Objectives To examine the diagnostic utility of canine cerebrospinal fluid samples collected into tubes containing ethylenediaminetetraacetic acid (EDTA) with and without the addition of 10% buffered formalin analysed within 6 to 20 hours. Materials and Methods Inclusion criteria were dogs presenting to a referral hospital with neurological signs and having cerebrospinal fluid analysis performed. Samples were submitted to an external laboratory in tubes containing ethylenediaminetetraacetic acid as paired‐samples; one with the addition of one drop of 10% buffered formalin and the other without formalin. Cytology report, total nucleated cell counts, and protein concentration were reviewed. Three different categories of cell preservation were defined: diagnostic, non‐diagnostic and unclassified. Each sample was included in one of these categories depending on cytological features, and the diagnostic quality between samples was compared. Samples were further divided in two groups depending on protein concentration using 50 mg/dL as cut‐off value and the diagnostic quality between samples was compared. Results 254 samples from 127 dogs were included. 47% of samples without formalin were non‐diagnostic, 46% diagnostic and 6% unclassified. In the formalin group, 2% samples were non‐diagnostic, 92% diagnostic and 6% unclassified. Samples with formalin preservation were statistically more likely to be diagnostic than samples without formalin preservation. In both protein groups (≥50 and <50 mg/dL) formalin samples were statistically more likely to be diagnostic as well. Clinical Significance The addition of one drop of 10% formalin is a simple, widely available method which can help to improve the accuracy of cytological assessment in canine cerebrospinal fluid by preserving cellular morphology when analysis is performed within 6 to 20 hours.