In short: Two solo founders hit the same gap and split on how to close it. FluentPrep AI (built by Noah, an AI scientist in the Netherlands) bet on breadth — all four sections, 12 item types, 1300+ practice items, and a cheaper three-month plan. TOEFLAIR (built by Harold after six TOEFL attempts and a 106) bet on depth — four Speaking and Writing task types on the official ETS 2026 1–6 band, with every Listen and Repeat miss classified by cause and repracticed at the phrase level. Choose FluentPrep for coverage; choose TOEFLAIR once you know which section is holding your score down.
TOEFL prep has an asymmetry built into it. Reading and Listening are self-gradable — get the question right or don’t. Speaking and Writing aren’t: a self-study taker finishes a response with no reliable way to know what just cost them points.
Noah and Harold each ran into a version of that gap, from opposite sides. Both are solo founders who’d taken the test themselves, and both eventually built a product to fix it — FluentPrep AI and TOEFLAIR.
The Same Problem, Read Two Ways
Noah hit the cost side of it in the Netherlands. Quality TOEFL feedback was, in his own words, “extremely difficult to get… without spending hundreds of dollars” — official tests run $30–50 with no feedback, platforms charge $100–300 a month, tutors run $50–100 an hour. FluentPrep AI started as a single free speaking tool and grew from there.
Harold hit the precision side of it across six TOEFL attempts, from 98 to 106. The gap wasn’t a shortage of practice — it was that he couldn’t see what was wrong with his own writing until he paid a private tutor ¥10,000 (about $1,400) for notes that actually named the problem. That loop is what TOEFLAIR tries to automate.
Same starting gap, read from two different angles — and two different bets on how to close it, unpacked below.
Noah’s FluentPrep: Betting on Breadth
Noah is an AI Scientist in the Netherlands with three years of research experience. He built FluentPrep AI alone, and the project reads like one: a single free speaking-practice tool that grew into a full exam platform over time.
Donations and ad revenue couldn’t cover FluentPrep’s own hosting and AI costs, which is why it eventually moved to a freemium model. It now serves “hundreds of weekly users.”
Given that starting point, his bet was breadth: cover the whole exam, keep the price down, ship real feedback everywhere instead of going deep on one section. FluentPrep AI covers 12 item types across all four TOEFL sections, 1300+ practice items, and a full mock test.
Harold’s TOEFLAIR: Betting on Depth
Harold is a full-stack engineer with five years of experience, studying for the TOEFL after work. He built TOEFLAIR alone after his own six-attempt climb from 98 to 106 — the full story is in why a 106-scorer rebuilt TOEFL prep with AI.
For years he could tell he’d lost points without ever being able to tell why. The break came in Writing, once he paid that private tutor for a real feedback loop: submit an essay, get notes on exactly what was wrong, then revise and drill the same prompt type again. His score jumped from 22 to 27 once that loop existed.
His bet was depth: cover less of the exam, but name the exact cause of every miss on the sections where self-assessment is hardest. TOEFLAIR covers four task types in the Speaking and Writing sections — Listen and Repeat, Take an Interview, Write an Email, Academic Discussion — scored under real exam conditions. Every response is graded against the ETS rubric on the 1–6 band, in 0.5-point steps.
Head-to-Head: Key Differences
| Feature | FluentPrep | TOEFLAIR |
|---|---|---|
| Coverage | Focus on breadth — all sections and task types are covered | Focus on Speaking and Writing — especially Listen and Repeat, Interview, Email Writing, and Academic Discussion |
| Pricing | • $19.99 / 1 month • $41.98 / 90 days | • $20 / 1 month • $51 / 3 months • $90 / 6 months |
| Scored rubric dimensions | • Writing — numeric per dimension • Interview — one overall score, no per-dimension numbers | • Writing — numeric per dimension • Interview — numeric per dimension |
| Reported scale | 0–5 per item | Official ETS 2026 1–6 band, 0.5-point steps |
| Targeted repractice | Sentence- or question-level restart, on screen only, not saved | Word/phrase-level on Listen and Repeat, saved and graded — not yet on Take an Interview |
FluentPrep covers every skill. TOEFLAIR covers Speaking and Writing on purpose, for now, and goes deeper inside them.
