California Community Colleges Waste Millions on Unreliable AI Chatbots
Key Takeaways
- ▸California community college districts have signed multi-year contracts totaling millions—the LA district alone committed $3.8M through 2029—for AI chatbots that frequently fail to provide accurate answers
- ▸Chatbot platforms relying on manually maintained FAQ databases and static websites generate outdated or incorrect information, forcing students to seek help through unofficial social media channels and Reddit
- ▸Testing revealed significant reliability issues, including inability to answer institution-specific questions correctly (e.g., East LA College's chatbot couldn't name its own president)
Summary
California's community college districts are spending millions of dollars annually on AI-powered chatbots—ranging from $151,000 to $500,000 per district per year—to assist students with admissions, financial aid, and campus services. However, investigative reporting by The Markup and CalMatters reveals these systems struggle to provide accurate, consistent answers, with some chatbots unable to answer basic questions or even correctly identify their own institution's leadership. The Los Angeles Community College District alone has approved $3.8 million in contracts through 2029 for chatbot platforms like Gravyty (formerly Ocelot) and Gecko, which rely on manually maintained FAQ databases and outdated campus websites that lead to frequent errors.
Students frustrated by unreliable answers are turning to unofficial channels—Reddit, Google searches, and social media—to find information they should be getting from these expensive systems. Even basic questions often result in outdated or incomplete responses. Some districts are beginning to pivot toward ChatGPT-integrated systems that scrape live website content, with officials suggesting these approaches appear more reliable. The widespread adoption of underperforming chatbots highlights a broader trend of educational institutions investing heavily in AI solutions that don't meet their fundamental requirements.
- Some districts are transitioning to newer ChatGPT-integrated systems that dynamically scrape college websites, suggesting the current generation of vendor platforms is fundamentally flawed
Editorial Opinion
This story exposes a painful gap between the hype around generative AI and the reality of vendor implementations in critical services. Community colleges—serving students who often juggle work, family, and education—deserve tools that actually work, yet these expensive systems are actively frustrating their users and wasting institutional resources. The pivot toward ChatGPT represents an indictment of the first-generation vendors, and it suggests that off-the-shelf LLMs fine-tuned to live data perform better than custom-built, static-data systems. Educational leaders should demand accountability: either vendors deliver systems that match student needs, or districts redirect these millions toward human support.
