In this work, we explore the differences between proctored and unproctored Internet administration for a Basque language low-stakes test considering demographic factors such as age, gender, and knowledge level in the subject. To this aim, we have developed an ad hoc application that allows us to establish a set of filters and techniques that successfully control dropout and non-serious test takers, two of the main threats of low-stakes testing. A total of 2,095 sessions has been registered. The obtained results show that age and knowledge level influences the execution of the test, whereas gender does not. Moreover, it is made explicit that conducting the test in an unproctored manner redounds to the achievement of better results. Finally, even if the time needed to complete the test is comparable in both cases, it is better invested in the unproctored version, i.e. less time is devoted to easy questions and more time is devoted to difficult questions if compared to the proctored version. These results seem to indicate that the unproctored version measures better the knowledge level for language low-stakes tests because they are carried out in an environment that is familiar to the examinee, and they lack the pressure of proctored testing.