Little is known about the impact of equine-assisted interventions (EAI) on equids’ perception of humans. In this study 172 equids, living in 12 riding centres, were submitted to a standardised human-horse relationship test: the motionless person test. Age, sex, type (horse/pony), housing, and feeding conditions of subjects were recorded. Overall, 17 equids worked in EAI, 95 in riding school lessons (RS), and 60 in both (EAI-RS). There were high inter-individual variations in the number of interactive behaviours directed towards the experimenter: negative binomial general linear models showed that activity was the most important factor: RS equids performed more interactive behaviours than EAI (p = 0.039) and EAI-RS (p < 0.001) equids. Daily quantity of hay appeared as the second most important factor (equids with more than 3 kg interacted more than equids with less than 3 kg, p = 0.013). Individual characteristics were also important as horses interacted more than ponies (p = 0.009), geldings more than mares (p = 0.032), and 3-15-year-old equids more than equids over 15 years (p = 0.032). However, there was no interaction between factors. The lower number of interactive behaviours of EAI equids leads to different hypotheses-namely, selection on temperament, specific training, or compromised welfare (apathy). In any case, our results raised new lines of questions on EAI.