The marginal reader of corporate disclosure is increasingly an algorithm rather than a human, yet a report that is visually polished for human readers may be technically inaccessible to the machines that now screen, rank, and trade on its contents. We provide a novel measurement of machine readability in an African frontier market, analysing 1,110 annual reports of 74 non-financial firms listed on the Nigerian Exchange over 2010 to 2024. Using two independent open-source parsing libraries, we classify each filing into three machine readability tiers and construct a continuous extraction-rate measure, validated through inter-parser agreement and blinded manual review. We document substantial heterogeneity: a meaningful share of filings remain image-based and effectively invisible to algorithmic investors, although the proportion of highly machine-readable reports rose from 41 percent in 2010 to 76 percent in 2024. To explain this variation, we develop the Disclosure Production Capacity framework, which treats machine readability as the output of a production process driven by external monitoring pressure and internal capacity. Consistent with the framework, firm size, Big-Four audit engagement, and foreign institutional ownership are robustly and positively associated with machine readability across ordered logit, OLS, and instrumental-variables specifications, and survive an independent parser-disagreement measure, dynamic controls, and a structural break. The findings show that the firms able to reach the machine-readable frontier are those facing globalised monitoring and possessing the capacity to respond. We discuss implications for structured-reporting mandates, cross-border capital access, and the selection bias embedded in algorithmic screening of frontier markets.