This page documents the criteria LingoLeap scores a written answer against and the exact fields the report hands back, task by task. It is a specification, not a sales page: every criterion named below is the one the grader is instructed to apply, and every field named below is one the result screen can display.
By the LingoLeap Research Team · Published:
You get five things. First, a task score from 0 to 5 in half-point steps, broken into the four criteria that belong to that task: content, syntactic and lexical variety, social conventions and accuracy for Write an Email; content and elaboration, response to the discussion, syntactic and lexical variety and language accuracy for Write for an Academic Discussion. Each criterion carries its own written comment and, where there is something to fix, quoted extracts from your own answer with a suggested rewrite and an explanation. Second, a minimal-edit grammar and spelling pass that lists each error as a one-to-three-word original, the correction, a reason and an error type. Third, a rewritten version of your answer at the level of a 5, together with the specific changes that got it there. Fourth, a vocabulary analysis: a CEFR distribution across A1 to C2, your total word count, your text annotated level by level, and a list of lower-level words with higher-level replacements. Fifth, a sentence-variety analysis that classifies every sentence as simple, compound or complex, records its opener type and length, and suggests rewrites; plus a mindmap and keyword list built from the rewritten answer. Alongside all of it the report carries your word, sentence and paragraph counts and the score converted to the 0-30 scale, the 1-6 band, CEFR and an IELTS equivalent.
The 2026 TOEFL Writing section has three task types and twelve questions in roughly 23 minutes, presented in a fixed linear order. Only two of the three are graded against a rubric, and those two are the ones this page documents. Nothing written for the older independent or integrated essay describes what you will be scored on now.
Build a Sentence (~6 minutes)
You arrange given words or word groups into one grammatically correct sentence. It is scored correct or incorrect, not on a 0-5 rubric, so there is no per-criterion written feedback to describe.
Write an Email (~7 minutes)
You read a campus scenario and write a short email for a specific purpose and audience. Scored 0-5 in half-point steps against four criteria.
Write for an Academic Discussion (~10 minutes)
You read a professor's prompt and two student replies, then write your own contribution. Official 2026 materials say an effective response contains at least 100 words. Scored 0-5 in half-point steps against four criteria.
Timings and the task inventory here are the ones published on our own TOEFL Writing 2026 guide, which sources them from official 2026 practice materials. /toefl/writing
Submitting an answer runs a fixed sequence of six steps. The report fills in as each one finishes, which is why the score can appear before the vocabulary and sentence panels do.
1. Scoring
The answer, the task prompt and a set of response statistics go to the grader with the full 0-5 rubric for that task type. It returns an overall score, the four criterion scores, a written comment on each and quoted examples.
2. Grammar and spelling pass
A separate, stricter pass lists errors as minimal edits. The instruction is explicit that each correction changes only one to three words and does not rewrite or rephrase the sentence.
3. Revision
A rewritten version of your answer at the level of a 5, built from your own content rather than a generic model. For the email task the instruction forbids placeholders such as [Last Name], so the rewrite uses the names the scenario actually gives.
4. Vocabulary analysis
Word-level CEFR classification. Where the CEFR dictionary is available it does the classification and the model is used only for the upgrade suggestions; otherwise the model does both.
5. Sentence-variety analysis
Every sentence is classified by type, opener and length, connectors are sorted into basic and advanced, and rewrites are proposed for the weakest ones.
6. Mindmap and keywords
An outline of the rewritten answer plus a keyword list, each entry carrying an IPA pronunciation, part of speech, definition and the exact sentence from the rewrite where the word appears.
The email grader is instructed to apply the four official 2026 scoring dimensions. The name on the left is the rubric dimension and the field name in the result; the name in brackets is the label the report prints above it.
content — shown as "Communicative Purpose"
Whether your elaboration supports the communicative purpose of the email. At 5 it effectively supports it; at 3 it only partially supports it; at 2 it is limited or irrelevant.
syntactic_lexical_variety — shown as "Language Facility"
Syntactic variety and idiomatic word choice. At 5 the variety is effective and the word choice precise and idiomatic; at 3 the range of syntax and vocabulary is moderate; at 1 the language is telegraphic with very limited vocabulary.
social_conventions — shown as "Social Conventions"
Politeness, register, organisation of information, and how you formulate actions such as requests, refusals and criticisms. At 5 these are consistently appropriate; at 3 there are some noticeable errors; at 2 they are inconsistent or inappropriate.
accuracy — shown as "Grammar & Mechanics"
Lexical and grammatical accuracy. At 5 there are almost no errors, with only the common typos expected under timed conditions; at 4 there are few; at 2 errors accumulate across sentence structure and language use.
