This study examines the impact of explicit and implicit pronunciation instruction on the segmental pronunciation accuracy of phonemes /s/ and /z/ in an EFL classroom of 11th-grade Spanish-speaking students. The research focuses on transparent and non-transparent words containing the grapheme <s> and the phonemes /s/ and /z/ and was conducted in a Chilean rural vocational high school over four weeks, using a quasi-experimental one-group repeated-measures design with convenience sampling. Four students participated in four 90-minute sessions with two different conditions: two sessions for the implicit instruction condition and two sessions for the explicit instruction condition, each followed by a segmental pronunciation accuracy post-test. The results showed that students consistently performed well with transparent words, benefiting from clear grapheme-phoneme correspondences aligned with their L1 orthography in both conditions. In contrast, non-transparent words posed greater challenges across conditions, indicating that the type of instruction does not play a role when it comes to grapheme-phoneme incongruencies. These findings imply that transparent words serve as effective foundational tools for pronunciation, while non-transparent words require tailored teaching approaches to address grapheme-phoneme incongruencies. The limitations and recommendations for pronunciation instruction of transparent and non-transparent words are also discussed in this article.