A short walkthrough of how the Virtual Examiner works, and why we built it this way.
I teach IELTS in Ho Chi Minh City, and I built English Me because the hardest part of preparing for this exam isn't practising — it's finding out whether your practice is actually any good. A band score you can trust, on the day you write, changes how you study.
So English Me marks your work the way an examiner does: against the official band descriptors, one criterion at a time. Reading and Listening are checked against the real answer key. Writing and Speaking go to the Virtual Examiner — an AI marker I built and then tested against IELTS's own officially-marked scripts, so I could tell you how often it agrees rather than just claiming that it does.
A photo of Kieron belongs here — ideally teaching, or at least a warm headshot. There isn't one in brand/ yet.
Until then this block stays as a marker rather than shipping a stock image, which would undercut the point the copy is making.
Two of the four papers have a right answer, so we score them with a deterministic checker, not an AI guess. The other two need real judgement, so we fine-tuned an AI Virtual Examiner against officially-marked scripts — and measured it honestly rather than rounding up.
Every dot below is one real script or clip we've evaluated, plotted at what we actually report to the student — the raw score below band 7, or the midpoint of the range once it reaches band 7. Not an average, not a cherry-pick. The diagonal is a perfect match with the official, certified band; the shaded strip is our one-band tolerance.
Five points cross band 7, where we report a range instead of a single number — each is plotted at that range's midpoint, since that's what the student is actually graded on. The one clear outlier that remains — an official 7.5 Speaking clip scored 6.0 every time, by every version we tried — never reached band 7, so no range applied. It's the single hardest case in the whole set, discussed honestly in the callout below rather than left off the chart.
Writing and Speaking aren't answer-key tasks. Marking them means weighing vocabulary against grammar against how well an argument holds together — real judgement calls, not a lookup. A tolerance of one band is us reporting that honestly rather than rounding a score up to look more precise than the marking actually is.
Figures are internal evaluations, not a certified result. Sample sizes are still small (5 Writing scripts, 12 Speaking clips) — we're growing the evaluation set as more officially-marked material becomes available.
Built and evaluated against IELTS's own published materials: the official 2023 Academic Writing sample tasks (officially-marked scripts and examiner comments) for Writing; certified score-band clips from ielts.org's score-setting resource and the IELTS Official YouTube channel for Speaking; and the official 2023 Listening and Reading sample tasks and answer keys for the deterministic engine. IELTS is a registered trademark of the British Council, IDP: IELTS Australia and Cambridge Assessment English — English Me is an independent study platform and is not affiliated with or endorsed by IELTS or its owners.
Full mock tests, marking that mirrors the real band descriptors, and feedback from people who've examined the actual exam. Know your score before test day — then close the gap.
Section-by-section drills, accent training and the note-taking system that stops you losing easy marks.
Skimming, scanning and question-type tactics so you finish all 40 with time to check.
Task 1 and Task 2 templates, model answers, and feedback on the exact descriptors examiners use.
The Part 2 long turn, recorded in your browser and graded on fluency, range, grammar and pronunciation.
The full diagnostic-to-target journey described here (a four-skill mock, a plan generated from your test date, weekly progress tracking) is the roadmap, not the current build. This block should stay honest about which parts are live until those exist.
Writing another essay nobody marks teaches you very little. Writing one that comes back with a band for each criterion — and the reason it landed there — shows you which of the four things you're actually losing marks on, and what to fix first.
Understand why, not just what, you're losing marks on.
Work through this week's Writing lesson, get the language and structure IELTS examiners are marking for, then write the paper it's built from.
Task 1 tests whether you can describe data accurately and concisely. Task 2 tests whether you can build an argument. Here's exactly what's being marked, before you sit down to write.
You're given a chart, graph, table, process diagram or map and about 20 minutes to summarise it accurately in your own words — no opinions, no outside knowledge, just what's actually shown.
An essay question — usually a viewpoint, argument or problem — with about 40 minutes to respond directly to every part of the prompt with a clear position and supporting reasons.