For the email task the grader is told explicitly that the overall score is not a simple average of the four criterion scores, and that it is limited by weaknesses in vocabulary and social conventions: the overall score should not exceed the lower of syntactic_lexical_variety and social_conventions by more than half a point. Correct grammar on its own does not produce a high score. That is why an answer can come back with a strong accuracy figure and a middling overall.
scoreThe overall task score, 0-5 in half-point steps.content, syntactic_lexical_variety, social_conventions, accuracyThe four criterion scores, each 0-5 in half-point steps.[criterion]_feedbackOne written comment per criterion — four in all.[criterion]_examplesPer criterion, a list of objects carrying original, suggestion and explanation, quoting your own text. The grader is instructed to return an empty list rather than invent an example when there is nothing genuine to fix, so an empty panel means no finding, not a missing one.overall_feedbackA summary comment, written in your interface language.revised_emailYour email rewritten to the level of a 5, using the names and details the scenario supplies rather than placeholders.grammar_errorsA list of objects carrying original_text, corrected_text and explanation, from the revision step.word_choice_improvementsA list of objects carrying original_text, improved_text and explanation.social_convention_issuesA list of objects carrying issue and suggestion — the register and politeness problems specifically, kept separate from grammar.The discussion grader applies a different set of four dimensions. One of them has no counterpart in the email task: whether you actually engage with what the other students wrote.
content_and_elaboration — shown as "Content & Elaboration"
Whether you give relevant, well-elaborated explanations, examples or details to support your position. At 5 the elaboration is well developed with relevant examples; at 3 it may be missing, unclear or irrelevant in parts.
response_to_discussion — shown as "Response to Discussion"
Whether you contribute to the ongoing discussion in your own words, typically by responding to the arguments in the short texts you were given. At 4 there is real engagement with others' arguments; at 2 that engagement is limited and the other students may be largely ignored.
syntactic_lexical_variety — shown as "Syntactic & Lexical Variety"
Variety of sentence structures and precise, idiomatic word choice. At 5 complex syntax is used effectively; at 2 the range of syntax and vocabulary is limited.
language_accuracy — shown as "Language Accuracy"
Lexical and grammatical errors. The rubric states that a top score requires a response with almost no errors; at 3 errors in sentence structure, word forms or idiomatic language are noticeable.
Where the email grader is given an arithmetic ceiling, the discussion grader is told the overall score reflects a holistic assessment of all four dimensions: strong content with adequate language can reach the 3.5-4 range, and excellent content with sophisticated language the 4.5-5 range. The two tasks therefore combine their criterion scores in genuinely different ways, and the same four subscores would not necessarily produce the same overall on both.
scoreThe overall task score, 0-5 in half-point steps.content_and_elaboration, response_to_discussion, syntactic_lexical_variety, language_accuracyThe four criterion scores, each 0-5 in half-point steps.[criterion]_feedbackOne written comment per criterion — four in all.[criterion]_examplesPer criterion, a list of objects carrying original, suggestion and explanation, quoting your own text, or an empty list where there is nothing genuine to fix.overall_feedbackA summary comment, written in your interface language.revised_responseYour contribution rewritten to the level of a 5, at roughly the target length for the task.grammar_errorsA list of objects carrying original_text, corrected_text and explanation, from the revision step.content_improvementsA list of objects carrying issue and improvement, saying what was lacking in the original and how the rewrite addressed it.Four more panels are produced for both rubric-scored tasks. They are the same for the email and the discussion, because the same modules run on both.
