Reverse transcription quantitative real-time PCR (RT-qPCR), a sensitive technique for quantifying gene expression, depends on the stability of the reference gene(s) used for data normalization. Only a few studies on reference genes have been done in fruit trees and none in litchi. In the present study, seven frequently used candidate reference genes, including actin (ACTIN), glyceraldehyde-3-phosphate-dehydrogenase (GADPH), elongation factor 1-alpha (EF-1α), poly ubiquitin enzyme (UBQ), α-tubulin (TUA), β-tubulin (TUB) and RNA polymerase-II transcription factor (RPII), were evaluated for their expression stability in litchi. A total of 78 samples, including different varieties, tissues, organs, developmental stages and treatments, such as NAA, shading and girdling plus defoliation, were addressed in this analysis. Our results showed that GAPDH was the most suitable reference gene among all the tested samples, different organs and NAA treatment. ACTIN was stably expressed in varieties and fruit developmental stages. RPII and UBQ exhibited better expression stability in tissues. EF-1α was the most stable gene in shading and girdling plus defoliation treatments. Moreover, using combination of two genes as reference genes might improve the reliability of gene expression by RT-qPCR in litchi. A better combination was GAPDH + EF-1α or GAPDH + ACTIN for all the examined samples. In addition, the validated reference genes were further relied on to quantify the expression of an interested gene, LcARF13 under different experimental conditions. These results first provide guidelines for reference genes selection under different experimental conditions and also a foundation for more accurate and widespread use of RT-qPCR in litchi.