Your recording is transcribed and evaluated against the same four criteria a real examiner uses: grammatical accuracy (verb endings, agreement, tense use), vocabulary range (precision, idiomatic expressions), pronunciation and fluency (nasal vowels, uvular R, register, pause patterns), and task completion (did you do what the prompt asked). You get a numerical score per criterion plus an NCLC band estimate within seconds of finishing the recording.

TEF has 2 tasks totalling 15 minutes — a 5-minute information-seeking role-play and a 10-minute persuade-the-examiner task. TCF has 3 shorter tasks: a 2-3 minute self-introduction, a 5-6 minute role-play, and a 5-6 minute argumentation task. Both grade on the same criteria; TCF's shorter discrete tasks suit candidates who freeze on long single tasks, TEF's fewer-but-longer tasks suit candidates who prefer one focused effort.

CLB 7 speakers attempt complex structures (subjunctive, conditional, concession) and use varied connectors (cependant, par conséquent, étant donné que) even with small errors. CLB 5 speakers stay safe with simple sentences and avoid risk. The biggest single jump from CLB 5 to CLB 7 is task completion on the persuasion tasks: persuading (not just describing) is what unlocks the higher band.

Three drills move the needle fastest for English-first speakers. (1) The uvular R: try gargling water then transition into the R sound. Practice in isolation before pairing with vowels. (2) Nasal vowels: drill the minimal pairs bain/banc/bon, vin/vent/vont — examiners hear shortened or missed nasals immediately. (3) Liaison: words ending in silent consonants (les, des, mes, ses) link to the next word's vowel — "les amis" sounds like "lez-ami". Missing liaison flags non-native speech.

Use the 2-minute prep time to outline a 3-act structure on scratch paper: problem (1-2 minutes), evidence (5-6 minutes — at least two distinct points), conclusion (1-2 minutes). The outline alone is often the difference between CLB 6 and CLB 7. Avoid trying to memorise sentences in prep; just commit to the structure and let the actual French come out unrehearsed.

Yes — AI speaking practice grades your recording against the same criteria a real examiner uses. The advantage over self-recording is that AI catches things you cannot hear yourself (pronunciation errors, repeated grammar mistakes, register slips). The advantage over human tutoring is volume: 5–10 short sessions per day adds up faster than weekly tutor lessons. Most candidates who hit CLB 7 use AI practice for daily reps and book a tutor only for the parts AI cannot coach well (accent, exam-day nerves).