Blog
The TOEFL Prep Journal
Speaking & writing strategy, real-exam breakdowns, and prep notes — data-driven guides from the team building Toeflair.
Why a 106-scorer is rebuilding TOEFL prep with AI
Apr 26, 2026Latest essays
What Is SpeechRater? TOEFL's AI Speaking Grader
SpeechRater is ETS's own AI engine for TOEFL Speaking, built in 2006 — not borrowed from the GRE. What it measures, and how one real answer scores.
Harold Guo
MySpeakingScore vs TOEFLAIR: Speaking Only, or Speaking and Writing?
Both score TOEFL Speaking on the same 1–6 band ETS uses today. What splits them is how much of the exam they cover, and how you pay to practice it.
Harold Guo
FluentPrep vs TOEFLAIR: Same Problem, Two Founders' Different Bets
Two solo founders hit the same TOEFL feedback problem and bet on different fixes. We tested both against real submissions to see how those bets play out.
Harold Guo
Magoosh TOEFL vs TOEFLAIR: What Each Report Tells You
Magoosh names the words you missed. TOEFLAIR names which ability failed: hearing it, holding it, or saying it. We ran both — here is what each report returns.
Harold Guo
How to Practice TOEFL Writing So Your Score Moves
Dozens of practice emails and discussion posts, and your band still hasn't moved. You need to know which of the four scored dimensions is holding it back.
Harold Guo
PrepEx AI vs TOEFLAIR: Which Gives Better Speaking Feedback?
We logged into PrepEx AI and ran real Listen and Repeat and Interview exercises, then compared its reports against what TOEFLAIR's feedback actually ships.
Harold Guo
TOEFL Scratch Paper Rules: What You Can Write Down and When
Scratch paper is allowed on the 2026 TOEFL, but the rules on materials, timing, and what to write differ by section — and by test center vs. Home Edition.
Harold Guo
The Best TOEFL AI Feedback Tools, Ranked by Depth
Several TOEFL AI tools now mark the exact sentence you got wrong. The split that's left is whether the redo you do about it gets saved and scored.
Harold Guo
How to use Listen & Repeat 2.0: from a score to a fix list
A walkthrough of the rebuilt Listen & Repeat report: read each sentence by cause — listening, pronunciation, delivery — then fix it until STILL turns to FIXED.
Harold Guo
TOEFL email writing: grammar isn't what raises your band
You can write a TOEFL email with zero grammar mistakes and still score a 3.5. Clean grammar is the entry ticket — three other things decide your band.
Harold Guo
TOEFL Academic Discussion: beyond the template trap
The TOEFL Academic Discussion task isn't scored on your template — it's scored on whether you add what the other two students didn't. Here's the move.
Harold Guo
TOEFL Listen & Repeat: one score, three problems
A real Listen & Repeat attempt scored 4.0 — and hid 21 problems across three different kinds of failure. Why a score can't tell you what to fix, and what can.
Harold Guo
How accurate is an AI TOEFL Speaking grader? 674 attempts
ETS scores Speaking with SpeechRater first: 100+ features across Delivery, Language Use, Topic Development — what it measures, where it falls short.
Harold Guo
Why memorizing fails the TOEFL interview
You can't memorize your way through the TOEFL interview — the prompts shift every time. What transfers is structure: an answer shape you drill into reflex.
Harold Guo
What's a good TOEFL Speaking score out of 6? 600+ analyzed
On the 2026 1–6 band, 4.0 is the floor most universities accept and 5.0 is competitive — scored from 600+ real attempts, plus the legacy 0–30 map.
Harold Guo
What Changed in TOEFL 2026? Shorter Test, New 1–6 Band Score
January 21, 2026: the TOEFL runs ~1.5 hours, scores on a 1–6 band, drops the Integrated essay, gives Speaking no prep time. Every change, section by section.
Harold Guo
Why 10 reps of Listen & Repeat fix nothing
A method for Listen and Repeat that changes your next attempt: hear the sentence, diagnose the smallest fix, repractice only that, then check it held.
Harold Guo