AI feedback or a private tutor for IELTS? Where each one is actually better

This is a comparison written by a company that sells one of the two options, so it is worth saying at the start what a private tutor does better, and meaning it. Reading and Listening are not the question: they are scored objectively, and practice with an answer key is practice with an answer key wherever you get it. The question is Writing and Speaking, where somebody has to look at what you produced and tell you what it is worth. Here are the six things that decide whether that feedback moves your band, with a straight answer on each.

1. Grading against the criteria

Your Writing and Speaking bands are decided by four criteria each, published by IELTS, and by nothing else. Feedback that does not name the criterion it is about is opinion.

A good tutor who has examiner training does this well, and a trained examiner's judgement on borderline cases is still the best there is. A tutor without that training often grades on general impression: "good essay, some grammar mistakes, maybe a 6.5." That is the most common kind of tutor, and impression grading is exactly the thing the descriptors were written to replace.

AI grading is consistent by construction: the same essay gets the same band on Tuesday and Thursday, and every evaluation comes back split by criterion with the reasons. It is not a trained examiner, and on an essay that sits exactly on a boundary it will sometimes land half a band from where an examiner would — there is a procedure for checking an AI band score against the descriptors before you rely on one. What it never does is grade on impression.

Verdict: a trained examiner wins; an untrained tutor loses to a consistent criterion-by-criterion grade; you rarely know in advance which tutor you have.

2. Speed and volume of feedback

Writing improves through the loop: write, get graded, fix the lowest criterion, resubmit the same task, see the band move. The loop is only as fast as the feedback.

A tutor grades an essay when the next lesson comes round. Two essays a week is a lot for most arrangements, and resubmitting the same essay three times is asking for a favour.

AI feedback returns in minutes, at any hour, and the tenth resubmission of the same essay costs no goodwill. A candidate with an hour a night can run the loop every night.

Verdict: AI, and it is not close. This is the single largest practical difference between the two.

3. Speaking interaction

Speaking is a conversation with an examiner who reacts to you, interrupts, asks a follow-up you did not expect, and notices when you dry up.

A tutor is a person. Practising Part 3 with someone who pushes back on your answer is something no recording can fully copy, and an experienced tutor hears pronunciation problems as they happen, in context.

AI feedback grades a recorded answer against the four Speaking criteria and tells you which one is holding the band down. It does not push back mid-answer. Pronunciation is graded on what the speech recognition heard, which is a proxy for what an examiner hears rather than the thing itself.

Verdict: a tutor, clearly, for live interaction and for pronunciation. Recorded practice, graded the same way every time, is still the fastest way to find your own pattern.

4. Knowing what to work on next

Most candidates are wrong about their weakest skill by half a band or more, and spend their hours on the skill they enjoy.

A tutor builds a picture of you over weeks and will usually steer you, if you keep turning up.

A diagnostic-led plan does it from the first hour: a timed diagnostic, then a study plan that puts the hours where the band is lowest and rebuilds itself as every graded task and practice result comes in. It does not need you to be a regular to know you.

Verdict: roughly even over a long engagement; the plan wins in the first two weeks, which is often all a candidate has.

5. Accountability

A tutor is an appointment. You show up because someone is waiting, and for many people that is the only thing that gets the hours done.

An app has a plan with today's tasks on it and a streak. That works for some people and does nothing for others.

Verdict: a tutor, for anyone who knows they will not study alone. Be honest with yourself about which person you are; it matters more than anything else on this page.

6. Cost per hour of feedback

A tutor is paid by the hour, and an hour with a good one is worth it. The arithmetic problem is the research on how long preparation takes: the largest published study found roughly 200 to 240 hours of instruction for about half a band on average, with slower gains above band 6 — the full arithmetic is in how long band 6 to 7 takes. Buying that many hours of a person is beyond most candidates, so tutoring in practice means a few hours a week and unguided study in between.

AI feedback is priced by the month or by the exam window rather than by the hour, and the unguided hours between lessons are exactly where it does its work. The prices are on the pricing page.

Verdict: for the hours you cannot afford to have a person in the room, AI feedback is the only feedback available at all.

The honest summary

If your budget stretches to a good tutor and you will show up every week, take the tutor and use AI feedback between lessons. If it does not, or if what you are short of is feedback rather than company, start with a free timed test and let the results decide where your hours go.