Speaking: One Score, or One Cause
Say your Listen and Repeat attempt comes back at 3.5. Noah’s AI shows you which words were weak and hands the sentence back so you can run it again. That costs almost nothing to deliver, it works identically across every item type he ships, and for a lot of people it is enough to start moving.
Harold’s AI starts one question later. A 3.5 can mean three unrelated things: you never caught the word in the audio, you caught it but could not produce it, or you rebuilt the sentence in your own words. Same number, three different drills, and picking the wrong one is how a week disappears. The full argument is in one score, three problems.
So TOEFLAIR spends its model budget on classification rather than coverage. Every miss is assigned a cause before anything is suggested, and while the repractice still has you say the whole sentence, it is graded against that one flagged issue instead of re-scoring the answer as a whole. Take an Interview runs the same way one level up, diagnosing a single move inside an answer rather than the answer as a whole (why memorizing fails the TOEFL interview). One thing FluentPrep still does better: its per-question “Re-answer” control has no TOEFLAIR equivalent.
Writing: Fix the Sentences, or Answer the Prompt
We submitted a 112-word email with four planted grammar errors and one requirement deliberately skipped. FluentPrep caught all of it, with rule-level explanations rather than generic advice. Anyone still claiming these tools only give vague feedback has not used this one.
That is Noah’s bet working as designed, and it also marks where the bet stops. Say your next email comes back clean at a 4. Error-level feedback has nothing left to tell you, because errors were never the problem: you answered two of the three things the prompt asked and wrote around the third.
Harold’s report is built for that reader. It opens on requirement coverage instead of sentence errors, marks what worked alongside what didn’t, and ends on a model email with the scoring standard attached to each move. A high-scoring answer is only worth something if you can see why it scores, which is the loop how to practice TOEFL writing so your score moves walks through.
Pricing
FluentPrep is the cheaper option at every published term.
| FluentPrep Pro | TOEFLAIR Pro | |
|---|---|---|
| 1 month | $19.99, auto-renewing | $20, one-time |
| 3 months | $41.98 total ($13.99/mo), one-time | $51 total ($17/mo), one-time |
| 6 months | Not offered | $90 total ($15/mo), one-time |
| Auto-renews | Monthly plan only | No |
| Refund | 24 hours, no questions | — |
At the monthly term the two are a penny apart, and FluentPrep’s renews while TOEFLAIR’s does not. Over three months FluentPrep is genuinely cheaper. If a low total price across every 2026 task type is the deciding factor, FluentPrep wins it.
The free tiers differ in shape rather than size. FluentPrep gives a daily per-item-type allowance and needs no account at all. On 2026-08-04 it ran from 1 free Interview up to 5 free Build a Sentence items, with the speaking and writing tasks at 1 or 2 each. TOEFLAIR gives unlimited practice on a hand-picked set of free questions, with full AI feedback and 2 Listen and Repeat repractices per session, behind a free signup.
Which Is Best for You?
Go with FluentPrep if breadth is the constraint. You want every section and every task type in one place, a full mock test, and the option to practice today without creating an account. It is also the cheaper of the two over three months, and its ETS disclaimer is the least hedged of the tools audited on this blog.
Go with TOEFLAIR if precision is the constraint. You have taken the test, you know Speaking or Writing is what is holding the score down, and you can see the number without seeing the fix. That is the reader Harold built for, and the one a breadth-first product is not trying to serve.
The Bottom Line
Two solo founders, the same problem, two defensible answers. Noah bet that TOEFL feedback should be cheap and everywhere; Harold bet that it should be specific enough to act on. Neither bet is wrong, and they mostly serve different weeks of the same prep.
A wider survey of the best TOEFL AI feedback tools rates five platforms on feedback depth, and how AI actually grades TOEFL speaking explains what these scoring engines measure before committing to any of them.
Start free at toeflair.com — unlimited practice on the free question set, full AI feedback, no payment required.
Further reading
- FluentPrep AI — homepage, pricing, and practice hub, checked 2026-08-04
- About FluentPrep AI — Noah’s own account of why and how he built it
- How TOEFL Listen and Repeat Is Scored (and How to Train) — FluentPrep’s own description of its Listen and Repeat feedback
- TOEFL Speaking Interview: Format, Scoring, and How to Train — FluentPrep’s own description of its Interview report
- ETS — About the TOEFL iBT test