Every App Asks You to Set a Goal. The Evidence Is Thin.
Setting a goal is the first thing almost every fitness product asks you to do. It feels like the responsible move, the thing serious people do. A large synthesis published in August 2026 suggests we know remarkably little about whether it changes how much anybody moves.
What 36 trials could not establish
A systematic review and meta-analysis pooled 36 randomised trials covering 7,275 participants, testing behavioural interventions that include goal setting against matched comparators.
On pain, the pooled effect was SMD −0.32 (95% CI −0.58 to −0.06), rated very low certainty by the authors. On disability, SMD −0.52 (95% CI −0.90 to −0.15), rated low certainty.
On physical activity, the number is the interesting one: SMD 0.88, 95% CI −0.32 to 2.07.
The point estimate is large. The interval spans zero and is roughly two and a half standard deviations wide. That is not a weak result so much as an uninformative one. The data are compatible with goal setting slightly reducing activity and with it transforming activity, and cannot distinguish between them. The authors’ own conclusion is that effects on physical activity are unclear (The Clinical Journal of Pain, PMID 42584484, August 2026).
Two things belong up front. The population is adults with chronic musculoskeletal pain, not general adults. And the comparison behind those numbers was built to isolate goal setting, against matched programmes with the component stripped out. What the abstract never says is how many of the 36 trials achieved that match, or how close it was, and the usual-care and wait-list comparisons cannot isolate it at all.
The same gap keeps appearing
That last problem is the recurring one in this field, and it is worth seeing the pattern.
A review of 87 eHealth trials pooled the 53 that could be combined and found a small effect on physical activity, SMD 0.22 (95% CI 0.13 to 0.31). The authors tell readers to treat it with caution: the funnel plot was badly asymmetric, Egger’s test came in at p < 0.001, heterogeneity was 85%, and among the low risk-of-bias trials the effect dropped to 0.13. Then they coded the interventions by technique. Feedback, goal setting, social support, self-monitoring, prompts. Every one of those techniques was positive in the trials that used it, goal setting included, at SMD 0.21 (95% CI 0.09 to 0.32). What the review could not show is that any of them added anything: the trials without goal setting came in at 0.26, marginally higher, and every between-group comparison was non-significant, p = 0.05 to 0.78 (Digital Health, PMID 42534183, July 2026). The participants were older adults, mean age 66.4, with only four of the 87 trials recruiting for frailty and none for sarcopenia, and the technique comparisons were made across trials rather than head to head, so read it as no technique being credited with the effect rather than no technique working.
A microrandomised trial randomised 70,552 send decisions and found no message characteristic predicted whether people moved afterwards, which the reminders piece covers.
A scoping review of just-in-time nudges, the research on catching people the moment they could act, lands in the same place and calls its own evidence preliminary.
Three syntheses and one large trial, and the shape repeats: the review that tried to pit matched programmes with goal setting against the same programmes without it came back too wide to read in either direction, the review that sorted trials by technique could not show that goal setting added anything, the nudge trial varied the messages and no variation predicted whether people moved, and the scoping review calls its own evidence preliminary.
What to do about your own goals
The read is not “stop setting goals”. It is narrower and more useful than that.
Treat a target as a planning tool until something shows it is a mechanism. The trials cannot yet say the number itself produces the behaviour, and that is not the same as showing it does not. If writing a target down helps you plan the week, keep doing it. Just hold it as a plan, not as the thing that makes you move.
When a week goes badly, shrink the session rather than respecifying the goal. The instinct is to rewrite the target more carefully. The evidence gives no reason to think a better-specified goal does more work than a worse one.
Watch which goal you pick. In an analysis of 2,771 real users, the goal category with the worst retention was weight loss. Muscle gain and general ageing outcomes did better (JMIR mHealth and uHealth 2026;14:e72201, PMID 41505740). One app, one country, needs replicating. But it points the same way as a separate study of a community exercise programme, in which improved mental health came through as a more prominent reinforcer than weight loss (Journal of Primary Health Care, PMID 42556994, August 2026).
The identity alternative, and its limits
There is a popular alternative to outcome goals: describe who you are rather than what you want. “Someone who moves twice a week” instead of “five kilos down”.
The evidence for it is suggestive and weaker than its popularity implies. A cross-sectional survey of 477 people living with and beyond cancer, mean age 48.5, most of them women and on average five years past diagnosis, used structural equation modelling and found a stronger exercise identity reported alongside more moderate-to-vigorous activity, stronger habit, more regulatory behaviour, higher perceived capability and a more positive attitude to exercise (Journal of Behavioral Medicine, PMID 42538475, July 2026). Identity was the outcome in that model rather than the predictor.
The limitation is fundamental. This is cross-sectional, so identity and behaviour cannot be causally ordered. People who train a lot naturally describe themselves as exercisers. All measures were self-reported and the sample was self-selected. Three of the reported coefficients fall outside the minus 1 to 1 range, which usually points to a scaling quirk rather than a huge effect, and the full paper that would confirm how the variables were scaled is paywalled, so none of these numbers should be quoted as an effect size or ranked against the others.
Nothing here shows that adopting an identity produces the behaviour. The honest framing is that identity and habit travel together, and which one arrives first is unestablished.
Which lands somewhere reasonable: an identity framing is probably a better description than a weight target, it costs nothing to prefer, and it should not be sold as a technique.
The part that does have evidence
While the motivational mechanics stay unresolved, the structural ones hold up.
A two-country study of 4,114 people separated two things usually bundled together: perceived capacity (“I could do this”) and perceived controllability (“my circumstances permit this”). Intentions predicted later increases in capacity. They did not uniquely predict change in controllability, inside a model already adjusting for prior behaviour and prior controllability, which is a failure to detect rather than a demonstrated absence. The pattern replicated across both national samples (Journal of Health Psychology, PMID 42576379, August 2026). Self-report throughout, lagged rather than causal, and a theory-refinement paper rather than an intervention.
The implication is sharp. Wanting to train seems to feed what you believe you are capable of, and not whether your Tuesday has room in it. Nobody in that study was messaged or coached, so this is about where motivation tracks, not about what an app can shift. If your blocker is “I have twelve minutes and no equipment”, nothing here shows a goal, badge or pep talk shifts that. A session that fits the twelve minutes removes the constraint rather than arguing with it, which is a design choice these studies did not test.
Do fitness apps work puts goal setting alongside the other mechanics the category leans on, and building a fitness habit that survives real life takes the practical route instead. What keeps people using a fitness app covers what held people in the one trial that managed it.
ZeroFit doesn’t ask what you want to weigh. It asks how you’re doing and how long you’ve got. Get beta access.
ZeroFit content is educational and not medical advice. If you have an injury, a health condition, or you’re pregnant, check with a qualified professional before starting any new movement routine.