Longitudinal associations between internet access and cognitive decline in older adults Authors: Hanqian Wang, Yueyao Wang, Yudong Wei, Lu Li Published: 2026 ABSTRACT AbstractThe Internet holds potential for alleviating cognitive decline through cognitive stimulation and social engagement. However, existing studies often overlook the crucial role of Internet access as the foundational layer of the digital divide. This study investigated the association between Internet access and cognitive function among middle-aged and older adults and examined age-specific differences in this relationship. Data were drawn from the 2015 and 2018 waves of the China Health and Retirement Longitudinal Study (CHARLS). Participants aged ≥ 50 years with complete follow-up cognitive assessments were included (n = 7721; mean age = 60.48, SD = 9.37; male, 56.1%). Lagged dependent variable models were used to evaluate associations between Internet access and cognitive function. Chain mediation analyses were used to test whether family connection mediated this association. Internet access was significantly associated with better cognitive outcomes over time. Participants with Internet access demonstrated greater significant improvements in episodic memory (β = 0.29,95%CI = 0.14–0.44) and mental status (β = 0.28,95%CI = 0.12–0.44) compared to those without access. Age-stratified models indicated stronger effects in adults aged 50–59 for episodic memory (β = 0.36,95%CI = 0.17–0.55) and aged 60–69 for mental status (β = 0.42,95%CI = 0.15–0.69). Chain mediation analyses revealed that Internet access enhanced family connection, thereby contributing to better cognitive function. Internet access is positively associated with cognitive function in mid- to later life, partly through strengthened family ties. These findings underscore the importance of digital inclusion policies to support cognitive health and promote healthy aging.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-50958-5.Keywords: Internet access, Cognitive function, Lagged dependent-variable model, Older adults, Longitudinal studySubject terms: Health care, Neuroscience, Psychology, Psychology Introduction As the global population ages, cognitive decline has become an increasingly pressing public health issue, imposing substantial medical, social, psychological, and financial burdens on families and society1. Cognitive decline is a precursor of dementia and may lead to physical disability and even death2. In China, about 15.5% of people aged 60 and above have mild cognitive impairment, and 6.0% have dementia3, underscoring the urgent need for strategies that prevent or delay cognitive deterioration in an aging population4. In recent years, the potential protective effects of Internet access and digital engagement have attracted considerable research attention. Existing studies indicate that Internet access or use is associated with improved physical and cognitive health among older adults5. Internet access or use provides a platform for cognitive stimulation, which may slow cognitive decline and enhance function6–9. Studies from high-income countries (HICs) suggest that Internet access may reduce the risk of cognitive decline and dementia through mechanisms involving cognitive training, information processing, and daily functional engagement8,10–12. Experimental and observational studies further support the notion that digital activities, such as online information searching or cognitively demanding digital games, can enhance cognitive performance or mitigate age-related decline10,13. However, considering the tremendous variation in digital levels across countries, there may be differences in the patterns of association between Internet access or use and cognitive function between low- and middle-income countries (LMICs) and HICs14. Despite recent increases in digital inclusion, a digital divide remains pronounced among Chinese older adults15. According to the 56th Chinese Statistical Report on Internet Development, Internet use among adults aged 60 and older continues to grow, yet many older individuals still face barriers related to affordability, device proficiency, infrastructure, and perceived usefulness15,16. As a result, a large share of older adults remains excluded from digital environments and their potential cognitive and social benefits. This reflects a broader digital divide, which is widely recognized as a multidimensional phenomenon, encompassing disparities in access to digital technologies, patterns of use, and digital literacy15. Among these dimensions, access represents a fundamental prerequisite for digital participation, particularly for older adults facing structural barriers to entering the digital environment, making it a critical starting point for understanding digital inequality. Existing studies have examined the relationship between Internet use and cognitive function. For instance, digital device ownership is protective against cognitive decline7, and different types of Internet activities may confer different cognitive benefits17. Moreover, the Internet can facilitate both social and cognitive engagement, which are important for maintaining cognitive reserve18. However, prior research has largely focused on Internet use, while less attention has been paid to the more fundamental prerequisite of digital participation, namely basic Internet access. In this study, Internet access is conceptualized as the ability and