IELTS speaking practice at home: a self-study plan that works

Most IELTS candidates prepare for speaking by doing two things: memorizing model answers and hoping. Neither moves your band score, because the examiner isn't grading your opinions — they're grading four specific criteria against published descriptors. The good news: three of the four are things you can measurably train at home, alone, for free. This is the plan.

Know the test: the 3 parts

The speaking test is an 11–14 minute face-to-face (or video-call) interview with a human examiner, in three parts:

1. 1.Part 1 — Introduction and interview (4–5 minutes). Familiar topics: your home, work, studies, hobbies. Short, natural answers.
2. 2.Part 2 — The long turn (3–4 minutes). You get a cue card, one minute to prepare, and must speak for up to two minutes without help. This is where under-prepared candidates dry up at 45 seconds.
3. 3.Part 3 — Discussion (4–5 minutes). Abstract follow-up questions related to the Part 2 topic: causes, comparisons, predictions, opinions. This part separates band 6 from band 7+, because you can't script it.

Know the scoring: 4 criteria, 25% each

Your band score is the average of four equally-weighted criteria:

• Fluency and coherence — can you keep going at a natural pace, link ideas, and self-correct without derailing?
• Lexical resource — range and precision of vocabulary, including paraphrasing when you don't know a word.
• Grammatical range and accuracy — complex structures attempted, and how often errors appear.
• Pronunciation — and this is the one most candidates train least, because until recently you couldn't get feedback on it without a teacher.

The pronunciation band descriptors, decoded

Read the public descriptors closely and the gap between band 6 and band 8 becomes concrete:

Band 6: uses a range of pronunciation features with mixed control; mispronunciations reduce clarity at times.

>

Band 8: uses a wide range of pronunciation features; sustains flexible use throughout; is easy to understand; accent has minimal effect on intelligibility.

"Pronunciation features" is examiner-speak for three trainable things:

• Individual sounds. Meaning-changing contrasts like /ɪ/ vs /iː/ (ship/sheep), /θ/ vs /s/ (think/sink), /v/ vs /w/. A survey of the hardest English sounds covers the usual suspects by first language.
• Word and sentence stress. Saying pho-TO-graph-er, not PHO-to-graph-er, and stressing the content words in a sentence. See our guide to English word stress and intonation.
• Intonation. Rising and falling pitch to signal questions, contrast, and finished thoughts. Flat delivery caps you even when every phoneme is right.

Note what band 8 does not say: it never mentions sounding native. Accent is fine; unclear sounds are not. That's exactly the line sayit's scoring draws — the engine is accent-tolerant by design, ignoring regional variation while flagging the contrasts that genuinely change meaning.

Why most app feedback fails IELTS candidates

Here's the trap with practicing into a generic transcription app: standard speech recognition is built to auto-correct your mistakes into real words. Say "I sink so" and it politely transcribes "I think so" — and tells you nothing. The error that would cost you at band 6 is invisible to the tool.

sayit uses a language-model-free phoneme recognizer instead: it scores the raw sounds you actually produced, shows a color-coded per-word heatmap, and puts the target IPA next to the sound you really made, with a concrete articulation tip. How that scoring works is worth two minutes if you're skeptical — you should be.

The self-study plan (45 minutes a day, 4 weeks)

Week 1 — diagnose

• Take a full IELTS-style speaking exam in sayit (all three parts, timed). Don't prepare; you want a true baseline.
• Run the Magic Wand coach — it auto-diagnoses your weakest phonemes across your recordings and builds a personalized drill plan. Write down your top three problem sounds.
• Do one minimal-pair session per problem sound. Look up specific words that trip you on the word-lookup hub.

Week 2 — drill sounds and stress

• 15 min: phoneme drills on your three target sounds, until the heatmap stays green across different words.
• 15 min: read-aloud mode on leveled passages, watching the stress and intonation scores, not just the phonemes.
• 15 min: one Part 2 cue card, self-timed. One minute prep, two minutes speaking, no stopping. Record it in freeform mode and review the fluency score.

Week 3 — Part 3 under pressure

Part 3 is unscriptable, so practice the skill, not answers. Use sayit's guided AI conversations to simulate Part 3-style follow-ups: the AI pushes into abstract territory ("Why do you think that's changed?", "Is that true everywhere?") and you must extend answers in real time. Aim for 30–45 second answers with a claim, a reason, and an example.

• 20 min: AI conversation on a Part 3-style topic.
• 15 min: shadowing a native-paced passage to pull your intonation range up.
• 10 min: re-drill any sound that reappeared red during conversation.

Week 4 — full rehearsals

• Two full timed IELTS-style exams this week, reviewed against the four criteria.
• Compare your score history in the progress analytics — you're looking for the pronunciation and fluency trends, not any single session.
• Keep vocabulary honest: when you paraphrase around a missing word in practice, note the word and drill its pronunciation the next day.

Everything above runs in the browser on the free tier, no card required; paid limits are on pricing.

Honest limits: where self-study stops

An app can measure sounds, stress, intonation, pace, and completeness. It cannot fully replicate two things:

• A live human examiner's follow-ups. AI conversation gets close for Part 3, but a human mock examiner reads your hesitation and probes it. Book at least one full mock with a qualified human — a teacher, a tutoring service, or an IELTS-experienced friend — in your final week. Our comparison of AI feedback vs a human coach covers exactly what each catches.
• Examiner-calibrated band prediction. Treat any app's numbers (ours included) as a training signal that tracks direction, not an official band guarantee.

Self-study gets you 90% of the way at roughly 0% of the cost. Spend your money on one or two human mocks at the end, not on months of tutoring for drills an app scores better anyway.

Start with a baseline today

The single highest-value thing you can do right now is find out which sounds are actually costing you marks — most candidates guess wrong. Record one IELTS-style answer in sayit, free, in the browser, no card, and you'll have your per-phoneme diagnosis in about two minutes. Then follow the four weeks above.