Objectives This systematic review and meta‐analysis aimed to examine non‐pharmacological interventions for helping people with dementia who experience feeding difficulties in order to improve their nutritional status. Methods The articles were searched using PsycINFO, Medline, PubMed, CINAHL and Cochrane. Two independent investigators critically appraised eligible studies. The PRISMA guidelines and checklist were used. The possibility of risk of bias was assessed using a tool to assess the quality of randomised control trials (RCT) and non‐RCT studies. A narrative synthesis was conducted as a method of synthesis. The Cochrane Review Manager (RevMan 5.4) was used for meta‐analysis. Results The systematic review and meta‐analysis included seven publications. Six interventions were identified and categorised as: eating ability training for people with dementia, staff training and feeding assistance and support. The meta‐analysis found evidence of the effect of eating ability training on feeding difficulty, quantified by the Edinburgh Feeding Evaluation in Dementia scale (EdFED) with a weighted mean difference of −1.36 (95% confidence interval: −1.84 to −0.89, p < 0.001) and on self‐feeding time. A spaced retrieval intervention showed a positive effect on EdFED. The systematic review discovered that while feeding assistance had a positive effect on feeding difficulty, staff training had no effect. According to the meta‐analysis, these interventions had no effect on improving the nutritional status of people with dementia. Conclusions None of the included RCTs met the Cochrane risk‐of‐bias criteria for randomised trials. This review found that direct training for people with dementia and indirect feeding support from care staff resulted in fewer mealtime difficulties. More RCT studies are needed to determine the efficacy of such interventions.