has the same three parts each time:
- A rationale — the total raw points available and exactly how they're earned. NCBE's sample for a short-answer question: 2 points for two correct answers, 1 for one, 0 for a response that is incorrect, insufficient or "I don't know." Its sample for a drafting set awards, per answer, one point for identifying a specific provision that conflicts with the law or the client's objectives and a second for explaining the inconsistency "without simply restating or summarizing the text."
- A list of correct and incorrect answers, built by subject-matter experts and NCBE staff and expanded with answers seen in pilot testing and pretesting.
- Benchmarks and annotations — sample answers at each possible score, each with an explanation of why it earned that score.
Performance tasks use an analytic rubric on top of that. Each legal issue in the assignment is scored separately, on two 0–4 scales: rules and application. A 4 means "accurate and thorough," 3 "mostly," 2 "somewhat," 1 "mostly inaccurate or so unthorough that it is cursory," 0 no rules (or no application) for the issue. Across the whole response graders assess four things: legal accuracy and completeness, analytical reasoning, organization and coherence, and writing quality for the intended audience. The research task's writing assignment is scored the same way, and its short-answer questions follow the short-answer rubric.
Three issues on two 0–4 scales is exactly the 24-point standard task; that arithmetic tells you where the points are. An issue you skip is up to eight points gone; an issue you state the rule for but never apply is four.
Absolute, not relative: who grades, and how
The change NCBE emphasizes most is the model. Legacy MEE and MPT answers were graded on a relative model applied within each jurisdiction; NextGen uses what NCBE's testing staff called "an objective, absolute grading model applied across all jurisdictions" — each question "will have a set scale and a clear definition of the knowledge and skills that an examinee must demonstrate to obtain each score level on that scale." The Examinees' Guide says it plainly: responses "will not be graded or ranked relative to other examinees' responses."
Who does it: graders appointed by your jurisdiction, grading only their own jurisdiction's examinees, on NCBE's common grading platform with NCBE's rubrics and NCBE's training. The process has three fixed features:
- Calibration. Before scoring live responses, graders score pre-scored sample responses for each question; more pre-scored responses are embedded throughout the grading window so drift gets caught.
- Double grading. Every constructed response is graded independently by two graders. If the two scores fall within the question type's allowed range of agreement, they're averaged.
- Reconciliation. If they don't, the response goes to a team leader or a group of graders, who review it against the scoring guide and set the final score.
So the answer to a question that has been showing up in search — are you graded against the whole country? — is no, twice over. Your written work is scored against a fixed guide, not a cohort. Your final number is then equated so a scaled score means the same thing in every jurisdiction and administration, which adjusts for exam-form difficulty, not for who else sat.
The instructions that cost points
The instructions NCBE shows during the exam are short, and the guide reproduces them. Several are point rules in disguise:
- One answer per field. Each short-answer component tells you how many answer fields to use. "Multiple answers submitted in a single answer field will be graded as only one answer." A blank field receives no credit.
- Recommended length is a floor for effort, not a ceiling. Each component states a recommended length, and "you will not be penalized for exceeding the recommended length." The sample sets recommend about a sentence per field.
- Copy-paste is allowed and can zero you. You may paste from the materials into your answer, but "an answer with excessive copying and pasting will not receive credit, even if the answer includes information that would otherwise be correct." NCBE's test: copying is excessive when the answer fails to show you understand which part is the correct part.
- Answer in order; assume the client's facts are true; don't invent facts. Components are built to be answered as asked; facts given by the client are true though possibly incomplete; and you are told not to assume facts that aren't given.
- Closed universe, defaults included. Unless told otherwise the scenario is set in the fictitious state of Franklin (Fifteenth Circuit, with Columbia and Olympia as neighbors); "Constitution" means the federal one; the Federal Rules of Evidence and Civil Procedure, the jurisdiction and venue sections of Title 28, and UCC Articles 1 and 2 apply. Law may come from the provided resources, your recalled knowledge, or both.
- The key words have definitions. NCBE defines ten terms that recur in prompts — argument, legal argument, dispositive fact, explain, factor, implication, issue, reason, specific, support. "Support" does not require a fact to be dispositive; "specific" means tied to the facts given; "explain" means give reasons. Its own tip is to learn the list so you can search the materials for exactly what was asked.
