This study presents a novel, discreet, and comfortable in-ear EEG device for long-term seizure detection. The device was tested against intracranial and scalp EEG in 20 patients with refractory focal epilepsy. The results showed that the in-ear EEG system detected 86.4% of seizures, with a false detection rate of 0.1 per day, which is significantly lower than that of ambulatory monitoring. The system was well tolerated by patients, with only minor adverse events reported. The in-ear EEG signals were found to be highly sensitive to temporal lobe seizures and provided complementary or superior signal quality compared to traditional scalp EEG. The study also demonstrated that the in-ear EEG system could detect seizures during sleep, which is often missed by conventional methods. The device showed high sensitivity for detecting temporal lobe seizures (86% for intracranial patients and 100% for scalp patients) and for seizures with secondary generalization. The system also captured all frontal lobe seizures and 100% of analyzed seizures with secondary generalization. The timing differences between seizure annotations from the in-ear EEG and gold standard EEG suggest that the signal characteristics used by epileptologists to identify seizure transitions vary depending on the recording modality. The in-ear EEG device offers a promising alternative to traditional EEG methodologies, with a non-intrusive design and reliable neurophysiological signals for seizure detection during both wakefulness and sleep. The study highlights the potential of using in-ear EEG for routine collection of complementary, prolonged, and remote neurophysiological evidence, which may assist clinicians in making an epilepsy diagnosis, assessing treatment efficacy, and optimizing medication titration. The device's potential for widespread adoption is underscored by its high patient tolerance and minimal adverse events. However, further research is needed to assess the device's performance in real-world settings and to refine its features for user-centric design.This study presents a novel, discreet, and comfortable in-ear EEG device for long-term seizure detection. The device was tested against intracranial and scalp EEG in 20 patients with refractory focal epilepsy. The results showed that the in-ear EEG system detected 86.4% of seizures, with a false detection rate of 0.1 per day, which is significantly lower than that of ambulatory monitoring. The system was well tolerated by patients, with only minor adverse events reported. The in-ear EEG signals were found to be highly sensitive to temporal lobe seizures and provided complementary or superior signal quality compared to traditional scalp EEG. The study also demonstrated that the in-ear EEG system could detect seizures during sleep, which is often missed by conventional methods. The device showed high sensitivity for detecting temporal lobe seizures (86% for intracranial patients and 100% for scalp patients) and for seizures with secondary generalization. The system also captured all frontal lobe seizures and 100% of analyzed seizures with secondary generalization. The timing differences between seizure annotations from the in-ear EEG and gold standard EEG suggest that the signal characteristics used by epileptologists to identify seizure transitions vary depending on the recording modality. The in-ear EEG device offers a promising alternative to traditional EEG methodologies, with a non-intrusive design and reliable neurophysiological signals for seizure detection during both wakefulness and sleep. The study highlights the potential of using in-ear EEG for routine collection of complementary, prolonged, and remote neurophysiological evidence, which may assist clinicians in making an epilepsy diagnosis, assessing treatment efficacy, and optimizing medication titration. The device's potential for widespread adoption is underscored by its high patient tolerance and minimal adverse events. However, further research is needed to assess the device's performance in real-world settings and to refine its features for user-centric design.