None of these are tricks. They're the mistakes that cost marks most often, in the order they tend to cost them.
It's worth twice as much as Task 1, and it's the one people run out of time on. Give it the 40 minutes it deserves, then spend the remaining 20 on Task 1.
Task Response is a full quarter of your band. A brilliant essay that answers a slightly different question — or only half of a two-part prompt — is capped before the examiner reads a single sentence for grammar.
A Task 1 answer with no overview paragraph cannot reach the top bands, however accurate the detail is. Write the trends in their own paragraph, straight after the introduction, before any numbers.
Deciding your position and your two body paragraphs before you write costs five minutes and buys back far more — most Coherence marks are lost in essays that were composed live.
Pre-learned paragraphs are obvious to an examiner and are penalised under Task Response — and they rarely fit the question you actually get. Learn structures and language you can adapt, not scripts. The same goes for stuffing in "advanced" vocabulary you're not sure of: Lexical Resource rewards precision, and a misused rare word scores worse than a well-used common one.
Your script is marked by the Virtual Examiner against the same four criteria above, each judged on its own rather than blurred into a single overall impression. You get a band for each criterion and the reasons behind it — not just a number.
On our evaluation set it lands within one band of the official mark on 80% of officially-marked Task 2 scripts. Where a script reaches the top of the scale we report a range rather than a single figure, because that's the honest width of the judgement at that level.
Free to write, no account needed. Band-scored AI feedback against the real examiner criteria requires a free English Me account — launching soon.
Reading has a right answer for every question, so it's the one skill where "am I correct" isn't a judgement call. Here's the format before you sit the real thing.
Everyday or general-interest topic — the easiest passage, meant to build early confidence and momentum.
Often tied to workplace or training contexts, with moderately more demanding vocabulary and question types.
A longer, more complex academic text with abstract ideas — where most of the time pressure really bites.
Every question has one correct answer, checked against IELTS's own published answer key. There's no examiner judgement involved — just whether you found the right information under time pressure.
Reading is the paper where technique closes the gap fastest, because the answers are all right there on the page. The problem is the clock.
Don't read the passage through before you start — you don't have the twenty minutes that costs. Skim for structure, then go to the questions and read closely only where you need to.
Within most question sets, the answers appear in the same order as the text. If you've found number 6, number 7 is below it — don't go back to the top and start again.
False means the passage contradicts the statement. Not Given means the passage simply doesn't address it. Deciding from your own knowledge instead of the text is the single most common way marks are lost here.
"NO MORE THAN TWO WORDS" is a rule, not a suggestion — three words scores zero even when the meaning is perfect. Copy the spelling from the passage exactly.
There's no penalty for a wrong answer in IELTS Reading, so an unanswered question is a guaranteed nothing where a guess is a chance. Watch the clock on Passage 3 especially — if you're running out of time, fill every remaining box before the hour ends. And remember there's no extra transfer time in Reading as there is in Listening: write answers straight onto the answer sheet as you go.
This is one 13-question passage, not a full 40-question paper, so it's marked against the answer key but no band score is estimated. Free to attempt, no account needed. Full three-passage papers are in development.
You hear each recording exactly once, so Listening rewards preparation as much as raw ability — knowing what's coming next matters. Here's the format before test day.
Two people talking in a social context — like booking accommodation or asking about a course.
One speaker giving information in an everyday context, such as a talk about local facilities.
Up to four people discussing an academic or training topic, like students planning an assignment.
A monologue on an academic subject, close to a university lecture — the hardest section.
Every question is checked against IELTS's own published answer key. There's no examiner judgement — the challenge is catching the right information as it goes past, once.
You only hear it once, so almost everything that helps happens before the recording starts.
The gaps before each section exist so you can read the upcoming questions. Spend them there, not checking answers you've already written — going in cold is what makes the recording feel fast.
Before you hear it, decide whether the gap needs a number, a date, a name or a noun. Knowing you're waiting for a price makes it far harder to miss when it arrives.
"It's on Tuesday — sorry, no, Thursday." Corrections like this are deliberate, and they're everywhere. The answer is almost always the second thing said, not the first.