opportunity to connect to the Internet, reflecting basic accessibility rather than digital literacy, frequency, or purpose of use. Focusing on this dimension allows us to capture older adults who may be socially vulnerable or excluded from the digital environment, providing insights into structural inequalities in digital participation. Furthermore, few studies have examined the social mechanisms through which digital access may influence cognitive aging18–20. In the Chinese context, family represents a primary source of emotional support and social interaction in later life21. Digital access may facilitate social interactions, particularly communication with family members, thereby enhancing emotional and social support22. According to social capital and cognitive reserve theories, stronger family connections can buffer cognitive decline by providing emotional support, stimulating cognitive engagement, and promoting adaptive health behaviors18,20,23. Empirical studies have shown that older adults who use digital devices to communicate with family experience greater family connectedness, reduced social isolation, and better cognitive outcomes19,20,23. However, empirical evidence on whether and how family connection mediates the relationship between digital access and cognitive function over time remains limited. Therefore, we hypothesized that family connection may serve as a mediating pathway linking digital access to cognitive function. To address these gaps, the present study used data from a nationally representative longitudinal survey, drawing on two waves conducted in 2015 (baseline) and 2018 (follow-up), to examine the association between Internet access and cognitive function among middle-aged and older Chinese adults. We further explored age-specific differences in these associations and tested a chain mediation model in which digital access influences cognitive function indirectly through changes in family connection across time. Based on previous evidence linking digital participation with social engagement and cognitive health7,24, we hypothesized that Internet access would be associated with better cognitive outcomes, and that family connection would serve as a meaningful social pathway linking digital access to cognitive function. Methods Participants Data for this study came from the China Health and Retirement Longitudinal Study (CHARLS), a nationally representative longitudinal survey of adults aged 45 years or older in China17. This study used data from two waves conducted in 2015 and 2018, corresponding to the baseline and follow-up surveys. Participants were restricted to those aged 50 and above in 2015 (n = 16476). Participants who had missing values on cognitive tests in 2015 (n = 3925) and 2018 (n = 4830) were excluded, which resulted in a final sample size of 7721 participants. Measures Cognitive function Cognitive function in CHARLS was assessed using a structured cognitive measurement developed based on the framework of the American Health and Retirement Study (HRS). This measurement has been widely used in prior studies based on CHARLS and has demonstrated acceptable reliability and validity in Chinese older populations25–30. Following prior literature, we constructed two domains of cognitive function: episodic memory and mental status. Episodic memory was measured using immediate and delayed 10-word recall test following the CHARLS protocol25,27. Participants were asked to recall the word list immediately and again after a five-minute delay. The episodic memory score was calculated as the average of immediate and delayed recall scores, ranging from 0 to 10. Detailed procedures are provided in the Supplementary Methods. Mental status was derived from selected components based on the framework of the Telephone Interview of Cognitive Status (TICS), including numeracy, orientation, and a figure drawing task. Numerical ability was assessed using the serial sevens test, in which participants were asked to successively subtract seven from 100 up to five times. Orientation was evaluated by asking participants to report the date, month, year, and day of the week. In addition, participants were asked to replicate as similarly as possible a picture of two overlapping pentagons, as was shown by the interviewers. Each correct response was assigned one point, and scores across all items were summed to generate a composite mental status score ranging from 0 to 10. Higher scores on both episodic memory and mental status indicate better cognitive function. Internet access Internet access was assessed using the question “Have you used the Internet in the past month?”. This measure identifies whether participants had the opportunity and means to connect to the Internet (i.e., physical connectivity), but does not capture frequency, duration, purpose, or digital literacy. In this study, we defined this measure as Internet access, reflecting basic connectivity rather than more complex dimensions of digital engagement. The response “yes” was categorized as having Internet access, while the response “no” was considered as not having Internet access. Family Connection Family connection was assessed using three binary items reflecting intergenerational contact patterns: (1) whether the participant lived with at least one child, (2) whether the participant contacted any of their children at least once per week, and (3) whether any child lived in the same community or neighborhood. These