Did your test change in 2026? What's real, and what's folklore
Every year, prep and AI-tool sites publish "your test changed!" articles to look fresh — and most of them are wrong, or describe an old change relabeled with this year's date. We checked each test against the official source. Here's the honest split: which tests actually changed, and which "reforms" have no primary source at all.
✅ Actually changed
IELTS
The delivery changed, not the test. Paper-based IELTS is being discontinued through mid-2026 and IELTS moves computer-delivered. The sections, question types and 1–9 band scoring are unchanged — so the test isn't "harder," it's just on a screen.
TEF Canada
Yes, but in September 2025, not "2026." The listening and reading sections were simplified to 40 questions each; writing and speaking were unchanged. This one is real and official — unlike the TCF "AI reform" myth.
TOEFL iBT
Yes — TOEFL iBT was materially redesigned in early 2026 (shorter, with adaptive sections). Because the test changed, we're rebuilding our TOEFL practice to match the new format before publishing it. Note: TOEFL is not accepted for Canadian immigration.
🚫 Didn't change — the "2026 reform" is a myth
TCF Canada
No. The "2026 AI reform" — adaptive AI listening, a new writing grid, "29 questions" — isn't in the official Manuel du candidat. The sourced myth-buster. →
CELPIP
No. CELPIP-General's format, difficulty and scoring are unchanged. "Getting harder in 2026" and "AI-human hybrid scoring" trace to marketing blogs with no official source — and CELPIP's own site says writing and speaking are human-rated, not AI.
PTE Core
No. The 2025 PTE overhaul applies only to PTE Academic. Pearson explicitly states "PTE Core and PTE Home are not affected." PTE Core is unchanged — and it's the version Canada accepts for PR.
Duolingo English Test
Not in 2026. The "second camera" requirement and the subscore updates that prep sites label "2026" are actually 2024 Duolingo rollouts. The current format — adaptive, about an hour, scored 10–160 with four subscores — is unchanged.
Every verdict is checked against the test-maker's own current documentation (or IRCC) and dated. When we can't confirm a claimed change from a primary source, we say so — rather than repeat it.