•NCCN maintains that END should be considered for cT1-2N0M0 OCSCC with a DOI > 3 mm.•DOI > 2.5 mm and poor differentiation predict adverse outcomes in cT1N0M0 OCSCC.•The decision to perform END should be guided by DOI and tumor differentiation.•This strategy allows avoiding END in 48.6% of patients, without compromising outcomes. According to the NCCN guidelines, there is weak evidence to support the use of elective neck dissection (END) in early-stage oral cavity squamous cell carcinoma (OCSCC). We sought to examine the indications for END in patients with cT1N0M0 OCSCC defined according to the AJCC Staging Manual, Eight Edition. Of the 3886 patients diagnosed with cT1N0M0 included in the study, 2065 underwent END and 1821 neck observation. The 5-year outcomes for patients who received END versus neck observation before and after propensity score matching (n = 1406 each) were as follows: neck control, 96 %/90 % (before matching), p < 0.0001; 96 %/90 % (after matching), p < 0.0001; disease-specific survival (DSS), 93 %/92 % (before matching), p = 0.0227; 93 %/92 % (after matching), p = 0.1436. Multivariable analyses revealed that neck observation, depth of invasion (DOI) > 2.5 mm, and poor differentiation were independent risk factors for 5-year outcomes. Upon the application of a scoring system ranging from 0 (no risk factor) to 3 (presence of the three risk factors), the following 5-year rates were observed: neck control, 98 %/95 %/84 %/85 %; DSS, 96 %/93 %/88 %/85 %; and overall survival, 90 %/86 %/79 %/59 %, respectively (all p < 0.0001). The survival outcomes of patients with scores of 0 and 1 were similar. The occult metastasis rates in the entire study cohort, DOI > 2.5 mm, and poor differentiation were 6.8 %/9.2 %/17.1 %, respectively. Because all patients who received neck observation had a score of 1 or higher, END should be performed when a DOI > 2.5 mm or poorly differentiated tumors are present. Under these circumstances, 48.6 % (1888/3886) of cT1N0M0 patients may avoid END without compromising oncological outcomes.