indicators were selected to capture key dimensions of family relationships, including structural proximity and interaction frequency, consistent with prior studies using similar aging surveys31,32. Co-residence and geographic proximity reflect opportunities for face-to-face interaction and instrumental support, while frequency of contact reflects ongoing social interaction and emotional connection31,32. These three indicators were summed to create a composite index ranging from 0 to 3, with higher scores indicating a stronger family connection. Covariates In this study, covariates were selected based on previous literature19,30, and additional variables related to cognitive function were deliberately excluded to streamline the model and reduce the risk of overfitting. Several demographic variables and behavioral, psychological, and clinical risk factors were included as covariates. Age, gender (male or female), marital status (married or otherwise), education level (illiterate, less than primary school, primary school, middle school, high school or above), and labor force status (currently not working or currently working) were measured by self-report. Health behaviors included currently drinking or not and currently smoking or not19. Chronic diseases, including hypertension, diabetes, and heart disease, were assessed by the following question: “Have you been diagnosed by a doctor with these conditions?” Functional ability was measured by the performance of activities of daily living (ADLs: dressing, bathing, eating, getting into or out of bed, using the toilet, and controlling urination and defecation) and instrumental activities of daily living (IADLs: preparing a hot meal, shopping for groceries, using the phone, taking medicines, and managing money)19,33. According to previous research33, both ADLs and IADLs were dichotomized into no limitation or at least one limitation. Depressive symptoms were assessed with the 10-item Center for Epidemiological Studies Depression Scale (CESD-10). Modified CESD scoring was used by excluding the loneliness question and calculating the sum of the remaining nine questions (CESD-9, range 0–27)33. Loneliness was measured with a single item included in the CESD: “In the last week, how often did you feel lonely?” and was dichotomized into 0 (not lonely =felt lonely rarely or none of the time), and 1 (lonely=felt lonely sometimes, occasionally, or most of the time). Statistical analysis Descriptive analyses reported frequencies and percentages, and means and standard deviations of all variables. We used chi-square tests and t-tests to compare sample characteristics between participants with and without Internet access. To examine relationships among key study variables, Pearson correlation analyses were conducted for digital access, family connection in 2015 (T1) and 2018 (T2), and cognitive function (T2). To examine the association between Internet access and cognitive function, we used lagged dependent-variable regression models with Ordinary Least Squares estimation. The lagged dependent-variable model is superior for analyzing the effects of predictor variables on an outcome with two-wave panel data while controlling for the influence of time-invariant variables34. Internet access was measured at baseline (2015) and treated as a time-lagged exposure in subsequent analyses. Models were fitted for both cognitive measurements and overall cognitive function. Model 1 A was constructed to examine the associations between Internet access in 2015 and episodic memory in 2018 by partially adjusting for control variables, including age, gender, marital status, education level, labor force status, and episodic memory in 2015. Other control variables, including chronic diseases, smoking and drinking status, ADL and IADL disabilities, were then added in Model 1B. Model 1 C added CESD-9 scores and loneliness in the fully adjusted model. Three similar models were fitted to test the independent associations of Internet access with mental status (Model 2 A, 2B, and 2 C) and overall cognitive function (Model A, B, and C). Stratified analysis by age was conducted to test for differential effects of Internet access and episodic memory and mental status across subgroups. Linear mixed-effect models were used to assess the longitudinal association of Internet access and cognitive function over time. A fixed effect for each variable was estimated, adjusting for the effect of the individual (a random effect). Sensitivity analyses were conducted in subgroups with higher education levels (middle school and above), where Internet access was more prevalent and illiterate participants were excluded. In addition, we performed a sensitivity analysis excluding participants with potential cognitive impairment at baseline, defined as having a total cognitive score (episodic memory and mental status) ≤ 1.5 standard deviations below the sample mean, following prior research35. In addition, a chain mediation analysis was conducted to test the mediating role of family connection between digital access and cognitive function. The significance of the mediation effects was tested using the bootstrap method with 5000 iterations, with mediation considered significant if the 95% confidence interval (CI) did not include zero. Mediation analyses were conducted using Mplus 8.3, while other statistical analyses were conducted in R 4.3.1. Statistical significance was set at P < 0.05. Results Characteristics of Participants Data from 7721 participants aged 50 or above (mean age 60.48, SD = 9.37) were included for analysis. Of all the participants, 56.1% were male and 43.9% were female (Table 1). A total of 90.0% of the participants were married or partnered, 82% had a middle school or below education, and 66.2% were working currently. Most of the participants did not have hypertension (61.2%), diabetes (81.6%), or heart disease (74.3%), nor did they smoke (69.0%) or drink alcohol (60.3%). The incidences of ADL disability and IADL disability were 16.2% and 14.6%, respectively. The mean scores (SD) for episodic memory and mental status of cognitive function and depression symptoms were 3.80(1.62), 8.06(1.86), and 8.23(4.41), respectively. 