Skip to main content
research explainers

What the research actually supports about brain exercise

A careful walk through the evidence on what daily cognitive practice can and cannot be claimed to do.

Updated Reviewed by Senwitt Editorial Team

What does the research actually support about brain exercise?

The literature supports a narrow, specific claim: practising a thinking skill keeps you better at that specific skill, and there is modest transfer to closely related tasks. It does not support broad claims about preventing cognitive decline, improving general intelligence, or protecting against dementia — the FTC, the Stanford Consensus, and the Simons et al. 2016 review have been explicit about that. The cognitive-offloading literature adds a complementary, recent finding: heavy delegation to AI and search engines is associated with measurable encoding and skill-formation effects, which makes regular unaided practice newly relevant.

The brain-exercise category has a credibility problem, and it earned it. Through the 2010s, several large companies in the space made broad cognitive-improvement claims that the published evidence did not support. The result was regulatory action, a high-profile consensus statement from the scientific community, and a generation of skepticism that hangs over the category to this day.

Senwitt is built inside that context. We are deliberate about what we will and will not claim, because the gap between the marketing instinct and the evidence is real and it matters. This post walks through what the literature actually supports, where it stops, and how that maps onto Senwitt's claim line.

The narrow claim the evidence supports

Two literatures matter here.

The first is the cognitive-training and brain-training literature. The most-cited synthesis is Daniel Simons and colleagues' 2016 review in Psychological Science in the Public Interest — roughly 200 pages and the closest thing the field has to a consensus document. The careful summary of what brain-training programs can be shown to do:

  • Practising a specific cognitive task produces improvement at that specific task. This is the most reliable finding in the entire literature.
  • There is modest transfer to closely related tasks — tasks that draw on similar underlying processes.
  • There is limited and inconsistent transfer to general cognitive outcomes — broad measures of intelligence, real-world cognitive performance, prevention of age-related decline.

This is not a dismissal. Practising specific thinking skills genuinely improves those skills, and that improvement is worth something on its own. The dismissal is of the bigger claim that practising puzzles in an app translates to broad cognitive enhancement or decline prevention.

The second is the cognitive-offloading and skill-disuse literature — newer, sparser, and more directly relevant to the AI moment. Risko and Gilbert's 2016 Trends in Cognitive Sciences review, the 2011 Sparrow et al. Google-effect paper in Science, and the more recent work on AI-mediated thinking (including the MIT Media Lab cognitive-debt 2025 preprint on arXiv, with the Stanković 2026 critique also in preprint) point toward a complementary finding. When cognitive work is delegated to external systems — search engines, GPS, AI assistants — the encoding pattern shifts and the underlying skill gets fewer reps. The 2026 Anthropic study on coding skill formation found a measurable difference in skill mastery between AI-assisted and unaided conditions.

Put the two literatures together, and the narrow defensible claim is roughly this: regular practice of a thinking skill maintains that skill better than non-practice; in an environment where AI and other tools are doing more of the everyday cognitive work, the daily existence of unaided practice is more relevant than it used to be.

What the evidence does not support

The same literatures are clear about what they do not establish.

They do not support broad cognitive-decline-prevention claims. This is the claim line that the 2014 Stanford Consensus Statement and the 2016 FTC Lumosity settlement were specifically about. The evidence did not support "our app prevents age-related cognitive decline," and the regulatory action was a direct consequence.

They do not support IQ-improvement claims. The intelligence literature is large, careful, and not friendly to the claim that any app or short daily practice can move general intelligence in a meaningful way.

They do not support clinical claims. No app should claim to prevent or treat dementia, Alzheimer's disease, or any other clinical condition unless it has been through the relevant regulatory process. None of the products in the brain-exercise category have. Senwitt has not, and we do not make those claims.

They do not support "makes you smarter" or "boosts your brain" framings. Both phrases are loose enough to imply far more than the evidence supports, and both are now closely associated with the marketing patterns that drew regulatory action. We avoid them in our copy.

