IELTS Writing Checker: What an Automated Check Can See, and What It Cannot

Candidates searching for an IELTS writing checker usually want one thing: paste the essay, read the band, know where they stand. Most tools will give you that number. Very few will tell you that an error count and a band are measuring different things.

This post separates the three different tools people mean by "checker", quotes the published band descriptors on the part a checker cannot see, and shows the difference on one paragraph rewritten twice. It also says plainly what Bandly's own estimates are worth, which is less than the number looks.

"Checker" means three different tools

The same search returns three categories. They answer different questions and are useful at different points in your preparation.

A spelling and grammar checker is a proofreading tool. It is genuinely useful, and it has nothing to say about three of the four things your writing is marked on.

A counter is more useful than it sounds. Word count, paragraph count and repeated word count are three of the easiest marks to lose and the easiest to check yourself, with no tool at all.

A band estimator is the thing this search usually wants. It is also the one to be most careful with, including ours.

What your writing is actually marked on

Your writing band comes from four criteria, each marked separately and each worth the same. For Task 2 they are Task Response, Coherence and Cohesion, Lexical Resource, and Grammatical Range and Accuracy. In Task 1 the first is called Task Achievement. A full explanation of each is in our guide to the band descriptors.

Notice what a spelling and grammar check touches. It touches the accuracy half of one criterion out of four. That is the whole of its reach.

The belief that costs candidates a band

Here is the thing almost every checker page implies and never states: a clean check means a good band.

The published descriptors contradict it directly, because they ask for two different things in the same criterion. Grammatical Range and Accuracy is not called "Grammatical Accuracy". These are the public wordings for three of its bands:

Band 7: uses a variety of complex structures; produces frequent error-free sentences; has good control of grammar and punctuation but may make a few errors

Band 6: uses a mix of simple and complex sentence forms; makes some errors in grammar and punctuation but they rarely reduce communication

Band 5: uses only a limited range of structures; attempts complex sentences but these tend to be less accurate than simple sentences

Read band 5 again. "Uses only a limited range of structures" says nothing about errors. An essay can have zero grammar errors and still meet that wording, because the range half is a separate judgement from the accuracy half.

Lexical Resource works the same way. Band 7 asks for a writer who "uses less common lexical items with some awareness of style and collocation". Band 6 asks for "an adequate range of vocabulary for the task". A spellchecker cannot see the difference between adequate and less common, because both are spelled correctly.

So a clean check and a modest band are not a contradiction. They are the normal result for a careful candidate who writes short, safe sentences. This is why "I made no mistakes and still got 6" is one of the most common complaints about IELTS writing, and why it is not a marking error. Our guide to complex sentences and Grammatical Range covers the structures that satisfy the range half.

One paragraph, checked twice

Here is a paragraph a grammar checker reports as completely clean.

Many people live in cities. Cities have many jobs. People move there for work. This causes problems. Traffic is very bad. Housing is also expensive.

Six sentences, no errors, no misspellings. Every automated proofreading tool passes this. Now read the descriptor wording again. Six simple sentences is "only a limited range of structures", which is the band 5 line, and no amount of error checking will reveal that.

Here is the same content, same facts, same length in words.

Because cities offer far more jobs than rural areas, large numbers of people move to them for work. This concentration creates two problems that are difficult to solve: congested roads, and housing that ordinary workers can no longer afford.

Two sentences instead of six. One subordinate clause with "because", one reference word doing real work ("this concentration"), one colon introducing a list, one relative clause ("that ordinary workers can no longer afford"). The information did not change. The structures did, and the structures are half of what that criterion marks.

A checker would report both versions as clean. Only one of them answers the range half of the descriptor.

What no automated checker can see

Beyond range, four things sit outside what any tool reads reliably.

• Whether you answered the question. Task Response is the criterion that asks this, and it is the one most commonly lost. An essay that discusses both views on a question asking for your own opinion has missed part of the task. A tool reading only your essay does not know what the question asked unless you give it the question too.
• Whether your example supports your point. An example can be well written, correctly spelled, and irrelevant to the sentence above it.
• Whether your overview is the right overview. In Task 1 the overview must state the main pattern. A tool can see that an overview exists. Whether it names the right pattern is a judgement about the chart.
• Whether your data is correct. This is the Task 1 specific one. If you write that a figure rose when the chart shows it fell, the sentence is perfectly grammatical and the mark is gone.

How to use a checker without being misled

Use it for comparison, not for the number. The number on its own has nowhere to go.

1. Write the answer under time first. A tool reading an untimed, dictionary assisted essay is measuring something you cannot produce in the exam.
2. Read the per criterion breakdown, not the overall band. The overall band is an average and averages hide the thing you needed. One criterion is almost always further behind than the rest.
3. Fix one criterion, then run the same essay again. The useful signal is the direction of change, not the absolute value. If you rewrote three simple sentences as one complex sentence and nothing moved, you have learned something either about your rewrite or about the tool.
4. Check the four cheap things by hand. Word count, paragraph count, whether you have an overview or a position, and your three most repeated words. No tool needed, and they are the most common losses.
5. Read a full answer at the level you want. A band is abstract. A paragraph is not. The sample library publishes complete answers with every scored sentence marked and the reasoning next to it, and the band 7 answers are the most useful comparison point for most candidates.

Marking your own work is a skill worth having for its own sake, because it is the only feedback available to you inside the exam room. The routine for that is in how to mark your own IELTS essay, and the error hunting half of it is in our guide to proofreading your IELTS writing.

What Bandly's own checker is and is not

Bandly marks a pasted answer against the four criteria separately, quotes the sentences behind each one, and shows the reasoning rather than just a number. That is the part we think is worth using.

What it is not: a mark. Bandly's scoring has never been compared against a qualified IELTS examiner. There is no agreement figure, because the comparison has not been run. It produces estimates from a language model reading the public descriptors, and the sensible way to use any such estimate, ours included, is as a pointer to the weakest criterion and to the sentences inside it. Treat the number as a hypothesis about where to work next, and treat a human teacher's reading as the better evidence when you can get one.

What to do today

Take the last answer you wrote. Run it through a grammar checker and fix every error it finds. Then count your sentences and mark the simple ones. If more than half are simple, you have found something no checker reported, and it is the cheapest band you will ever buy back.