21.8% of participants reported feeling lonely sometimes, occasionally, or most of the time. Participants with Internet access were more likely to be younger, male, better educated, currently not working, drinking alcohol, and less likely to have hypertension, ADL disability, and IADL disability. We also found that participants with Internet access had lower levels of depressive symptoms, and higher episodic memory and mental status scores compared to those without Internet access (all P-values < 0.05). Table 1Characteristics of participants by Internet access (N = 7721).VariablesFull sampleN (%)Internet access P No (n = 7161)Yes (n = 560)Age, M(SD)60.48(7.59)60.74(7.61)57.17(6.48)< 0.001Gender0.015Male4335(56.1)3993(55.8)342(61.1)Female3386(43.9)3168(44.2)218(38.9)Marital status0.164Married or partnered6952(90.0)6438(89.9)514(91.8)Otherwise769(10.0)723(10.1)46(8.2)Education level< 0.001Illiterate713(9.2)711(9.9)2(0.4)Less than Primary school1238(16.0)1231(17.2)7(1.3)Primary school2184(28.3)2111(29.5)73(13.0)Middle school2194(28.4)2040(28.5)154(27.5)High school or above1392(18.0)1068(14.9)324(57.9)Labor force status0.001Currently not working2505(32.4)2990(32.4)215(39.6)Currently working5110(66.2)4782(67.6)328(60.4)Missing106(1.4)--Hypertension0.003No4724(61.2)4372(66.0)352(72.6)Yes2383(30.9)2250(34.0)133(27.4)Missing614(8.0)--Diabetes0.127No6303(81.6)5880(89.5)423(87.2)Yes754(9.8)692(10.5)62(12.8)Missing664(8.6)--Heart disease0.231No5737(74.3)5356(81.1)381(78.9)Yes1352(17.5)1250(18.9)102(21.1)Missing632(8.2)--Currently smoking0.704No5327(69.0)4936(69.0)391(69.8)Yes2391(31.0)2222(31.0)169(30.2)Missing3(0.0)--Currently drinking< 0.001No4653(60.3)4404(61.5)249(44.6)Yes3063(39.7)2754(38.5)309(55.4)Missing5(0.1)--ADL disability< 0.001No6469(83.8)5943(83.0)526(93.9)Yes1252(16.2)1218(17.0)34(6.1)IADL disability< 0.001No6597(85.4)6052(84.5)545(97.3)Yes1124(14.6)1109(15.5)15(2.7)CESD-9, M(SD)8.23(4.41)8.31(0.48)7.23(3.28)< 0.001Loneliness< 0.001No6034(78.2)5561(77.7)473(84.5)Yes1687(21.8)1600(22.3)87(15.5)Family connection, M(SD)1.81(1.03)1.74(1.02)1.80(1.04)0.139Cognitive function11.86(2.80)11.68(2.77)14.20(1.95)< 0.001Episodic memory, M(SD)3.80(1.62)3.71(1.59)4.94(1.48)< 0.001Mental status, M(SD)8.06(1.86)7.97(1.88)9.25(1.01)< 0.001 Characteristics of participants by Internet access (N = 7721). Internet access and cognitive function Tables 2 and 3 shows the results of lagged regression on overall cognitive and episodic memory and mental status, respectively. Higher scores of episodic memory and mental status in 2015 were associated with higher corresponding cognitive function scores three years later. Similar results were observed in overall cognitive function. Specifically, Internet access was significantly associated with better cognitive function (episodic memory: β = 0.34, 95%CI = 0.19–0.49; mental status: β = 0.33, 95%CI = 0.17–0.49) after controlling sociodemographic variables and episodic memory and mental status scores in 2015 (Model 1 A and Model 2 A, respectively). These associations remained significant (episodic memory: β = 0.30, 95%CI = 0.15–0.45; mental status: β = 0.29, 95%CI = 0.13–0.45) when other covariates such as chronic diseases, health behaviors, and disabilities were included in the models (Model 1B and Model 2B, respectively). Internet access was significantly associated with better cognitive function (episodic memory: β = 0.29, 95%CI = 0.14–0.44; mental status: β = 0.28, 95%CI = 0.12–0.44) after controlling all the variables (Model 1 C and Model 2 C, respectively). Sensitivity analyses restricted to participants with higher education levels yielded consistent results (Supplementary Table 4). Similarly, after excluding participants with potential cognitive impairment at baseline, the associations remained statistically significant (Supplementary Table 5). Table 2Lagged regression models on cognitive function in 2018.VariablesCognitive function in 2018Model Aβ(95%CI)Model Bβ(95%CI)Model Cβ(95%CI)T1 Internet access0.48(0.22,0.73)***0.44(0.18,0.69)**0.43(0.17,0.69)**T1 Cognitive function0.55(0.53,0.58)***0.54(0.52,0.57)***0.54(0.51,0.56)***Age-0.04(-0.05,-0.03)***-0.04(-0.05,-0.03)***-0.04(-0.05,-0.04)***Gender0.14(0.01,0.27)*0.01(-0.16,0.17)0.04(-0.13,0.20)Marital status0.66(0.60,0.72)***0.64(0.58,0.70)***0.64(0.58,0.70)***Education level-0.43(-0.64,-0.21)***-0.41(-0.63,-0.20)***-0.40(-0.62,-0.18)***Labor force status-0.29(-0.43,-0.15)***-0.28(-0.42,-0.13)***-0.27(-0.41,-0.12)***Hypertension-0.05(-0.19,0.09)-0.05(-0.19,0.09)Diabetes-0.03(-0.23,0.18)-0.02(-0.23,0.19)Heart disease0.34(0.18,0.51)***0.36(0.19,0.53)***Currently smoking-0.33(-0.49,-0.18)***-0.33(-0.48,-0.17)***Currently drinking0.08(-0.06,0.23)0.09(-0.05,0.23)ADL disability-0.17(-0.36,0.01)-0.13(-0.32,0.05)IADL disability-0.19(-0.39,0.00)-0.16(-0.35,0.04)CESD-9-0.02(-0.03,-0.00)*Loneliness-0.06(-0.22,0.11)ADL, activities of daily living; IADL, instrumental activities of daily living; CESD, Center for Epidemiologic Studies Depression scale.