What research says about brain games and cognitive training

Three things that get called the same thing

A puzzle you do for fun is general mental activity. Cognitive training, in the research sense, is guided practice on set tasks aimed at a named skill such as reasoning or speed of processing. A commercial brain-training app is a product. It may borrow the look of training without matching the dose, the tasks, or the people in a trial.

Getting better at the game you practised is near transfer. Getting better at something you did not practise — another test, your job, or living independently — is far transfer. Reviews of commercial programs have not found convincing evidence that practice on one task reliably lifts unrelated abilities.

What official sources say

The World Health Organization’s 2019 guideline says cognitive training may be offered to older adults with normal cognition and to people with mild cognitive impairment to reduce the risk of cognitive decline and/or dementia. The recommendation is conditional. The WHO rated the evidence very low to low. “May be offered” is not “has been shown to prevent dementia,” and it does not name any app.

The U.S. National Institute on Aging describes the ACTIVE trial: healthy adults over 65 who trained in memory, reasoning, or speed of processing improved in the area they trained, with reasoning and speed gains still seen up to 10 years later. NIA also says it is still unknown whether those long-term benefits help people live independently longer. On commercial computer games, NIA’s position is direct: there is not enough evidence that they have the same effect as the training used in trials.

A separate 2016 review in Psychological Science in the Public Interest reached a similar limit for brain-training programs sold to the public. A 2010 Nature paper that put a commercial program to the test found improvement on the trained tasks and little evidence of generalisation.

A newer finding, and why it is not a product claim

In February 2026, NIH described a 20-year follow-up of the ACTIVE trial. Among adults who had been 65 or older, the group that received speed-of-processing training and later booster sessions had fewer subsequent dementia diagnoses in Medicare claims. Speed training without those boosters, and the memory and reasoning training arms, did not show that association in the same analysis. Claims data are not a full clinical exam. The training was a defined protocol from a trial that began in the late 1990s, not an off-the-shelf puzzle app, and it was not Two5.

One encouraging follow-up does not license a sentence like “brain games prevent dementia.” It is a reason to keep the question open and to describe the exact training, the population, and the outcome.

What this means for Two5

Two5 is building short scored challenges in numbers, logic, and later computer science and memory. A future score will measure that round. It will not diagnose anyone, estimate dementia risk, or stand in for medical care. We have not run a trial. The waitlist is for the game, not enrollment in a study.

If you are worried about your own memory or someone else’s, talk to a qualified clinician. An article and a puzzle score are not an assessment.