AI IELTS Writing Checker Accuracy: Test Method

A single number feels like an answer, and it is the wrong object to test. An official Writing band is awarded by trained examiners using the published descriptors, on test day, under the test instructions. This site states on its methodology page that the product is a diagnostic and revision tool, not a replacement for the official exam, and that a band estimate is there to reveal a likely bottleneck rather than to guarantee an official score. The editorial policy repeats the limit: the site does not promise a specific official result. Any accuracy claim that skips those sentences is larger than the product.

The test that fits those limits uses one script. Read it against the public descriptors before you open a tool. Decide which criterion it fails most clearly. Then see whether the checker names that criterion and a next change that would move the script toward the next descriptor line. Revise only that change, read the descriptor again, and repeat on a second script before you treat one run as a pattern. That is the method behind the IELTS Writing Checker.

What the public descriptors let you test

IELTS publishes separate Writing descriptors for Task 1 and Task 2. Both use four criteria. Task 2 calls the first criterion Task Response. Academic Task 1 calls it Task Achievement, because the task is to summarize visual information rather than to argue a position. The other three criteria have the same names in both tasks: Coherence and Cohesion, Lexical Resource, and Grammatical Range and Accuracy. The band descriptors guide on this site keeps that map in one place. The test method uses the public wording, not a private scale.

For Task Response, Band 6 addresses all parts of the task, though some parts may be covered more fully than others. The position is relevant, though conclusions may become unclear or repetitive, and some main ideas may be inadequately developed. Band 7 addresses all parts, keeps a clear position throughout, and presents, extends, and supports main ideas, with room for some over-general support. Under this method, a comment is on target when it describes a thin idea or an unclear position in those terms and the revision adds coverage, a stable position, or developed support.

For Academic Task Achievement, Band 6 presents an overview with information appropriately selected, and it allows that details may be irrelevant, inappropriate, or inaccurate. Band 7 presents a clear overview of the main trends, differences, or stages, and it clearly presents and highlights key features. A comment that only corrects spelling, while the overview is missing or a figure is wrong, has not been tested against Task Achievement. The chart-upload article walks through that figure check on its own: IELTS Task 1 checker with chart upload.

Coherence and Cohesion at Band 6 has clear overall progression, while cohesion may be faulty or mechanical and referencing may be unclear. Band 7 is logically organised, with cohesive devices that may still be slightly under-used or over-used. Lexical Resource at Band 6 is an adequate range; less common words may be inaccurate. Band 7 has enough range for some flexibility and precision, with occasional errors. Grammatical Range and Accuracy at Band 6 mixes simple and complex sentences, with errors that rarely reduce communication. Band 7 produces frequent error-free sentences and keeps a few errors inside otherwise good control.

Those lines are the whole marking key this method is allowed to use. A comment that invents a fifth criterion, or that praises “advanced words” while word choice is inaccurate, fails the test even if the prose sounds confident. The plain-English descriptor page is the shorter companion when you want the Band 6, 7, and 8 differences without leaving the site.

The four steps the checker is built to follow

The methodology page states the product workflow in four steps. The accuracy method checks whether a report you receive still follows them.

1. Prompt and essay intake

The system checks essay type, task intent, and length before it explains a score, so grammar-only feedback is not treated as an IELTS score. A discuss-both-views essay that receives only comma comments has missed this step. An Academic Task 1 report that is asked for a personal opinion has missed the task type. Length belongs here too: the task instructions set a minimum, so idea comments on a response far under that minimum are not a full band judgement.

2. Four-criteria diagnosis

Feedback is organised around Task Response, Coherence and Cohesion, Lexical Resource, and Grammatical Range and Accuracy. For a Task 1 report, read the first of those as Task Achievement. The free diagnosis shows the likely band direction, the weakest criterion, and the main reason this draft is held back. Deeper feedback adds criterion evidence and a clearer rewrite. The page also displays three design signals: four criteria, one next fix, and a band direction marked “+/-0.5” that is designed to stay useful for revision decisions. Treat that label as the page’s design statement, not as a measured trial error.

3. Sentence and pattern review

The tool looks for repeated weaknesses: a missing overview, weak idea support, repetitive vocabulary, and unstable grammar in comparative or complex sentences. One odd sentence is a slip. The same thin example in both body paragraphs, or the same trend verb on every line, is a pattern. A missing overview is Task Achievement. Weak support is Task Response. Repetitive vocabulary is Lexical Resource. Unstable complex sentences are Grammatical Range and Accuracy.