β = standardized regression coefficient.*P < 0.05, **P < 0.01, ***P < 0.001. Lagged regression models on cognitive function in 2018. ADL, activities of daily living; IADL, instrumental activities of daily living; CESD, Center for Epidemiologic Studies Depression scale. β = standardized regression coefficient. *P < 0.05, **P < 0.01, ***P < 0.001. Table 3Lagged regression models on episodic memory and mental status in 2018.VariablesEpisodic memory Score in 2018Mental status score in 2018Model 1 Aβ(95%CI)Model 1Bβ(95%CI)Model 1 Cβ(95%CI)Model 2 Aβ(95%CI)Model 2Bβ(95%CI)Model 2 Cβ(95%CI)T1 Internet access0.34(0.19,0.49)***0.30(0.15,0.45)***0.29(0.14,0.44)***0.33(0.17,0.49)***0.29(0.13,0.45)***0.28(0.12,0.44)**T1 EM0.39(0.36,0.41)***0.38(0.35,0.41)***0.37(0.35,0.40)***T1 MS0.48(0.46,0.51)***0.47(0.45,0.49)***0.47(0.44,0.49)***Age-0.04(-0.05,-0.04)***-0.04(-0.05,-0.04)***-0.04(-0.05,-0.04)***-0.01(-0.01,-0.00)**-0.01(-0.01,-0.00)*-0.01(-0.01,-0.00)**Gender0.27(0.19,0.35)***0.18(0.08,0.28)***0.21(0.11,0.30)***-0.10(-0.19,-0.01)*-0.15(-0.26,-0.05)**-0.13(-0.23,-0.02)*Marital status-0.32(-0.44,-0.20)***-0.30(-0.42,-0.18)***-0.25(-0.37,-0.13)***-0.28(-0.41,-0.15)***-0.25(-0.38,-0.13)***-0.21(-0.33,-0.08)**Education level0.41(0.38,0.45)***0.39(0.36,0.43)***0.39(0.36,0.43)***0.30(0.26,0.34)***0.28(0.25,0.32)***0.28(0.24,0.32)***Labor force status-0.06(-0.15,0.02)-0.07(-0.16,0.01)-0.06(-0.15,0.02)-0.16(-0.25,-0.07)**-0.17(-0.27,-0.08)***-0.16(-0.26,-0.07)**Hypertension-0.09(-0.17,-0.01)*-0.09(-0.17,-0.01)*-0.02(-0.10,0.07)-0.01(-0.09,0.08)Diabetes-0.07(-0.18,0.04)-0.07(-0.17,0.04)0.08(-0.04,0.19)0.08(-0.03,0.20)Heart disease0.20(0.10,0.29)***0.22(0.12,0.31)***0.10(0.00,0.20)*0.12(0.02,0.22)*Currently smoking-0.26(-0.35,-0.17)***-0.26(-0.35,-0.16)***-0.17(-0.27,-0.07)**-0.16(-0.26,-0.06)**Currently drinking0.08(-0.01,0.17)0.08(-0.01,0.16)0.03(-0.07,0.12)0.02(-0.07,0.11)ADL disability-0.12(-0.24,-0.01)*-0.06(-0.18,0.06)-0.29(-0.41,-0.17)***-0.23(-0.35,-0.10)***IADL disability-0.17(-0.28,-0.06)**-0.12(-0.24,-0.01)*-0.24(-0.36,-0.12)***-0.19(-0.31,-0.07)**CESD-9-0.02(-0.03,-0.01)***-0.02(-0.03,-0.01)***Loneliness-0.18(-0.27,-0.09)***-0.18(-0.28,-0.09)***EM, episodic memory; MS, mental status; ADL, activities of daily living; IADL, instrumental activities of daily living; CESD, Center for Epidemiologic Studies Depression scale.β = standardized regression coefficient.*P < 0.05, **P < 0.01, ***P < 0.001. Lagged regression models on episodic memory and mental status in 2018. EM, episodic memory; MS, mental status; ADL, activities of daily living; IADL, instrumental activities of daily living; CESD, Center for Epidemiologic Studies Depression scale. β = standardized regression coefficient. *P < 0.05, **P < 0.01, ***P < 0.001. Internet access and cognitive function by age groups We further explored whether the association between Internet access and cognitive function (episodic memory and mental status) varied by age among the three age groups: 50–59 years, 60–69 years, and 70 years or older. As shown in Supplementary Table 1, a significant relationship was found between Internet access and episodic memory for the group aged 50–59 years (β = 0.36, 95%CI = 0.17–0.55). In this age group, with Internet access every year, there was an additional 0.36 increase in scores of episodic memory. A significant relationship could also be observed between Internet access and mental status for the group aged 60–69 years (β = 0.42, 95%CI = 0.15–0.69) (Supplementary Table 2). In this age group, with Internet access every year, there was an additional 0.42 increase in scores of mental status (Table 4). Table 4Adjusted longitudinal association between Internet access and cognitive function over time (N = 7721).VariablesPoint estimates (95%CI) P T1 Episodic memory ScoreNo Internet accessrefHave Internet access0.64(0.49,0.79)< 0.001T2 Episodic memory ScoreNo Internet accessrefHave Internet access0.47(0.35,0.59)< 0.001T1 Mental Status ScoreNo Internet accessrefHave Internet access0.61(0.44,0.78)< 0.001T2 Mental Status ScoreNo Internet accessrefHave Internet access0.62(0.48,0.75)< 0.001 Adjusted longitudinal association between Internet access and cognitive function over time (N = 7721). The effect of Internet access on cognitive function over time In the episodic memory score in 2015, participants with Internet access were 0.64 (95% CI = 0.49–0.79) higher than those without Internet access (Table 3). Likewise, in the mental status score in 2015, participants with Internet access were 0.61 (95% CI = 0.44–0.78) higher than those without Internet access. In congruence with the observed patterns in the independent relationship between Internet access and cognitive function, a longitudinally protective association was discerned during the extended three-year follow-up period (episodic memory: β = 0.47, 95% CI = 0.35–0.59; mental status: β = 0.62, 95%CI = 0.48–0.75). Mediation analysis This study employed Pearson correlation analysis to examine the associations among digital access, family connection, and cognitive function at T1 and T2 (Supplementary Table 3). All the variables were significantly correlated with each other. Figure 1 indicates that digital access at T1 was