<b><i>Objectives:</i></b> This study aimed at testing whether CSF levels of amyloid β<sub>42</sub> (Aβ<sub>42</sub>), Aβ<sub>40</sub>, total tau, and phosphorylated tau (P-tau<sub>181</sub>) individually contribute to the identification of atypical phenotypes among a retrospective cohort of probable Alzheimer’s disease (AD) patients diagnosed by means of the ratio between Aβ<sub>42</sub> and Aβ<sub>40</sub> (Aβ<sub>42/40</sub>). <b><i>Methods:</i></b> The retrospective study cohort comprised 50 probable AD patients diagnosed by means of the ratio between Aβ<sub>42</sub> and Aβ<sub>40</sub> (Aβ<sub>42/40</sub>) and for whom total tau and P-tau<sub>181</sub> values were also available. Patients were clinically classified as typical, amnestic-predominant AD (<i>N</i> = 39; 16 males; mean age: 73.4 ± 7.6 years; mean disease duration: 27.4 ± 24.7 months), or atypical phenotypes (<i>N</i> = 11; 6 males; mean age: 70.2 ± 6.5 years; mean disease duration: 35.5 ± 24.9 months) – i.e., posterior cortical atrophy (<i>N</i> = 4), logopenic-variant primary progressive aphasia (<i>N</i> = 4), and behavioural variant AD (<i>N</i> = 3). A logistic regression allowed predicting the occurrence of atypical versus typical phenotypes based on age, sex, and Aβ<sub>42</sub>, Aβ<sub>40</sub>, total tau, and P-tau<sub>181</sub> levels. <b><i>Results:</i></b> Atypical and typical AD patients were comparable for Aβ<sub>42/40</sub> values. Only Aβ<sub>40</sub> and P-tau<sub>181</sub> levels positively (<i>p</i> = 0.015) and negatively (<i>p</i> = 0.019) predicted the occurrence of atypical AD phenotypes, respectively. Classification precision was of 86%, yielding excellent specificity (94.9%) but poor sensitivity (54.5%). <b><i>Conclusions:</i></b> The present study delivers promising, albeit preliminary, evidence on the utility of Aβ<sub>40</sub> and P-tau<sub>181</sub> CSF biomarkers in differentiating atypical from typical Aβ<sub>42/40</sub>-confirmed AD phenotypes, prompting further research and confirmation on larger cohorts.