Show HN: GhostBench – do AI models recommend dead SaaS? I tested my own thesis (github.com)

GhostBench, a new verification convention for SaaS vendors, has been introduced to address the challenge of identifying "ghost" companies—those that are no longer operational but still appear viable online. Utilizing machine-verifiable facts published in a standardized format, GhostBench enables AI agents to determine a company's aliveness by analyzing its operational history. A recent test revealed that while 19 out of 20 inactive SaaS products still returned HTTP 200 status codes, a leading AI model effectively flagged all dead sites without additional tools. Notably, a smaller, faster model misidentified some deceased vendors as operational due to its reliance on limited verification metrics. This initiative is significant for the AI/ML community as it enhances the reliability of vendor selection processes within AI applications. The convention establishes a hierarchy of verifiable facts—ranging from basic domain verification to comprehensive continuity logs—thereby promoting transparency and trust in the information that AI systems utilize. By addressing the complexities of maintaining accurate operational data, GhostBench lays the groundwork for more informed decision-making in AI-driven environments, ensuring models are trained on valid, up-to-date information rather than outdated or misleading signals.