IELTS Writing Mistakes: Which of the Four Criteria Each One Is Marked Under

Every page about IELTS writing mistakes gives you the same flat list. Misreading the question, weak paragraphs, too few words, bad grammar, bad spelling, memorised phrases. The list is not wrong, but it is useless in the order it arrives, because it does not say what each mistake costs.

IELTS writing is not marked as one score. It is marked under four separate criteria, and each criterion is worth a quarter of your writing band. A mistake only takes marks off in the criterion that covers it. That single fact changes the whole list: two mistakes that look equally careless can cost you nothing and half a band.

This guide sorts the common mistakes by the criterion that marks them, then shows what a count of real marked answers says about which ones to fix first.

The four criteria, in one line each

IELTS writing is marked against four criteria, each carrying equal weight:

1. Task Response in Task 2, or Task Achievement in Task 1. Did you answer the question that was asked, fully, with a clear position?
2. Coherence and Cohesion. Can the reader follow the order of your ideas, paragraph by paragraph and sentence by sentence?
3. Lexical Resource. Vocabulary. Range, precision, spelling and word formation.
4. Grammatical Range and Accuracy. Sentence structures, verb forms, articles and punctuation.

The four marks are averaged, so each one moves your writing band by up to a quarter of the difference it moves on its own. A criterion that drops from 7 to 5 pulls the writing band down by half.

The sorted list

Read down the right hand column and something becomes clear. Spelling sits in Lexical Resource, not in grammar, which is why spelling mistakes behave differently from the way candidates expect. Punctuation sits in grammar, not in coherence. And word count is a Task Response matter, because the criterion is about how fully you answered, and a short Task 2 answer cannot develop enough to satisfy it.

The three mistakes that are not really mistakes

Three items that appear on almost every list belong nowhere in the table above, because they do not match any descriptor.

**Using I in an academic essay.** Nothing in the public descriptors forbids first person. A clear position stated as I believe that satisfies Task Response exactly as well as an impersonal construction.

Contractions. No descriptor mentions them. They read informally, which can matter in a Task 1 letter where tone is assessed, and in a Task 2 essay they are a style preference rather than an error.

Handwriting. Your handwriting is not a criterion. It matters only when a word becomes unreadable, at which point it is marked as a spelling or vocabulary problem, because the reader cannot tell what you wrote.

Spending exam minutes on any of these is spending minutes on nothing.

What a count of real marked answers says

Here is the part the ranking pages leave out. They tell you that a mistake is common without telling you which criterion is actually holding candidates back.

I counted it in Bandly's own published sample library. There are 25 samples with all four criterion estimates on them, across Task 1 Academic, Task 1 General Training and Task 2. For each answer I took the four estimates and found which criterion came back lowest.

Lexical Resource was the weakest criterion, alone or tied, in 23 of the 25 answers. Task Response was never the weakest in any of them.

That points at the bottom third of the table above. Vocabulary mistakes, the ones candidates treat as cosmetic, sat under the criterion that came back lowest almost every time, while the question handling mistakes that every list puts first scored highest.

Two cautions, both important. These bands are estimates produced by Bandly's own scoring engine, which has not been checked against a qualified human examiner. And the Task Response column is the one to trust least: an estimate that never once falls below 7 across 25 different answers is more likely to be telling you something about the estimator than about the writing. The vocabulary finding is the useful half, and it agrees with what the descriptors say, because Lexical Resource is the only criterion that penalises range, precision and spelling all at once.

One paragraph, with only the expensive mistakes removed

This is a band 6 body paragraph of the kind the count above describes. Task handling is fine. Vocabulary is doing the damage.

Nowadays, in the modern world of today, there are many persons who think that goverments should spend more money on public transportation. Firstly, public transportation is good. It is cheaper than cars and it is good for the enviroment. Also, public transportation is good for traffic because many people can use one bus. So, in conclusion, public transportation is a good thing for a city.

Now the same paragraph with nothing changed except what Lexical Resource marks. The ideas, the order and the sentence structures are left alone.

Many people argue that governments should fund public transport more heavily. The first reason is cost: a bus fare is a fraction of what running a car costs over a year, and the emissions per passenger are lower. A full bus also takes dozens of private vehicles off the road at rush hour, which eases congestion without any new road being built. Public transport is therefore cheaper for the household and cheaper for the city at the same time.

What changed, and under which criterion:

1. Nowadays, in the modern world of today removed. A memorised opener that says nothing, marked as inappropriate for the task under Lexical Resource.
2. persons, goverments, enviroment corrected. Word formation and spelling, Lexical Resource.
3. good used four times, replaced with cheaper, lower and eases. Range, Lexical Resource.
4. dozens of private vehicles, rush hour, per passenger added. Precision, which is what band 7 asks for in this criterion.
5. So, in conclusion removed from a body paragraph. That one is Coherence and Cohesion, an inaccurate cohesive device, and it was too cheap to leave in.

Not one sentence was made longer or more complex, and no new idea was introduced. The paragraph still has the same four thoughts in the same order. Grammatical range was deliberately untouched, so whatever that criterion scored before, it scores now.

What to do today

Take an answer you have already written and mark it four times, once per criterion, reading only for that criterion each time. Four passes over one essay teaches you more than one pass over four essays, because the four passes are looking for different things and the single pass is looking for everything and finding nothing.

Then look at which pass produced the longest list. That criterion is where your next band is, and the mistakes in the other three can wait. If you want a second opinion on which of the four is weakest, paste the same answer into Bandly and compare its four estimates with your own four marks; the estimates are not examiner checked, but the disagreements are the part worth reading.