significantly and directly associated with cognitive function at T2 (β = 0.096). The mediation model revealed that digital access at T1 was significantly and directly associated with family connection at T1 (β = 0.233), and family connection at T1 was significantly associated with cognitive function at T2 (β = 0.393). Digital access at T1 was also significantly linked to family connection at T2 (β = 0.572), and family connection at T2 had a significantly positive association with cognitive function at T2 (β = 0.113). The path coefficient of the chain mediation model. Discussion In this study, a nationally representative sample of Chinese older adults was selected to examine the association between Internet access and cognitive function. Our findings supported the initial hypothesis that Internet access was linked to improvements over time in episodic memory, mental status, and overall cognitive function. Older adults with Internet access exhibited greater gains in both episodic memory and mental status scores than those without access. Moreover, we identified age-specific effects, with Internet access related to higher episodic memory among adults aged 50–59 years and higher mental status among those aged 60–69 years. This study also contributes new evidence by demonstrating that family connection serves as a mediating pathway in the association between digital access and cognitive health. The chain mediation model revealed that digital access in earlier stages may facilitate stronger emotional and social ties with children, and that such ties continue to exert positive influence over time, ultimately promoting better cognitive function. This finding reflects the enduring significance of family-based social support in cognitive aging. Our results revealed a consistent correlation between Internet access and cognitive function across both baseline and follow-up assessments. The finding aligns with a previous study showing that accessing or using the Internet may be associated with significantly lower risk of cognitive decline7 and provides cognitive benefits or mitigates cognitive decline over time36,37. Internet access may provide cognitive benefits through multiple mechanisms, including enhanced opportunities for social interaction, engagement in cognitively stimulating digital activities, and exposure to new information and learning environments8,12. These activities may help preserve cognitive reserve, foster adaptive neural responses to age-related degeneration, and delay cognitive deterioration12,17,38. Accordingly, addressing inequalities in digital access could have meaningful public health implications for reducing cognitive decline in later life. However, Internet access may also function as a proxy for broader socioeconomic and health-related advantages accumulated over the life course. Individuals with access to the Internet are more likely to have higher levels of education, better socioeconomic conditions, and more favorable health profiles, all of which are independently associated with cognitive function7,19,28. Although we adjusted for several of these factors in our models, residual confounding cannot be fully ruled out. Therefore, the observed associations should not be interpreted as purely causal, but rather as reflecting both the potential effects of digital engagement and underlying social and health inequalities. In this study, we observed a trend toward a significant longitudinal association between episodic memory scores and Internet access, as well as between mental status scores and Internet access. Although the regression coefficient is small, our findings are significant given that the Internet has a positive impact on cognitive function. Previous research has suggested that individuals with access to digital devices (such as desktops or cellphones) have better cognitive function and a lower rate of cognitive decline than those without access7. General Internet use is associated with improved cognitive function and a reduced risk of dementia in several previous observational studies39–41. Additionally, the protective effects strengthened over longer follow-up periods. This is consistent with previous longitudinal studies reporting greater cognitive benefits associated with longer exposure to digital environments7. These results suggest that sustained and structured interventions aimed at promoting Internet access and basic digital proficiency may further enhance cognitive outcomes in aging populations42–44. Age-stratified analyses revealed heterogeneity in the effects of Internet access across cognitive domains. There is a significant relationship between Internet access and episodic memory scores in the population aged 50–59 years, and a significant relationship with mental status scores among participants aged 60–69 years. The findings are partially in line with our hypothesis and previous evidence, suggesting that different domains of cognitive function follow distinct age-related trajectories45–47. Word recall performance tends to decline with age, particularly after the mid-60s45. The association between Internet access and higher episodic memory in the 50–59 age group may reflect relatively preserved memory function before a more pronounced age-related decline. However, the observed age-specific patterns should be interpreted with caution. Differences across age groups may also be influenced by statistical or distributional factors, such as variation