Minimal grammar corrections
Each item carries original, corrected, reason and type, where original and corrected are one to three words. This pass is deliberately separate from the criterion feedback: it is a proofreading list, not an assessment.
Vocabulary analysis
vocabulary_distribution gives the share of your words at each CEFR level from A1 to C2; total_words gives the count; annotated_text returns your own text with levels marked inline; vocabulary_upgrades lists lower-level words with a higher-level replacement, both levels and an explanation; vocabulary_level_summary is a short written assessment.
Sentence-variety analysis
sentence_analysis classifies every sentence as simple, compound or complex, records its opener type and word count, and attaches a rewrite where one is needed. Alongside it come sentence_type_distribution, opener_distribution, sentence_lengths, average_sentence_length, a verdict on length variety, a connector_analysis splitting the connectors you used into basic and advanced, a variety_score and a ranked top_improvements list.
Mindmap and keywords
An outline of the rewritten answer — for an email, greeting, self-introduction, purpose, background, supporting details, request, closing and sign-off — plus a keyword list where each entry carries the word, its IPA pronunciation, part of speech, definition and the exact sentence from the rewrite that uses it.
Counts and conversions
The report also carries n_words, n_sentences and n_paragraphs for your answer, and the score expressed four ways: raw_score out of 5, score out of 30, band_score out of 6, and cefr, plus an ielts equivalent.
A task score is converted in three steps: 0-5 to the familiar 0-30 scale, 0-30 to the 2026 band scale of 1-6 using the writing-specific table, and the band to a CEFR level. The table below is computed at build time from the same constants the application uses, so it cannot drift from the app.
| Task score (0-5) | 0-30 scale | 2026 band (1-6) | CEFR |
|---|---|---|---|
| 5.0 | 30 | 6.0 | C2 |
| 4.5 | 27 | 5.5 | C1 |
| 4.0 | 25 | 5.0 | C1 |
| 3.5 | 22 | 4.5 | B2 |
| 3.0 | 20 | 4.0 | B2 |
| 2.5 | 17 | 4.0 | B2 |
| 2.0 | 14 | 3.0 | B1 |
| 1.5 | 11 | 2.5 | A2 |
| 1.0 | 8 | 2.0 | A2 |
| 0.5 | 4 | 1.5 | A1 |
| 0.0 | 0 | 1.0 | A1 |
Because the 0-30 band table treats 17 to 20 as a single band, a 2.5 and a 3.0 both land on band 4.0. The gap between them is real on the 0-30 scale and invisible on the band scale. If you are tracking progress, track the raw 0-5 criterion scores, not the band.
ETS publishes the 0-5 writing rubric and publishes the 1-6 band scale, but has not published how raw writing performance is converted into a band. The 0-5 to 0-30 step above is our own mapping, chosen so the familiar 0-30 figure stays available; it is not an official conversion and should not be read as one. We track this and the other unpublished details at /toefl/2026-open-questions
The report is long. This is the order that gets the most out of it in the least time.
Three things are deliberately absent, and their absence is the point.
No accuracy or calibration figure
We have not published a study comparing these scores against official ETS writing scores, so this page states no agreement rate, margin of error or accuracy claim. Treat the score as practice feedback, not a prediction.
No sample score or sample feedback
There is no worked example on this page. We hold no graded 2026 writing response we are able to publish, and inventing one to illustrate the format would defeat the purpose of a specification.
No invented ETS internals
Where ETS has not published a detail — most importantly how raw writing performance becomes a band — this page says so rather than filling the gap.
Write a 2026 email or academic discussion response in a full mock test and get the criterion scores, quoted corrections, revision, vocabulary and sentence analysis described above.
Take a free mock testTOEFL 2026 writing rubrics
The official 0-5 scoring guides for both rubric-scored writing tasks.
What TOEFL 2026 speaking feedback contains
The same specification for spoken answers: criteria, returned fields and the band conversion.
TOEFL email writing template
Structure for the Write an Email task, including the register and politeness moves the social conventions criterion scores.
TOEFL Writing 2026 guide
Task inventory, timing and format for the whole writing section.
TOEFL 2026 open questions
The details ETS has not yet published, including how raw performance becomes a band.