4. One next revision

The stated goal is the next fix. Ignore comment length and look for one change that matches the weakest descriptor line. If the position disappears in the conclusion, the revision restates that position. If the overview is missing, the revision states the main trend or difference and leaves unverified figures out. A list of synonyms, offered while the position is unclear, misses this step.

How to run the method on a script you wrote

Before you open the tool, write four short notes, one per criterion, using only the descriptor language above.

• Task Response or Task Achievement: which part of the question is thin, and is the position clear throughout, or is the overview clear and the detail accurate?
• Coherence and Cohesion: can a reader see a new point in each paragraph, or does the linking feel mechanical?
• Lexical Resource: where does word choice become vague or inaccurate, as opposed to merely simple?
• Grammatical Range and Accuracy: do errors appear in many sentences, or only as occasional slips inside otherwise controlled complex sentences?

Circle one criterion. That circle is your prediction. Then open the checker and submit the same text you just marked. Compare two outputs only: the weakest criterion, and the revision it tells you to make. Agreement means the tool selected the criterion you circled, or one you can defend with a sentence in the script. A percentage after one essay is outside this method.

Make that one revision and leave the rest of the script alone. Read the same descriptor line and ask whether the feature it names has changed. Inadequate development changes when the main idea gains a specific extension that stays on the question. A grammar problem changes when more sentences are error-free and the complex sentences you kept stay under control.

Repeat with a second script of the same task type. The methodology page says to compare several pieces of writing and to look for a repeated weak criterion before you change a study plan. Two scripts that both fail Task Response, with the checker naming Task Response both times, are a reason to practise development. Two scripts that fail different criteria are a reason to keep testing. The page on moving from 6.5 toward 7 uses the same order: find the criterion that is blocking the next band, then revise that criterion.

What a passing comment looks like, criterion by criterion

These are checks on the comment, not sample scores. No script is awarded a band in this section.

A Task Response comment passes when it names the thin part of the question and the rewrite adds support that stays on that part. If the question asks for causes and solutions, a note about the introduction fails while the body never presents solutions. A note that extends the solutions paragraph passes. The method asks the tool to aim at what Band 7 describes: ideas presented, extended, and supported, with a position that stays clear.

A Coherence and Cohesion comment passes when it talks about progression and paragraph roles. Band 6 already allows cohesive devices; the risk it names is mechanical cohesion and unclear referencing. Adding “moreover” to every sentence moves toward that risk. Giving each paragraph one main idea, and making “this” refer to a noun the reader can find, matches Band 7’s demand for logical organisation and clear progression.

A Lexical Resource comment passes when the replacement is precise and still means what you wrote. Band 6 is where less common vocabulary is attempted with some inaccuracy, so a rarer word that changes the claim fails that line. Band 7 asks for flexibility, precision, and some control of collocation. Keep an adequate word when the offered collocation does not fit the question.

A grammar comment passes when it separates a slip from a loss of control. Band 6 expects a mix of simple and complex forms and accepts errors that rarely reduce communication. Band 7 expects frequent error-free sentences and only a few errors. Correct simple sentences are already inside the criterion. Range asks for some complex structures, and a repaired complex sentence that keeps the same meaning is the useful comment.

Limits that stay in force after a good run

Agreement on one script is evidence about that script. It is not an examined band and it is not a forecast of test day. The methodology page says the product is not the official exam, the estimate is not a guarantee, and one draft is not a final prediction. Stronger examples, clearer structure, and better idea support matter more than surface polishing.

The free diagnosis is the right first sample because it shows direction, the weakest criterion, and the main blocker. If you cannot find that criterion in the descriptors, reread the script before you continue. If you can, decide whether you need the deeper evidence the methodology page separates from the free diagnosis. The comparison of free and paid checker feedback describes that product difference. It is a separate decision from the accuracy test. Pay for more detail only after the free pass has named a real descriptor problem in your own writing.

Two further boundaries keep the method honest. The site is not affiliated with IELTS, the British Council, IDP, or Cambridge English, and the descriptors used here are the public assessment criteria, paraphrased for study. A checker comment, including ours, is study material you can accept or reject by pointing at a sentence and a descriptor line. When you want to apply the method to the script you are revising today, use the IELTS Writing Checker on the finished text, write down the criterion you expected, and revise once before you write anything new.