aceModerate evidence
More
Chatbots nudged people to walk and sleep more, review of 19 trials finds
Study title: Systematic review and meta-analysis of the effectiveness of chatbots on lifestyle behaviours
- Evidence type
- Systematic review or meta-analysis
- Sample
- n = 3,567
- Sources
- 1 sourcePeer-reviewed
- Published
- 2 min read
- Truvace record
- TRV-2026-2756cite or check it
Study title: Systematic review and meta-analysis of the effectiveness of chatbots on lifestyle behaviours
A systematic review and meta-analysis from the University of South Australia pooled 19 trials with 3,567 participants and found that chatbot interventions improved physical activity, diet and sleep. The pooled effects were equal to about 735 more steps a day, roughly one extra serving of fruit and vegetables a day and 45 more minutes of sleep a night.
Read the study abstract
Chatbots (also known as conversational agents and virtual assistants) offer the potential to deliver healthcare in an efficient, appealing and personalised manner. The purpose of this systematic review and meta-analysis was to evaluate the efficacy of chatbot interventions designed to improve physical activity, diet and sleep. Electronic databases were searched for randomised and non-randomised controlled trials, and pre-post trials that evaluated chatbot interventions targeting physical activity, diet and/or sl…
The gain
A systematic review and meta-analysis of 19 trials with 3,567 participants found chatbot interventions raised daily steps by 735, fruit and vegetable intake by about 1 serving a day and sleep by 45 minutes a night.
In brief
Standardised effects were small to moderate, ranging from 0.28 for total physical activity and daily steps to 0.59 for fruit and vegetable consumption. Moderate-to-vigorous activity rose by about 103 minutes a week, based on two studies with 184 participants. Participants ranged from age 9 to 71, and effects held across short and longer interventions.
Most trials were small and weak in quality, and only seven of the chatbots used AI or natural language processing. Nearly three-fifths of studies lasted six weeks or less, so durability remains unknown. Reviewers called for larger, longer trials with objective measures to confirm the findings.
Main points
- The review searched 14 databases for studies published before 1 September 2022 and included 19 trials: 11 randomised trials, 2 non-randomised trials, 5 single-group pre-post studies and 1 panel design trial.
- Of the 19 studies, 14 (74%) were rated weak quality, 4 moderate and 1 strong; 15 targeted physical activity, 7 healthy eating and 5 sleep.
- Total physical activity rose by an SMD of 0.28 across 10 studies and 1,603 participants; in randomised trials alone the effect was 0.27.
- Daily steps rose by 735 on average and moderate-to-vigorous activity by 103 minutes a week; the steps effect in randomised trials alone (SMD 0.18) just missed significance at p = 0.06.
- Fruit and vegetable intake rose by about 1 serving a day across 4 studies with 289 participants; sleep duration rose by 45 minutes a night across 3 studies with 1,184 participants.
- Text-based chatbots and AI or NLP chatbots outperformed voice chatbots and non-AI chatbots for fruit and vegetable intake, while multicomponent programmes beat chatbot-only ones for sleep.
What the review found when it combined the studies
These are the review’s own pooled results, copied from its abstract. Truvace did not recalculate them.
Moderate-to-vigorous physical activity increase in minutes per week with chatbot interventions
Mean difference 103 (95% CI 48.20 to 159.24)
2 studies · 184 people · search to 1 September 2022
The review’s own words
Meta-analysis using mean differences showed the increase in MVPA was 103 (95% CI = 48.20, 159.24) minutes per week
Total physical activity with chatbot interventions (randomised and pre-post trials)
SMD 0.28 (95% CI 0.16 to 0.40)
10 studies · 1,603 people · search to 1 September 2022
The review’s own words
Meta-analyses of RCTs and pre-post trials showed significant effects of chatbot-based interventions for increasing total physical activity (SMD = 0.28 [95% CI = 0.16, 0.40], 10 studies, 1603 participants, I 2 = 12%, p < 0.01 [Hedge’s g])
Sleep duration with chatbot interventions
SMD 0.44 (95% CI 0.32 to 0.55)
3 studies · 1,184 people · search to 1 September 2022
The review’s own words
Significant increases were observed for sleep duration (SMD = 0.44 [95% CI = 0.32, 0.55], 3 studies, 1184 participants, I 2 = 0%, p < 0.01 [Hedge’s g])
Daily steps with chatbot interventions (randomised and pre-post trials)
SMD 0.28 (95% CI 0.17 to 0.39)
6 studies · 1,276 people · search to 1 September 2022
The review’s own words
daily steps (SMD = 0.28 [95% CI = 0.17, 0.39], 6 studies, 1276 participants, I 2 = 0%, p < 0·01 [Hedge’s g])
Fruit and vegetable consumption with chatbot interventions (randomised and pre-post trials)