A misspelled word is marked wrong even if you clearly heard it. So is a plural where the sentence needs a singular. Names and places get spelled out aloud — write them down letter by letter.
The most expensive mistake in Listening isn't missing an answer — it's still thinking about it while the next three go past. Put a mark next to it, move on, and come back during transfer time. On the paper test you get ten extra minutes at the end to copy your answers across; on the computer test you get two, so write straight into the boxes as you go.
All four section scripts for the first full Listening paper are written. What's outstanding before this can go live: the four audio recordings, the campus-map graphic for Section 2, and the test page itself.
It will be free to attempt with no account needed, and marked instantly against the answer key.
English Me's Speaking test is Part 2 only — the long turn. One cue card, one minute to prepare, two minutes to speak. It's the one part of the exam we can reproduce exactly, so it's the part we do.
This is the Part 2 long turn on its own — three minutes end to end, in your browser. There's no interview and no discussion round.
On test day you'll sit 11–14 minutes with one examiner, covering all three parts below. Only the middle one is what you practise here — Parts 1 and 3 are listed so you know what else is coming.
4–5 minutes of general questions about yourself — home, work, study, familiar everyday topics — to settle you in before the harder parts.
Not yet in our practice test.
You get a cue card, one minute to prepare notes, then speak for up to two minutes on that topic without interruption.
This is what English Me practices & marks.
4–5 minutes of more abstract questions that build on the Part 2 topic — you're pushed to explain and justify opinions, not just describe.
Not yet in our practice test.
Speaking rewards the student who keeps going. Most of what costs marks here is hesitation and under-answering, not grammar.
In Part 2 you get a minute to make notes — use it, and write key words rather than sentences. Then keep going until the examiner stops you. Finishing at fifty seconds costs you marks you'll never get back.
Never answer just the question. Add a reason, an example, or a contrast — "because…", "for instance…", "although…". A two-word answer gives the examiner nothing to award a band for.
Pronunciation is marked on whether you're easy to understand — stress, rhythm and clear sounds. You are not being marked on sounding British or American, and trying to is a distraction.
Fluency counts hesitation and self-correction. Noticing an error and fixing it briefly is fine; stopping dead and restarting the sentence three times is what actually costs you.
The examiner is marking your English, not fact-checking your life. If the cue card asks about a holiday you've never taken, invent one — a fluent invented answer scores far better than an honest, halting one. What you must not do is deliver a memorised speech: examiners are trained to spot it, and it's penalised.
Parts 1 and 3 are conversations. Reproducing them honestly means an examiner who listens to your answer and asks a real follow-up — twelve minutes of back-and-forth, marked live. That's a much harder thing to build than a cue card, and a far more expensive one to run for every student who wants a practice go.
Part 2 is a monologue. You get a card, you speak for two minutes, nobody interrupts — which means a recording of your Part 2 is genuinely the same task you'd sit in the exam, not an approximation of it. It's also where the four criteria show up most clearly, because you're sustaining language on your own rather than being carried by the examiner's questions.
So we do the part we can do properly, and we say so rather than implying you've had a full mock exam. Parts 1 and 3 are on the roadmap, not pretended.
Free to attempt as many times as you like, no account needed — recordings stay on your device. Band-scored AI feedback on a recording costs 20 credits and needs an English Me account. Unlike Reading, Listening and Writing, Speaking has no free marked test — marking a two-minute recording is the most expensive thing we do.
A structured, week-by-week Task 1 Writing course — each lesson pairs a short Prepare walkthrough with a Write practice paper. New lessons publish weekly; members will get full access once accounts go live.
More weeks launching soon — the full course will sit behind a member login.
English Me is built and taught by one person, not a content farm. Everything on this site — the lessons, the practice papers, the marking standard the Virtual Examiner was built against — comes out of real IELTS classrooms in Ho Chi Minh City.
Short lessons, band-score breakdowns and exam-day advice — posted where students already are.
