The evidence, in full.

A 2026 review in Addiction

Twenty studies pooled, covering 1,249 participants. The 12 studies that measured craving, covering 696 participants, found it fell by a small but consistent amount: Hedges' g = 0.30, 95% CI 0.12 to 0.48, with low heterogeneity between studies (I² = 23.5%). The effect held across nicotine, alcohol, opioids, stimulants and polysubstance use.

Roxburgh, Rowland, Lubman, Bains, Pek, Tso, Manning and Arunogiri (2026). DOI 10.1111/add.70542

Limits. A g of 0.30 is a small effect. It takes the edge off. It does not remove an urge, and any product implying otherwise is overstating this paper. The pooled figure of 20 studies is often quoted for the craving result; the craving result comes from 12 of them.

The mechanism paper

Four experiments in abstaining smokers. A visual imagery task reduced cigarette craving and an equivalent auditory task did not. A visuomotor task, making shapes from modelling clay, beat a verbal task of counting backwards. The visual advantage held even when imagery had not been used to induce the craving in the first place.

May, Andrade, Panabokke and Kavanagh (2010). Behaviour Research and Therapy, 48(6), 476–485. PMID 20189549

Why this one matters. The modality controls are what establish that the effect belongs to the visuospatial channel rather than to distraction, engagement or novelty. Limits: sample sizes are not stated in the abstract and we have not obtained the full text.

Tested in daily life

Thirty-one undergraduates, split into an intervention group of 15 and a monitoring-only control of 16, prompted seven times a day for a week. The intervention group played three minutes of Tetris and then re-rated their craving. Mean reduction was 13.9 points on a 0 to 100 scale, consistent across the week, across targets including alcohol, nicotine, caffeine, food and activities.

Skorka-Brown, Andrade, Whalley and May (2015). Addictive Behaviors, 51, 165–170. DOI · PMID 26275843

Limits. The effect size is f² = 0.11, which sits below Cohen's threshold for a medium effect. Thirty-one people. Undergraduates, not people in treatment.

The one study with a behavioural outcome

Forty-eight undergraduate women, four weeks, using a hand-held device showing dynamic visual noise whenever a food craving occurred. Craving intensity fell about 23 percent, and craving-related consumption fell too. This is the only study in the set that measured what people actually did rather than what they reported feeling.

Kemps and Tiggemann (2013). Appetite, 68, 152–157. PMID 23685086

Limits. Forty-eight people, all women, all undergraduates, food cravings only.

Two more, with a weakness worth stating

A 2014 lab study tested naturally occurring rather than induced cravings, and found lower craving and less vivid craving imagery after Tetris. A clay-modelling study found craving reduced by loading the visuospatial sketchpad but not by loading the phonological loop, which is the same modality finding from a different angle.

Skorka-Brown, Andrade and May (2014). Appetite, 76, 161–165. PMID 24508486. Andrade, Pears, May and Kavanagh (2012). Appetite, 58(3), 955–963.

Limits. The 2014 control condition was waiting for a computer program designed never to load. That is a frustrating null activity, not an active control, so that experiment on its own cannot separate visuospatial interference from doing anything at all. The 2012 clay study is summarised here from a secondary source; we have not read the original. And the specific puzzle we built is untested. The evidence covers Tetris, clay, visual noise and generic visual working-memory tasks. Ours is a close inference from those, and an inference is what it is.

The trial, and its null result

One hundred and twenty-three undergraduate smokers, interested in changing but not enrolled in a cessation programme, randomised to brief instruction in the technique or to no instruction at all. Urge intensity did not differ between the groups. The instructed group smoked significantly fewer cigarettes over the following week.

We lead with this rather than bury it. The technique does not claim to make urges smaller, so a result showing unchanged intensity and changed behaviour is the mechanism working as described rather than failing.

Bowen and Marlatt (2009). Psychology of Addictive Behaviors, 23(4). DOI 10.1037/a0017127

Limits, and they are real. The comparison group received no instruction of any kind, so the trial cannot separate the technique from being handed something to do. Undergraduates wanting to cut down, not people in treatment. Cigarettes self-reported over seven days. One trial, no replication. The research group behind it is the group behind much of the interference work above, so it is not independent confirmation.

Where the technique comes from

Alan Marlatt named urge surfing in the relapse-prevention literature in 1985. The clinical training summary sets out the sequence: watch the breath without altering it, notice thoughts without feeding or fighting them, then notice the urge in one area of the body and stay curious about how it changes.

Marlatt and Gordon (1985), Relapse Prevention, Guilford. Technique as taught: Walsh (2006), Turning Point, Melbourne. PDF

Limits. We have not obtained Marlatt and Gordon's primary text, so everything describing the technique here comes from clinical secondary sources. And that source cuts against us: it states the skill is difficult to understand without practice and recommends practising on ordinary urges first. Our app delivers it to someone who has never practised, at high intensity. On the source's own account, that is the condition under which it works least well.