Parafoveal processing plays an important role in reading, the information obtained from parafovea is used to begin pre-processing upcoming words and to guide where to move the eyes next. Evidences have showed that parafoveal processing makes an important contribution to skilled and more effective reading. Older adults read more slowly than young adults and also show reduced peripheral visual processing in non-reading tasks. This raises the possibility that visual declines in later life limit older adults’ parafoveal processing. Consistent with this view, studies found that older adults obtain less rightward parafoveal information compared to young adults. Similarly, other studies using the boundary paradigm suggest eye movements are disrupted more for young than older adults when rightward parafoveal information is denied, consistent with older adults processing parafoveal information less effectively. However, several studies provide conflicting findings showing no such age differences, especially about the word N+2 processing. Accordingly, we conducted an eye movement experiment to assess parafoveal preview benefits for word N+2 in Chinese younger and older adults during sentence reading. 40 older adults (aged 66.23±1.83 years) from a community and 40 undergraduates(aged 20.13±1.18years) from a university participated in our experiment. Two groups of participants were asked to read 60 sentences which were presented using the boundary paradigm, with an invisible boundary placed after word N (two-character word), followed by two single-character words (word N+1 and word N+2). Prior to a reader making a saccade that crossed the boundary, word N+2 were shown normally (identical previews) or as invalid previews replacing with a visually similar pseudo-character. The sentences were therefore shown in one of two preview conditions, which reverted to normal as soon as a saccade crossed the boundary. We analyzed the data for sentence and three word regions (including word N, word N+1 and N+2) using Linear Mixed-Effects Models. The results showed that older adults read more slowly compared with young adults, by having more and longer fixations, making more regressions and shorter forward saccades. Moreover, the delayed parafoveal-on-foveal effect occurred for both young and old adults, showing that participants spent shorter fixation time on word N+1 under the condition of identical preview of word N+2 compared to pseudo-character preview, and skipped word N+1 more frequently. However, there was no age difference on the size of effects. Most important, we found age differences on the preview benefit of word N+2, due to an word N+2 preview effect for young adults but not for older adults. Consequently, the preview benefit from word N+2 in parafovea for older adults showed up when they processed the word N+1 but not word N+2. Then we reanalyzed the data for word N+2 when word N+1 was skipped and fixed respectively, and found that younger adults showed robust preview benefit regardless of word N+1 was fixed or not, but older adults did only when word N+1 was skipped. In summary, both older and younger adults can process up to two words parafoveally, however, older adults have difficulty in using the information obtained from word N+2 in parafovea. These findings shed light on revealing age-related reading difficulty in Chinese, and indicated that older readers’ parafoveal processing is impaired which might result in inefficiency word processing as well as the saccade targeting.