This block will become a live social feed once the accounts are confirmed. It needs two things that don't exist yet: the account handles/URLs, and a decision on how to embed them.
Note that embedded TikTok and Facebook feeds load third-party scripts, which conflicts with the strict content-security policy the rest of this site runs under — so the likely answer is a periodically-refreshed grid of post links rather than a native embed. Worth settling before this page goes public.
The Virtual Examiner isn't a general-purpose chatbot asked to guess a band. It was built by someone who marks to the official descriptors, then measured against IELTS's own published scripts until the numbers held up.
Kieron to supply: teaching background and years in the classroom, IELTS examining/marking credentials, qualifications (CELTA/DELTA or equivalent), time in Vietnam, and the short version of why English Me exists.
Also needed: a photo for this page. There isn't one in brand/ yet.
Most AI feedback tools will happily award you a band. Very few will tell you how often that band matches the one a real examiner gave the same script — because most have never checked.
Feedback is only worth having if someone checked it against the real thing.
Grammar you'll actually remember, study routines that survive real life, and honest advice about exams.
You don't lose a band to one big error — you lose it to five small ones repeated across every paragraph. Here's what examiners see most, and the tiny habit that fixes each.
One picture and three examples that make "I have eaten" vs "I ate" click for good.
Four filler phrases that buy you thinking time without sounding lost.
The science of spaced practice, and how the app builds it in for you.
A timing plan and three question-types worth doing last.
Learn words in the sentences you'll actually reuse — with a simple template.
How context and tone change what a sentence actually means — and why word-for-word translation misses it.
Short and practical. No spam, unsubscribe any time.
Last updated: July 15, 2026. This page covers the English Me website and services. Plain language first, details below.
English Me proofreads sentences from the panel you open — nothing is read from any other page or app on your screen, even while it floats above them. Every correction and translation is generated by an AI model, so your text is sent over the network. Where it goes depends on how you're using the panel:
Your OpenRouter API key (if you add your own), chosen writing style, and translation-language preference are saved in your browser's local storage, directly on your device. This data is never synced to us and isn't accessible to English Me — only to the panel running in your own browser, and to OpenRouter when your key is used to authenticate a request you make.
While the panel is in beta, if something breaks we may send a bare error report — an error message, HTTP status code, model name, and a stack trace from our own code — to our error-tracking tool, Sentry. We deliberately never include the text you typed in these reports. This diagnostic reporting is a beta-only measure and is planned to be removed once the panel leaves beta.
No account or email signup, no browsing history, no content from other pages or apps on your screen, no advertising identifiers, and no analytics that track you across sites. This website (englishmeapp.com) does not run any third-party analytics or ad-tracking scripts.
Saves your settings and API key in your browser, on your device, so you don't have to re-enter them every time. Nothing else on your computer is touched.
Detaching the panel uses your browser's built-in Picture-in-Picture feature, triggered only when you click the button. It does not grant access to any other tab, app, or window.
The panel only talks to openrouter.ai (or our own relay when you're on the free tier) to process corrections/translations, and sentry.io for beta error reporting. It reaches no other site.
Processes the text you submit for correction/translation, under its own privacy policy at openrouter.ai. We don't control what OpenRouter or the underlying model provider retains — check their policy if you want details on their side.
Only used on the free tier. Forwards your request to OpenRouter and enforces the daily allowance. It's our own code, but it doesn't log or store the text you submit.
Receives beta-only error diagnostics (never your typed text) under its own privacy policy at sentry.io. This integration is removed once the panel leaves beta.
We don't run a backend that stores your text, so there's no English Me database of it to retain or delete. Your local settings and API key live in your browser's storage until you clear them yourself — through the panel's settings, or your browser's own site-data controls. Any text sent to OpenRouter is retained only per OpenRouter's own policy, not ours.
English Me is not directed at children and we do not knowingly collect personal information from children.
If this policy changes, we'll update the "Last updated" date at the top of this page. Meaningful changes to how your data is handled will be reflected here before they take effect.
Privacy questions, data requests, anything else — email support@englishmeapp.com.