Study selection

A total of 7998 studies were initially screened using the established search strategy, 3156 duplicates were excluded, 4699 unrelated literatures were excluded by reading titles and abstracts, 143 literatures were obtained after primary screening, and 110 literatures were excluded by reading the full text. Finally, 32 literatures were included. The literature screening process is shown in Fig. 1.

Fig. 1

Flow diagram of literature screening

General information regarding the included literature

A total of 33 articles were included, including 16 English articles [12, 15,16,17,18,19,20,21,22,23,24,25,26,27,28,29] and 16 Chinese articles [11, 30,31,32,33,34,35,36,37,38,39,40,41,42,43,44], all of which were RCTs. The combined sample size was 9514 cases, including 4514 cases in the experimental group and 5000 cases in the control group, General information about the 33 studies was shown in Table 1.

Table 1 Characteristics of included studiesQuality evaluation of the included literature

A total of 32 RCTs were included in this study, all of which clearly defined the inclusion and exclusion criteria of subjects, were comparable at baseline, and the measurement results of the experimental and control groups were measured using the same tools and the same statistical methods. All 32 studies [12, 15,16,17,18,19,20,21,22,23,24,25,26,27,28,29] described the generation method of random sequence, and 16 studies [12, 15,16,17, 19, 21,22,23,24,25, 27, 30, 32, 36, 39, 44] described allocation hiding; 20 studies [12, 15,16,17,18,19, 21, 22, 24, 26, 28,29,30,31, 34, 39] describe blinding the research subjects and intervention implementers, and 20 studies [12, 19,20,21, 24, 26,27,28, 33,34,35,36,37,38, 40,41,42,43,44,45] describe blinding the outcome evaluators. The quality level of 5 studies is A level, while the rest are B level, and the overall quality was at the medium level or above. The results of the risk of bias assessment were shown in Fig. 2.

Fig. 2

Risk of bias in the included studies

Meta-analysis resultsEffect of the Internet intervention model on breastfeeding knowledge

Six studies [14, 17, 22, 25, 28, 37] reported the impact of the Internet intervention model on breastfeeding knowledge. Due to different evaluation tools used, SMD was selected for the combination of effect sizes, and the results showed significant heterogeneity (P < 0.01, I2 = 95%). After excluding the study of Guo Yuxin[44]by sensitivity analysis, there was no significant decrease in heterogeneity, and the random effects model was used for meta-analysis. The results showed that after intervention, the score of breastfeeding knowledge in the experimental group was higher than that in the control group, and the difference was statistically significant [SMD = 1.88, 95% CI (1.09, 2.67), P < 0.01], as shown in Fig. 3.

Fig. 3

Influence of the Internet intervention model on breastfeeding knowledge

Effect of the Internet intervention model on breastfeeding attitude

Five studies [16, 20, 34, 35, 45] reported the effect of the Internet intervention model on attitudes towards breastfeeding. SMD was used to combine effect sizes, and the results showed significant heterogeneity (P < 0.01, I2 = 87%). After excluding the study of Hannula [16] by sensitivity analysis, there was no significant decrease in heterogeneity, and the random effects model was used for meta-analysis. The results showed that the scores of breastfeeding attitude in the experimental group were higher than those in the control group after intervention, and the difference was statistically significant [SMD = 0.56, 95%CI (0.17, 0.95), P = 0.004], which was shown in Fig. 4

Fig. 4

Influence of the Internet intervention model on breastfeeding attitude

Effect of the Internet intervention model on breastfeeding confidence

Six studies [16, 24, 25, 32, 36, 43] reported the effect of the Internet intervention model on attitudes towards breastfeeding. SMD was used to combine effect sizes, and the results showed significant heterogeneity (P < 0.01, I2 = 98%). After excluding the study of Hannula [16] by sensitivity analysis, there was no significant decrease in heterogeneity, and the random effects model was used for meta-analysis. The results showed that the scores of breastfeeding confidence in the experimental group were higher than those in the control group after intervention, and the difference was statistically significant [SMD = 1.69, 95%CI (0.55, 2.84), P = 0.004], which was shown in Fig. 5

Fig.5

Influence of the Internet intervention model on breastfeeding confidence

Effect of the Internet intervention model on the rate of exclusive breastfeeding within 6 weeks postpartum

Six studies [16, 17, 22, 23, 32, 41, 45] reported the effect of Internet intervention mode on the rate of exclusive breastfeeding at discharge, and the results showed significant heterogeneity (P < 0.01, I2 = 80%). After excluding the study of Seguranyes [17] by sensitivity analysis, there was no significant decrease in heterogeneity, and the random effects model was used for meta-analysis. The results showed that the rate of exclusive breastfeeding in the experimental group was higher than that in the control group at discharge after the intervention, and the difference was statistically significant [RR = 1.12, 95%CI (1.00, 1.24), P = 0.04], as shown in Fig. 6. Ten studies [