Background: There are 9.9 million new cases of dementia in the world every year. Short-term conversion rate from mild cognitive impairment (MCI) to dementia is between 20% and 40%, but long-term in 5–10 years ranges from 60% to 100%. It is particularly important to prevent or prolong the development of MCI into dementia. Both auriculotherapy and vagus nerve stimulation are effective on improving cognitive functions. However, there is no double blinded randomized clinical trial to support the effectiveness of transcutaneous electrical stimulation of auricular acupoints in patients with MCI. Methods: This randomized controlled trial involved patients with MCI, aged from 55 to 75 years old. Patients were randomly allocated to transcutaneous auricular vagus nerve stimulation (taVNS) group or sham taVNS group. In the taVNS group, two auricular acupoints were stimulated, including heart (concha, CO15) and kidney (CO10), which are in the distribution of vagus nerve. While in the sham taVNS group, two other auricular acupoints were stimulated, including elbow (scaphoid fossa, SF3) and shoulder (SF4,5), which are out of the distribution of vagus nerve. The primary outcome was the Montreal cognitive assessment-basic, MOCA-B. The secondary outcomes included auditory verbal learning test-HuaShan version (AVLT-H), shape trails test A&B (STT-A&B), animal fluence test (AFT), Boston naming test (BNT), Pittsburgh sleep quality index (PSQI), rapid eye movement sleep behavior disorder screening questionnaire (RBDSQ), Epworth sleepiness scale (ESS) and functional activities questionnaire (FAQ). These outcome measures were taken at baseline, 24 weeks later. Results: After 24 weeks of intervention, the data of 52 patients were intended for analysis. After intervention, there was significant difference in the overall scores of MoCA-B between taVNS group and sham taVNS group (p = 0.033