For the performance task, NCBE's development team adds five tips in the guide, and they read like a rubric in prose: address only the issues the prompt lists (no credit for extra issues, lost credit for missed ones); include only the document sections asked for; assume no facts or law outside the file and library; follow a traditional IRAC/CREAC/CRAC structure, because every assignment is testing legal analysis whatever document it wears; and write in the tone the prompt sets, to the audience it names.
A short answer, graded
Here is a counseling-set component from our bank, from the free set The Coffee Supply Dispute (Contracts and Business Associations), with the scoring logic laid next to it.
Prompt. You are advising whether Harbor — not merely Dana — is bound. Identify two facts in the file that show Dana had apparent authority to bind Harbor Coffee Co. to the order. Provide one answer in each answer field. The length of each answer should be about one sentence.
Model answer. (1) Harbor holds Dana out to the public as its "General Manager," including on Harbor's website — a position that customarily carries authority to buy supplies for the business. (2) Dana had placed and signed Harbor's supply orders with Riverbend before, and Riverbend knew Dana was Harbor's general manager, so Riverbend reasonably relied on Harbor's own manifestations in dealing with Dana.
How it scores. Two fields, two points. Credit goes to any two facts that rest on Harbor's manifestations to Riverbend — the General Manager title, the two years of running the cafés and ordering their supplies, the earlier signed orders, Riverbend's knowledge of Dana's role. A fact resting on Dana's own claim of authority earns nothing, because it isn't apparent authority. Two good facts in one field earn one point. A paragraph in each field earns the same two points as a sentence in each. That is the whole format: find the fact, tie it to the rule the question turns on, put one in each box.
A performance-task rubric, in practice
Our standard tasks are built on the same anatomy as NCBE's generic rubric — each issue scored on rules and application — so a candidate can self-grade the way a jurisdiction grader would. From The Withdrawn Stall, the free task (Contracts, a client advice letter on promissory estoppel), the rubric has three issues, each scored 0–4 for rules and 0–4 for application, 24 points in all:
- Whether the market's assurance was a promise clear and definite enough to support promissory estoppel — rules: the requirement of a clear and definite promise and the line between a promise and a statement of intention or an invitation to keep negotiating; application: the email fixing the stall, the rent and the term, weighed against "final terms to follow."
- Whether the client's reliance was reasonable and foreseeable — rules: actual, reasonable reliance the promisor should expect to induce, unreasonable where essential terms are known to be open; application: the instruction to "get ready," the irrevocable notice on the old space, the counter commissioned for the new one.
- Whether injustice can be avoided only by enforcement, and whether the Statute of Frauds bars the claim — rules: the detriment element and the estoppel exception to the statute; application: what the client spent and gave up.
Read that against NCBE's tips and the scoring becomes predictable. Miss the Statute of Frauds because the memo didn't list it? It did list it — so that's up to eight points. State the promissory-estoppel elements beautifully and then never touch the email's wording? Four points on rules, one or two on application. Write a full letter with a greeting, a facts section and a closing when the assignment asked for the analysis body only? No credit for the extra sections, and the minutes they cost come out of the analysis that is graded. The 60-minute attack method is built on this arithmetic.
What this changes about practice
If written work is half the score and graded against fixed guides, practice has to look like grading. Three habits follow from the guide:
- Write to the field count and the call. In sets, one answer per box, tied to the rule in the prompt; in tasks, the issues listed and nothing else, in IRAC/CREAC, in the tone asked for.
- Self-grade against a rubric, not a feeling. A written answer can feel complete while missing the two things the scoring guide pays for. Every question set and performance task on our platform carries a model answer and an issue-by-issue rubric on NCBE's scales; NCBE's own sample scoring guides are the calibration set to start with.
- Keep the clock in the drill. Twenty-four minutes for six components is four minutes each; sixty for a task. The first cohort's reports were about time more than about content, and written formats are where the minutes go.
The written half of NextGen is not a return of the essay. It's a set of smaller, better-defined jobs with the answer key published in advance — which is the most useful thing NCBE has told candidates about the exam so far.
