When I was preparing for Canada PR, I learned quickly that a competitive CRS score is critical. Language points are a big part of that, so the English test mattered a lot.
The established options, IELTS and CELPIP, had plenty of prep resources. What I still found hard was knowing exactly how to push toward a high score. There was material everywhere, but not enough clarity on what would actually move the needle.
Then PTE Core launched in 2024. I was genuinely excited. A fully machine-scored exam felt like a better fit for how I think about practice: measurable, consistent, and less dependent on the luck of a human examiner on a given day.
The catch was the prep landscape. There were not many resources built for PTE Core. Most existing platforms were designed for PTE Academic, and they mostly shipped those Academic materials over to Core. The overlap was incomplete, and a lot of it was not really relevant for immigration candidates.
I saw that gap as an opportunity: build a platform and a community specifically for learners on the Canadian immigration path. Even better, I could put my love for languages and AI into something meaningful for people going through the same journey I was on.
That became Phrasel.
The problem I couldn't ignore
PTE Core is Pearson's English test for Canada immigration (work, study, PR). For a lot of people in Toronto, and around the world, the score isn't just a number. It's CLB levels, Express Entry points, and a timeline for a fresh start.
The existing prep market is loud on volume:
- Practice 1000 questions
- Memorize these templates
- Watch this 45-minute tip dump
Volume helps until it doesn't. Blind repetition builds confidence that collapses under exam pressure. Templates help until the task doesn't match the template. More material without a diagnosis just makes the haystack bigger. Shipping Academic content into a Core-shaped hole does the same thing: more practice, less relevance.
My thesis became simple:
Learners do not need more practice material. They need clarity on what to improve next.
That line still sits on the About page, because it's the product.
What Phrasel actually is
Phrasel is focused on PTE Core (not Academic), intentionally. The Canada PR path has different stakes, different score goals, and different workflows than university admissions.
What I ship today:
- Practice across Speaking, Writing, Reading, Listening: exam-style tasks with guides and examples
- AI scoring for speaking & writing: pronunciation, fluency, content, grammar, vocabulary; word-level feedback where it helps
- Full-length timed mock tests, with a report that shows overall + skill scores, strengths, and focus areas
- Canada PR tooling: CLB conversion, CRS language points, test comparisons, so practice maps to immigration goals
- A free tier: practice + one scored mock, because people should feel the loop before paying
Live product: phrasel.com · App: app.phrasel.com
Product decisions that mattered more than features
1. Clarity over volume
I refused to win by having the biggest question bank.
The unit of value isn't another unanswered Read Aloud. It's leaving a session knowing: oral fluency is the bottleneck; here's what to do next.
Mocks are diagnostics first, score vanity second. If a mock doesn't tell you where to spend the next week, it's entertainment.
2. Build for the immigration journey, not a generic English exam
Canada PR learners think in CLB, not abstract "overall 79." So I put CLB conversion and Express Entry language points next to practice.
That also shaped content: I write deeply about PTE Core tasks, booking, fees, score charts, and Canada PR pathways.
3. Real ability over tricks
Templates are aids. They are not a substitute for skill.
I say this out loud because the market sells shortcuts. Overnight CLB jumps without building real English don't hold up on test day, and they don't hold up in a new country either. I want scores that grow with ability, not despite it.
4. Clear AI scoring
AI scoring for speaking/writing is the hard part, and the part people care about most.
I use speech and language models to score against PTE Core style criteria, with component-level feedback. I'm clear with users: my scores are training signals I refine continuously, not a guarantee of Pearson's official mark.
That honesty matters. Overclaiming "exact Pearson scores" might convert better in ads. It would also burn trust the first time reality disagrees.
A quick peek under the hood
For the builders reading this:
| Layer | Stack |
|---|---|
| Marketing site | Next.js (static export), MDX blog, Tailwind |
| App | React + Vite |
| API | Flask, SQLAlchemy, Docker |
| Scoring | AI Services for Speech Recognition |
| Payments | Stripe |
The interesting work isn't the framework choice. It's making speaking audio, scoring pipelines, mock timing, and feedback UX feel like one coherent study session instead of five disconnected tools.
What I believe (and what I don't)
I believe
- Real language ability and test scores grow together
- Clear feedback beats blind repetition
- You should always know which skill is holding your score back
- Preparation should feel like progress, not guesswork
I don't believe
- You can jump a CLB level overnight without building real skills
- Confidence grows from volume, not from feedback
- More practice alone is enough to raise your score
- Memorizing templates is enough to handle every PTE Core task
If you're preparing for PTE Core
Try a free scored mock and see whether the feedback names the thing you've been guessing at:
If you're building something adjacent (AI feedback for skills, immigration tooling, language learning), I'd love to hear what you're learning. Feedback (and critique) welcome in the comments.
Phrasel is educational software. Not affiliated with Pearson or IRCC. Not immigration advice.
Top comments (0)