OBJECTIVE: This preliminary study tested whether non-invasive, remote Elastic Scattering Spectroscopy (ESS) measurements obtained in the oral cavity can be used as a proxy to accurately differentiate between patients with laryngeal cancer versus laryngeal leukoplakia. METHODS: 20 patients with laryngeal lesions [cancer (n = 10),leukoplakia (n = 10)] were clinically assessed and categorized by otolaryngologists per standard clinical practice. Patient demographics of age, race, sex, and smoking history were collected. A machine-learning artificial intelligence (AI) algorithm was applied to classify patients using ESS spectra of patients with benign laryngeal leukoplakia or laryngeal cancer. Specificity, sensitivity, positive predictive value (PPV), negative predictive value (NPV), F1, and area-under-the-curve (AUC) were calculated. Additional algorithms stratified spectral data by sub-anatomical site and smoking status to explore diagnostic capability. RESULTS: Overall, the algorithm had a sensitivity = 74 %, specificity = 40 %, PPV = 51 %, NPV = 64 %, F1 = 0.61 and AUC = 0.65. When stratifying by former and active smokers, algorithm sensitivities increased to 85 % and 77 %. Analysis by sub-anatomic location yielded an AUC = 0.77 for lateral tongue, and when stratified by (former/current) smoking status, demonstrated AUC = 0.94 and 0.83, sensitivities = 98 % and 76 %, and specificities = 85 % and 86 %. Algorithm output from the mucosal lip yielded sensitivity = 89 %, specificity = 88 %, PPV = 83 %, and NPV = 92 % in former smokers. CONCLUSION: This pilot study demonstrated ESS technology coupled with AI-assisted statistical modeling, could differentiate between patients with laryngeal leukoplakia versus cancer with good precision, especially with smoking status and anatomic subclassification. If ESS can be utilized in the oral cavity as a non-invasive screening tool for laryngeal cancer, it would greatly facilitate early detection in specialized/non-specialized clinics, and under-resourced regions.