How to Choose a Tutor: Stop Asking Questions. Start Verifying Claims.

1. The problem with asking the tutor

Ask "do you know the IB?" or "are you experienced?" and the answer is fairly predictable. That doesn't make the answer false — it simply means it's a claim made by the person being evaluated, not evidence.

This isn't about assuming tutors are dishonest. It's about recognising that marketing language and independently verifiable evidence are different categories of information, and only one of them should carry weight on its own.

A better process is claim, then independent evidence, then verification, then judgment — not claim, then trust.

3. A review is evidence. It is not automatically proof.

Reviews are useful, and easy to overinterpret.

Google's contribution policy is explicit that a review should reflect a genuine, unbiased experience with the business, and it prohibits several specific forms of manipulation: paid reviews, reviews posted from multiple accounts at one person's request, and — notably — businesses that discourage or prohibit negative reviews, or selectively solicit positive reviews. Google also states plainly that it uses a mix of human review and machine-learning detection to enforce this, which is itself an admission that violations happen at meaningful scale.

None of that means Google reviews can't be trusted. It means they should be read as evidence within Google's review system, not as an independently audited record of every transaction that occurred. For context, Google updated its own policies in April 2026 to prohibit review quotas and pre-written review content, and its enforcement report for 2025 cites more than 292 million policy-violating reviews blocked or removed — a sign of how actively these systems need to be moderated.

5. Superprof: what a review count doesn't show you

A review doesn't exist in a vacuum. The system that produced it is itself information, and Superprof is a useful, well-documented example — not because its rules are unusual, but because they're unusually easy to check.

Its own help centre states that any tutor without "Super Tutor" status is required to offer a free introductory session, and that both tutor and student become eligible to leave a review seven days after a lesson request is accepted — a timer tied to acceptance of the request, not to a completed, paid lesson. In practice, a free first session is enough to generate a review. That doesn't make the review fake. It does mean a review on this kind of platform tells you a student showed up once, not that they became a client.

Superprof's own published policy for students also describes a second detail worth knowing: what happens when a review is negative. Its team contacts the student, then the tutor, who receives a warning — there isn't a published equivalent process described for positive reviews. That's simply a difference worth knowing about, not evidence that the platform is doing something improper, and it's the kind of detail that isn't visible from a star rating alone.

What exactly triggers a review, and is negative feedback handled the same way as positive feedback?

7. Notice the shape of a review profile, not just the count

A review count on its own doesn't tell you much. The pattern behind it often tells you more.

The European Commission's 2021 coordinated sweep of 223 major consumer websites — shops, marketplaces, booking sites, comparison services — found that authorities could not confirm review authenticity on 144 of them, that 104 gave consumers no information about how reviews were collected, and that a comparable number gave no information about how fake reviews were prevented. That's a common pattern across the market, not an exception.

One pattern worth noticing yourself: a review profile made up entirely of five-star ratings, with no variation at all, is unusual. Genuine feedback collected from different people over months or years tends to include at least some spread — a four-star review, a mixed comment, something less than glowing. It isn't proof that anything is wrong, but it's worth a second look before treating a review count as settled evidence.

The right question is never just "does this tutor have reviews?" It's "what does the actual pattern of those reviews look like?"

9. Knowing the subject and teaching it are not the same skill

A person can be excellent at mathematics and a poor mathematics tutor. The reverse is also true.

The Education Endowment Foundation's evidence review on one-to-one tuition identifies specific implementation factors that separate effective tuition from simply reducing the student-to-teacher ratio: accurately identifying which students need support, understanding the specific learning gaps involved, and, critically, well-planned feedback. That's a far more demanding picture of tutoring than "the tutor knows the subject," and it's also a picture that's hard to fake in a single conversation, but easy to test with the right question: not "do you know IB Mathematics," but "how do you determine whether a student is struggling with the underlying maths, the use of the GDC, exam technique, or how the question is phrased?" A tutor who has actually done this work can answer immediately, with specifics. A tutor reciting a marketing line cannot.

11. Build an evidence stack

There's no single test for choosing a tutor. Build a stack instead.

The interview hasn't disappeared. It has moved to the correct place in the process — after the research, not before it. Once you know a tutor has real IB experience, the useful question stops being "do you know IB Mathematics?" and becomes "how do you tell the difference between a student who doesn't understand the maths and one who doesn't understand what the question is asking?" That's a question a marketing slogan can't answer.

Call it what it is: an evidence stack — a short, repeatable list of checks, not a single gut-feeling conversation.

13. The bigger picture

None of this is unique to tutoring. Unverifiable experience claims, inflated review counts, and content written for search engines rather than accuracy are symptoms of a wider shift: generative tools have made credible-looking content easier to produce than ever, and regulation is still catching up. In the US, the FTC's 2024 rule banning fake and incentivised reviews reflects that shift — it allows for civil penalties of up to $53,088 per violation (adjusted annually under 16 CFR 1.98), though enforcement against smaller, individual actors in a global market remains difficult in practice.

The takeaway isn't cynicism — it's a bit of methodical care. Everything in the checklist above takes under thirty minutes, and while it won't guarantee you find the right tutor, it meaningfully improves the odds of seeing past a well-built website to what's actually there.

In a market where credible-looking content is increasingly easy to produce, the signals that keep their value are the ones that are hard to fake: a dated review from a verifiable student, a multi-year record of activity, and claims that stay consistent wherever you happen to check them.