Over the past decade, several education technology companies built around the "adaptive learning" label were acquired, folded into larger companies, or shut down โ€” a consolidation industry commentators have called an adaptive learning market shakeout. The pattern behind it is instructive for any parent evaluating today's crop of AI tutoring products: the word "adaptive" turned out to be much easier to put on a landing page than to actually build.

Why the label got attached to so much that wasn't

Genuine adaptivity โ€” tracking an individual student's mastery skill by skill and recalculating, in real time, the single best next question โ€” is difficult, expensive engineering. It requires either a real statistical model (Item Response Theory, Bayesian Knowledge Tracing, or similar โ€” see how adaptive learning algorithms actually work) built and validated on real student data, or a genuinely well-designed rules engine doing similar work.

A fixed content sequence โ€” the same order of topics and questions for every student, regardless of performance โ€” is dramatically cheaper and faster to build. Wrap it in a friendly interface, add the word "adaptive" or "personalised" to the marketing copy, and it's very difficult for a parent evaluating it in a ten-minute demo to tell the difference from the outside.

What the industry commentary actually documented

Analysts and practitioners tracking the sector have pointed to a specific, recognisable pattern: products marketed heavily around "adaptivity" that, on closer inspection, forced users down a largely pre-determined path rather than genuinely responding to performance. Industry critiques from within the edtech space itself (not just outside sceptics) have called this out directly โ€” the phrase "you're not adaptive learning ifโ€ฆ" became something of a genre in edtech commentary, cataloguing the tell-tale signs of the gap between the label and the reality.

The market consequence was real: a wave of companies whose core identity was built around the "adaptive" brand were acquired, merged, or wound down over the following years โ€” not proof that adaptive learning as a concept doesn't work, but reasonably strong evidence that a lot of what was marketed under that name didn't survive contact with actually delivering it.

A useful, sobering statistic on UK edtech evidence

Independent reviews of the UK education technology sector have found the evidence base behind many products' claims is thinner than marketing suggests. Surveys of UK edtech companies have found only a small minority โ€” in the single-digit percentages in some reviews โ€” have run randomised controlled trials on their own products, and a similarly small minority have sought independent third-party certification of effectiveness claims. This doesn't mean the majority of products don't work; it means most claims about whether they work rest on internal, unaudited data rather than independent verification โ€” worth knowing before treating any efficacy claim, including ones from aitutors.me, as settled fact rather than a claim to interrogate.

Why Carnegie Learning is a useful, honest case study rather than a cautionary tale to avoid

One of the most established, genuinely research-grounded adaptive systems โ€” Carnegie Learning's Cognitive Tutor, built on real Bayesian Knowledge Tracing โ€” has a track record that's mixed, not uniformly positive. Some independent comparative-effectiveness studies found meaningfully positive results; others, in different districts, found flat or even negative results on specific sub-topics. This is worth sitting with rather than dismissing: even a system built on genuinely sound underlying technology doesn't guarantee good outcomes everywhere, because implementation, classroom integration, and how students actually use it matter as much as the algorithm itself. "Real adaptivity" is necessary for the effect to be possible โ€” it isn't sufficient to guarantee it.

What this means practically

The lesson isn't cynicism about every "adaptive" claim โ€” it's a habit of checking rather than assuming. Ask what specific mechanism a product is using (a fixed IRT ability score? mastery-gated BKT? something else, or nothing?). Ask what independent evidence exists beyond the company's own dashboard. And treat "adaptive" on its own, without either of those answers, as marketing copy rather than a fact about the product.

FAQ

What was the 'adaptive learning market shakeout'?

A period in which several edtech companies branded around "adaptive learning" were acquired, consolidated, or shut down, partly because delivered products didn't match personalisation claims made in marketing.

Why do so many products marketed as adaptive actually run a fixed path?

Genuine item-by-item adaptivity is difficult and expensive to build. A fixed curriculum sequence with a friendly interface and the word "adaptive" attached is far cheaper to ship.

Is there independent evidence on whether UK edtech products actually work?

Evidence is thinner than most parents assume โ€” independent reviews have found only a small minority of UK edtech companies have run randomised controlled trials or sought third-party certification of their claims.

Duke Harewood ยท founder, aitutors.me ยท Updated 11 Jul 2026.