in sample size, score variability, or statistical power within subgroups. Moreover, the absence of significant associations among adults aged 70 years or older may be partly due to smaller sample sizes or more advanced cognitive aging reducing sensitivity to digital influences. Nonetheless, these findings highlight the potential value of considering age when designing digital inclusion and cognitive health interventions. A key contribution of this study lies in the demonstrated chain mediation effect of family connection. Digital access provides older adults with additional channels for maintaining contact with children, especially when living apart or facing mobility limitations. This finding is consistent with research showing that ICT use enhances intergenerational relationships, reduces loneliness, and strengthens emotional support networks48. Within the frameworks of social capital and cognitive reserve theories, these findings suggest that digital access enhances social interactions and family support, which are protective for cognitive health18,20,23. Stronger and more sustained family ties can promote psychological resilience, reduce stress, and provide emotional meaning, all of which are protective for cognitive health19. Continued social engagement may stimulate cognitive processes and help maintain cognitive reserve, thereby delaying cognitive decline20. These findings suggest that cognitive benefits of digital access may arise not only from individual cognitive engagement, but also from its capacity to maintain meaningful social integration and family-based support systems. This is particularly relevant in China, where family remains a central source of emotional companionship and caregiving in older age. Thus, interventions should not only focus on improving digital skills among older adults, but also encourage family members to actively engage in digital communication, thereby strengthening intergenerational ties and supporting cognitive health. The present findings suggest that access to the Internet can decrease the likelihood of cognitive decline in older adults, highlighting the importance of strategies to bridge the digital access divide. The Internet provides numerous benefits for older adults, including access to health-related information and alternative ways of accessing healthcare49–51 as well as new channels for social interaction and entertainment19. These benefits can promote social participation, which is beneficial for cognitive function. Additionally, Internet access and use can directly stimulate older adults to learn new things and increase their knowledge and skills in terms of digital technology7. A growing number of studies have demonstrated the effectiveness of digital health interventions on cognitive function, such as online cognitive training and the use of the Internet or e-mail52,53. Older adults who are exposed to the Internet can also strengthen their visuospatial abilities through activities, such as playing digital games and online shopping, which can enhance their visuospatial processing speed54. Enhancing digital access among older adults has the potential to support cognitive health, strengthen family relationships, and improve overall well-being. Policymakers should therefore prioritize efforts to bridge the digital divide through community-based digital literacy programs, intergenerational digital learning initiatives, and the development of age-friendly digital platforms. Reducing material, skill-based, and motivational barriers to digital participation is essential for ensuring equitable access to the cognitive, social, and health benefits that Internet use may afford. Nonetheless, further research is needed. Additional longitudinal and experimental studies should examine the causal pathways linking digital access, family interaction patterns, and cognitive aging, as well as variations across rural and urban settings and socioeconomic groups. Tailored interventions should also be tested to determine the most effective strategies for integrating digital engagement into daily life for older adults. Our findings highlight the importance of age-sensitive strategies to promote digital participation. Younger older adults (50–64 years) may benefit most from programs emphasizing early engagement and skill development, whereas interventions targeting older adults (65 + years) should account for potential mobility or cognitive limitations, provide additional guidance, and encourage family support to facilitate digital access. Tailoring digital inclusion programs to specific age groups can enhance cognitive resilience, reduce social isolation, and maximize the benefits of digital engagement across the older population. Our study makes several important contributions. First, this study provides empirical evidence from a nationally representative longitudinal cohort of middle-aged and older Chinese adults on the link between basic Internet access and cognitive function. Second, by integrating social capital and cognitive reserve theories, it highlights family connection as a potential mediating pathway, illustrating the social mechanisms through which digital participation may influence cognitive aging. Additionally, focusing on basic access rather than usage or digital skills emphasizes structural digital inequalities and identifies socially vulnerable older adults, advancing understanding of the digital divide’s role in cognitive health. The