The diagnostic accuracy of contrast-enhanced CT for detection of cervical lymph node metastasis in dogs is unknown. The purpose of this retrospective, observational, diagnostic accuracy study was to assess the efficacy of CT for detection of mandibular and medial retropharyngeal lymph node metastasis in dogs. Histopathology of dogs with cancer of the head, CT and bilateral mandibular and medial retropharyngeal lymphadenectomy was reviewed. A single radiologist measured lymph nodes to derive short axis width and long-short axis ratios. Two blinded radiologists separately assessed lymph node margins, attenuation and contrast enhancement and each provided a final subjective interpretation of each node site as benign or neoplastic. Where radiologists' opinions differed, a consensus was reached. Sensitivity, specificity and accuracy were calculated for mandibular and medial retropharyngeal sites. Agreement between radiologists was assessed. Fisher's exact test and the Kruskal-Wallis H-test were used to assess associations between variables. Forty-one primary tumours were recorded in 40 dogs. Metastasis to mandibular or retropharyngeal lymph nodes occurred in 16 out of 40 dogs (43/160 nodes). Agreement between radiologists was almost perfect for margination, attenuation and enhancement, strong for interpretation of mandibular lymph node metastasis, and weak for interpretation of medial retropharyngeal lymph node metastasis. Sensitivity of CT was 12.5% and 10.5%, specificity was 91.1% and 96.7%, and accuracy was 67.5% and 76.3% for mandibular and medial retropharyngeal lymph nodes respectively. No individual CT findings were predictive of nodal metastasis. Given the low sensitivity of CT, this modality cannot be relied upon alone for assessment of cervical lymph node metastasis in dogs.