Modern web applications rely heavily on client-side API calls to fetch data, render content, and communicate with backend services. However, the quality of these network interactions (redundant requests, missing cache headers, oversized payloads, and excessive third-party dependencies) is rarely tested in a systematic way. Moreover, many of these quality deficiencies carry security implications: missing cache headers enable cache poisoning, excessive third-party dependencies expand the supply-chain attack surface, and error responses risk leaking server internals. In this study, we present an automated software testing framework that captures and analyzes the complete HTTP traffic of 18 production websites spanning 11 categories (e-commerce, news, government, developer tools, travel, and more). Using automated browser instrumentation via Playwright, we record 108 HAR (HTTP Archive) files across 3 independent runs per page, then apply 8 heuristic-based anti-pattern detectors to produce a composite quality score (0-100) for each site. Our results reveal a wide quality spectrum: minimalist server-rendered sites achieve perfect scores of 100, while content-heavy commercial sites score as low as 56.8. We identify redundant API calls and missing cache headers as the two most pervasive anti-patterns, each affecting 67% of sites, while third-party overhead exceeds 20% on 72% of sites. One utility site makes 2,684 requests per page load, which is 447x more than the most minimal site. To protect site reputations, all identities are anonymized using category-based pseudonyms. We provide all analysis scripts, anonymized results, and reproducibility instructions as an open artifact. This work establishes an empirical baseline for HTTP API call quality across the modern web and offers a reproducible testing framework that researchers and practitioners can apply to their own applications.

2,684 requests on a single page load is wild. Would be interesting to see how much of that is just analytics and tracking scripts piling up on each other.

The redundant API call detection is something I wish more dev tools had built in. I've definitely shipped things where the same endpoint was getting hit multiple times on mount because of re-renders, and you don't notice until you actually look at the network tab.