study has several limitations. First, as an observational longitudinal study, causal inferences between Internet access and cognitive function are limited. Internet access was used only at baseline, and changes in access status over time could not be captured. Future research with more detailed and long-term panel data is needed to explore these relationships and mechanisms. Second, the key independent and mediating variables were measured using simplified indicators. Internet access was assessed using a single-item indicator reflecting basic connectivity, without fully addressing the use divide, patterns of Internet activity, or digital literacy. Family connection was measured using structural and interactional indicators (co-residence, contact frequency, and geographic proximity), which may not fully reflect the emotional quality or depth of family relationships. These simplified measurements may limit the interpretation and the ability to fully represent the multidimensional nature of digital engagement and family relationships. Future research should take more comprehensive measures, including the timing and frequency of Internet use, changes in access over time, digital literacy, and qualitative aspects of family ties, to better understand the mechanisms linking digital access and cognitive function. Third, the relatively low proportion of participants with Internet access led to imbalanced exposure groups, and illiterate individuals were overrepresented in the non-Internet group. Although adjustments and sensitivity analyses were applied, the findings should be interpreted with caution. Fourth, differences in subgroup analyses by age may also reflect statistical or distributional effects rather than true heterogeneity, and should therefore be interpreted cautiously. Fifth, although we conducted sensitivity analyses excluding participants with potential cognitive impairment at baseline, residual confounding and reverse causality cannot be fully excluded, particularly with respect to baseline cognitive status and depressive symptoms. Moreover, by excluding samples with missing data on key variables, we inevitably excluded a significant number of subjects, potentially impacting the representativeness of the results. However, this approach ensures data integrity and prevents bias introduced by simple data imputation. Future research should consider employing rigorous multiple imputation methods and implementing stricter data collection to minimize missing data. Conclusion This study demonstrates that Internet access is positively associated with cognitive function among middle-aged and older adults, with effects that persist and strengthen over time. The age-differentiated patterns observed indicate that digital engagement may support distinct cognitive processes across the life course. Moreover, the mediating role of family connection highlights a social mechanism through which digital access promotes cognitive health. These findings underscore the importance of digital inclusion as an integral component of healthy aging policy. Expanding equitable Internet access and fostering supportive digital environments may help prevent or delay cognitive decline and mitigate the broader societal burden associated with population aging. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements We thank the China Health and Retirement Longitudinal Study team for providing data and training in using the datasets. We thank the students who participated in the survey for their cooperation. We thank all volunteers and staff involved in this research. Abbreviations Activities of daily living Instrumental activities of daily living Center for Epidemiological Studies Depression Scale Geriatric Depression Scale Author contributions HW contributed to the study design, data analysis and interpretation and drafted the manuscript. YueW contributed to the data analysis and interpretation and critical revision of the manuscript. YuW and LL contributed to the critical revision of the manuscript. LL has the primary responsibility for the final content. All authors read and approved the final manuscript. Funding This study was funded by the National Social Science Fund of China in 2021 (No. 21BGL235). Data availability The data are available in a public, open access repository and can be freely downloaded from the official website (https://charls.pku.edu.cn/) of the CHARLS. Declarations Competing interests The authors declare no competing interests. Ethics approval and consent to participate The Biomedical Ethics Review Committee of Peking University has given ethics approval to data collection, and all participants provided written informed consent (IRB00001052–11015). As this study utilized anonymized, publicly available data from these sources, no additional ethical approval or informed consent was required. The CHARLS was conducted in accordance with the ethical principles of the Declaration of Helsinki. Consent for publication Written informed consent for publication was obtained from all participants. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Yudong Wei and Lu Li contributed equally to this work. Contributor Information Yudong Wei, Email: ydwei@cdc.zj.cn. Lu Li, Email: lilu@zju.edu.cn. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement The data are available in a public, open access repository and can be freely downloaded from the official website (https://charls.pku.edu.cn/) of the CHARLS.