What Senwitt actually claims

Senwitt's position is narrow on purpose. We claim:

  • A daily, mixed Set of short, deliberate, unaided cognitive practice across writing, math, code, memory, reading, and reasoning gives those specific skills regular reps.
  • In an environment where AI tools are doing more of the everyday cognitive work, the existence of unaided practice somewhere in the day is increasingly relevant — and that's what we're built for.
  • We track a self-reported Sharpness signal so that users have a private read on their own pattern, not so that we can claim population-level cognitive outcomes.

We do not claim our product prevents cognitive decline. We do not claim it improves general intelligence. We do not claim it protects against dementia or any clinical outcome. We do not make brain-improvement promises beyond the deliberate-practice base rate that the literature supports.

The reason for the discipline is straightforward. The category has a history of overreach. The cognitive-offloading argument we do make is strong enough to stand on its own. Adding the broader claims would undermine the narrower, defensible ones — and would put us on the wrong side of the same evidence line the Stanford Consensus and the FTC drew in the 2010s.

Why the AI angle matters newly

The Senwitt position is not a brain-training position re-packaged. The argument is structurally different.

The brain-training claim was: "our specific puzzles improve your general cognitive performance." That claim did not survive the evidence.

The cognitive-offloading claim is different. It is: "when more of your everyday thinking is mediated by external systems, the underlying skills receive fewer reps, and the evidence for skill-specific atrophy is real." That claim is in the literature, and it has been growing as AI use has spread. The right response to it is not a transfer-of-training story; it is a deliberate-practice story applied to thinking skills the same way it has long been applied to musical, athletic, and craft skills.

In other words: brain training was selling cognitive enhancement. Senwitt is selling deliberate practice for skills that the AI environment is putting in less daily use. The framing is narrower, the evidence is more direct, and the claim line is honest.

Why the language matters

The phrasing we use across the site reflects the claim line above. We say "keep your thinking skills sharp," not "make you smarter." We say "deliberate practice," not "brain training." We say "daily reps," not "cognitive enhancement." When we cite preprints, we flag them as preprints. When we summarise a study, we describe what it actually measured rather than the headline version of what the press release implied.

This is partly editorial discipline and partly a structural feature of the position. The deliberate-practice case is strong enough on its own evidence. The cognitive-offloading case is real and growing. The combination is enough to build a serious product around without overreaching. Overreaching would put us in the company of the 2010s brain-training industry — and that is not a place we want our credibility to live.

The research/scope-of-evidence page is the formal statement of all of this. It is one of the most important pages on the site, and it is the one we update most carefully when new research lands.


From Senwitt · advertisement

The text above is editorial. What follows is a promotional message from Senwitt, the maker of this site. Senwitt is a brain-exercise app and is not a medical product. Read the full disclaimer in the footer.

Sources

  1. 1.A Consensus on the Brain Training Industry from the Scientific Community Stanford Center on Longevity, 2014.
  2. 2.Do 'Brain-Training' Programs Work? Psychological Science in the Public Interest 17(3):103–186 (DOI 10.1177/1529100616661983), 2016.
  3. 3.Lumosity to Pay $2 Million to Settle FTC Deceptive Advertising Charges for Its Brain Training Program Federal Trade Commission, 2016.
  4. 4.Your Brain on ChatGPT: Accumulation of Cognitive Debt (preprint) arXiv, 2025.
  5. 5.Comment on: Your Brain on ChatGPT arXiv (Stanković et al.), 2026.
  6. 6.Cognitive Offloading Trends in Cognitive Sciences 20(9):676–688 (DOI 10.1016/j.tics.2016.07.002), 2016.
  7. 7.Google Effects on Memory: Cognitive Consequences of Having Information at Our Fingertips Science (Sparrow et al.), 2011.
Get the app

Take this argument with you. Daily practice in the app.

Download on the App StoreGet it on Google Play

Free download. Super Senwitt available in-app.

We use cookies to make the site work, measure aggregate usage, and (if you opt in) attribute organic app installs. You can accept all, reject all, or customize.

See our cookie policy and privacy policy.