SMD 0.59 (95% CI 0.25 to 0.93)
4 studies · 289 people · search to 1 September 2022
The review’s own words
There was a significant effect of chatbot interventions for increasing fruit and vegetable consumption in meta-analyses of RCTs and pre-post studies (SMD = 0.59 [95% CI = 0.25, 0.93], 4 studies, 289 participants, I 2 = 42%, p < 0.01, Fig. 4 )
Daily steps increase with chatbot interventions, mean difference
Mean difference 735 (95% CI 441 to 1029)
6 studies · 1,276 people · search to 1 September 2022
The review’s own words
the increase in daily steps was 735 (95% CI = 441, 1029, p < 0.01 [Hedge’s g]) steps per day
Total physical activity with chatbot interventions in randomised trials only
SMD 0.27 (95% CI 0.07 to 0.46)
6 studies · 641 people · search to 1 September 2022
The review’s own words
Meta-analyses of RCTs showed a significant and consistent effect in favour of chatbot interventions for total physical activity (SMD = 0.27 [95% CI = 0.07, 0.46], 6 studies, 641 participants, I 2 = 14%, p < 0.01 [Hedge’s g])
Sleep quality with chatbot interventions
SMD 0.50 (95% CI 0.09 to 0.90)
4 studies · 1,302 people · search to 1 September 2022
The review’s own words
sleep quality (SMD = 0.50 [95% CI = 0.09, 0.90], 4 studies, 1302 participants, I 2 = 80%, p = 0.02, Fig. 5 )
Moderate-to-vigorous physical activity with chatbot interventions
SMD 0.53 (95% CI 0.24 to 0.83)
2 studies · 184 people · search to 1 September 2022
The review’s own words
MVPA (SMD = 0.53 [95% CI = 0.24, 0.83], 2 studies, 184 participants, I 2 = 0%, p < 0.01 [Hedge’s g])
Sleep duration increase in minutes per night with chatbot interventions
Mean difference 45 (95% CI 33.2 to 56.9)
3 studies · 1,184 people · search to 1 September 2022
The review’s own words
Meta-analyses of sleep duration using mean differences showed the effect equated to an increase in sleep duration by 45 (95% CI = 33.2, 56.9) minutes per night.
Code checks every number above against the review’s text before it is shown. The short labels are an AI model’s reading of that text, so the review’s own sentence sits under each one. A pooled figure is only as strong as the studies behind it. How Truvace uses pooled results
The full story
The authors searched ACM Digital Library, CINAHL, Cochrane, Embase, IEEE Xplore, MEDLINE, PsycINFO, PubMed, Scopus and others for English-language peer-reviewed trials. Screening, data extraction and bias assessment were done in duplicate. Of 2,514 records, 74 full texts were retrieved and 19 included. Samples ranged from 25 to 958 people and interventions from 2 weeks to 1 year, with a median of 6 weeks.
Participants included apparently healthy adults, older adults, adolescents, families, inactive adults, cancer survivors, adults with a body mass index of 25 to 35, adults with insomnia and children with sleep difficulties. Eleven interventions were chatbot-only and eight added items such as wearable trackers, mailouts or websites. Eight chatbots were text-only, three voice-only and eight mixed formats.
Evidence on sleep was graded B and evidence on physical activity and fruit and vegetables A. Sleep quality results varied widely between studies (I2 = 80%). Subgroup tests for most outcomes found no difference by chatbot duration, output type or AI use, apart from fruit and vegetable intake.
The authors also flagged privacy risks, noting chatbots may store chat logs and intimate data. They noted that AI cannot truly empathise and that people with low digital literacy may be excluded.
Sources
1 on file- npj Digital Medicinedoi:10.1038/s41746-023-00856-1
Talk it over
Who gets this benefit, and who was left out of the study or story?
Chatbots nudged people to walk and sleep more, review of 19 trials findshttps://truvace.com/item/systematic-review-and-meta-analysis-of-the-effectiveness-of-chatbots-on-lifestyl
Who gets this benefit, and who was left out of the study or story?
The gain A systematic review and meta-analysis of 19 trials with 3,567 participants found chatbot interventions raised daily steps by 735, fruit and vegetable intake by about 1 serving a day and sleep by 45 minutes a night.
Answer in the threadHow do we know? What can this kind of evidence show, and what can’t it?
The evidence This rests on a systematic review or meta-analysis with a sample of 3,567, from npj Digital Medicine.
Answer in the threadWhat does it leave open? What else would you want to know before acting on it?
The record’s own caveat Over half of the studies (58%) had 75 or fewer participants and most ran six weeks or less with none longer than a year; participants were mostly self-referred and may be more tech-savvy.
Answer in the threadHave you seen AI used this way where you study, work or live? Did it go the same way?
Answer in the thread
The questions are the same on every record. The details under them are taken from this record as written.
What readers say