Previous research on cognates, false cognates (FCs), and non-cognate translation equivalents (NCTEs) has largely centered on proficient bilinguals, emphasizing orthographic similarities between the first language (L1) and second language (L2) while often neglecting phonological factors. In contrast, the current study investigated intermediate learners’ recognition of the correct L1 meanings of L2 cognates, FCs, and NCTEs in two genetically related languages with distinct orthographies. A test-retest design was implemented with 23 Arabic-speaking high school students learning Hebrew (L2). Participants were presented with 32 bolded words embedded in Hebrew sentences and asked to choose the correct Arabic translation from four options, with linguistic features influencing responses carefully controlled. The task was first administered in the 11th grade (2019) and then again in the 12th grade (2020). The words selected for the translation task were tailored to match the learners’ proficiency level, guided by insights from a pilot study involving a larger pool of words and participants. The results indicated that accuracy in recognizing the correct translations of cognates was significantly higher than for FCs and NCTEs at both time points. In the 11th grade, students’ recognition of FCs was significantly lower than that of NCTEs. However, as students progressed in their L2 learning, they became less confused by misleading phonological similarities, and no significant difference in performance was found between FCs and NCTEs, suggesting that with increased L2 learning, the misleading effect of FCs (i.e., interlingual homophone inhibition) decreased. In addition, significant correlations between cognates, FCs, and NCTEs emerged only in the 12th grade.