**Background:** Autosomal recessive cerebellar ataxias (ARCAs) are a heterogeneous group of neurodegenerative disorders primarily affecting the cerebellum and its afferent tracts, often accompanied by damage to other neurological or extra-neurological systems. Diagnosis remains challenging, especially in adult-onset sporadic cases, despite advances in molecular characterization. The examination of the afferent (retinal nerve fiber layer, ganglion cells) and efferent (fixation, saccades, pursuit, VOR) visual systems may provide relevant structural and neurophysiological findings. This review suggests using robust eye findings as markers for parametrization of ARCAs, proposes a diagnostic classification based on ocular features, and discusses quantification of ganglion cells (GC) and retinal nerve fiber layer (RNFL) thickness as markers of neurodegeneration, as well as the role of artificial intelligence (AI) in analyzing ocular parameters.
**Methods:** This is a narrative review summarizing the literature on ocular findings in ARCAs. The authors categorize ARCAs into three common pathological pathways: mitochondrial metabolism, DNA repair/genome stability, and complex lipid metabolism. For each pathway, they describe the associated genetic disorders, their underlying molecular mechanisms, and the characteristic oculomotor and ophthalmoscopic findings. They also review studies using optical coherence tomography (OCT) to measure RNFL and GC thickness in ARCAs, particularly in Friedreich's ataxia (FRDA), spastic paraplegia 7 (SPG7), and autosomal recessive spastic ataxia of Charlevoix-Saguenay (ARSACS). The potential of AI and machine learning in analyzing eye movement data and OCT images is discussed.
**Key Results:** The review identifies distinct ocular patterns for each pathway:
- **Mitochondrial pathway** (FRDA, POLG, COQ8A, SPG7, SPAX5, ARSACS): FRDA shows fixation instability with square wave jerks (SWJs), gaze-evoked nystagmus, impaired smooth pursuit, increased saccadic latency, and saccadic dysmetria. Optic nerve pallor is present in 30% of FRDA patients at onset. OCT reveals decreased average RNFL thickness, with predominant involvement of the superior quadrant. In SPG7, adult-onset progressive external ophthalmoplegia (CPEO) and ptosis are common, and OCT shows global RNFL thinning, more pronounced in the temporal quadrant. ARSACS is characterized by increased peripapillary inner retinal thickness, foveal hypoplasia in 100% of patients, macular microcysts, and papillomacular folds.
- **DNA repair/genome stability** (AT, ATLD, AOA1, AOA2, CANVAS): These disorders share oculomotor apraxia (OMA), long-latency multistep saccades, and oculocephalic dissociation. AT and ATLD show saccadic intrusions, spontaneous nystagmus, and impaired VOR. AOA1 and AOA2 exhibit hypometric saccades with a staircase pattern, excessive blinking, and fixation instability. CANVAS features cerebellar nystagmus, impaired VOR, and oscillopsia due to bilateral vestibular impairment.
- **Complex lipid metabolism** (NPC, CTX, AVED): NPC is characterized by vertical supranuclear gaze palsy with progressive slowing of vertical saccades. CTX presents with juvenile cataracts, gaze-evoked nystagmus, and increased saccadic latency and directional errors. AVED shows retinitis pigmentosa, nystagmus, and oculomotor apraxia.
**Clinical Implications:** Ocular features can serve as biomarkers for disease assessment in ARCAs. Eye movement quantification helps localize cerebellar-extracerebellar network involvement, while OCT provides a measure of neurodegeneration, particularly in mitochondrial pathway-related forms. The proposed classification based on ocular findings may aid in differential diagnosis and guide genetic testing. The integration of AI and machine learning with oculomotor and OCT data could enhance diagnostic accuracy and provide quantitative outcome measures for clinical trials. Early identification of treatable forms (e.g., CTX, NPC, AVED) is crucial, as specific therapies can improve or prevent symptoms.