Inside Capitol Hill's AI Toolkit: ChatGPT Purchases Outpace Every Rival Combined
News Summary
A new analysis of U.S. House of Representatives expense records shows that OpenAI's ChatGPT has become the overwhelmingly dominant artificial intelligence tool purchased by congressional offices, accounting for roughly 88% of identifiable AI spending and 96% of transactions during the year ending March 31, 2026, Eastern Time. The finding offers a rare, itemized look at how one of the world's largest legislative bodies is adopting generative AI tools for day-to-day staff work, even as broader questions about AI procurement, security, and oversight continue to be studied across government.
How the Numbers Break Down
Between April 2025 and March 2026, House offices spent at least $113,740 on named AI vendors, according to the review of disbursement records. Of that total, ChatGPT-related purchases accounted for about $100,580 across 798 separate transactions. Anthropic's Claude, the next most common tool, trailed far behind with $13,160 spent across 37 transactions. Other AI products appeared in the data only sporadically, underscoring how concentrated early institutional adoption has been around a single platform.
Researchers who reviewed the disbursement data caution that the figures likely undercount actual AI usage on Capitol Hill. The totals exclude free-tier accounts, AI features bundled into existing software subscriptions such as Microsoft's productivity suite, and most Senate spending, which is tracked through a separate reporting system. As a result, the real footprint of generative AI inside Congress is almost certainly larger than the itemized purchase records suggest.
What Staff Are Using AI For
Congressional staff have described using chatbot tools to summarize lengthy legislative text, draft internal memos, prepare materials for committee hearings, analyze proposed bills, and help respond to the large volume of constituent correspondence that flows into every office. These tasks are traditionally time-intensive for small legislative teams, and staffers say AI assistance has made it easier to keep pace with the volume of information moving through the institution at any given time.
At least 70 House offices, roughly one in six, used identifiable AI tools during the period studied, according to the analysis. That adoption rate illustrates how quickly generative AI has moved from an experimental curiosity to a standard part of the software toolkit available to legislative staff, mirroring similar adoption trends seen in other large organizations and public-sector agencies.
Access and Platform Differences
Access to AI tools varies by chamber. According to internal guidance, Senate offices currently permit official use of a specific set of approved tools, while Anthropic's Claude is not authorized for official use in that chamber. The House, by contrast, has seen broader experimentation across multiple vendors, even though ChatGPT remains the clear preference among the tools that show up in itemized spending records.
Earlier in the year, the General Services Administration, which manages procurement across federal agencies, announced an agreement making ChatGPT available government-wide at a steeply discounted rate through existing acquisition contracts. That arrangement is separate from the itemized congressional office spending detailed in this analysis but reflects a broader push to make commercial AI tools more accessible to federal employees, including through low-cost or no-cost onboarding programs designed to encourage adoption while agencies develop longer-term usage guidelines.
A Broader Technology Race
The concentration of spending around one platform comes as multiple AI developers, including OpenAI, Anthropic, Google, and Microsoft, continue to refine offerings aimed at government and enterprise customers. Each company has built specialized versions of its models with additional security, data-handling, and compliance features intended for institutional use. As agencies and legislative offices gain more experience with these tools, procurement patterns are expected to evolve, and analysts expect the current gap between leading and secondary providers to narrow as more offices formalize their AI purchasing decisions and evaluation processes.
For now, the disbursement data offers one of the clearest publicly available snapshots of how a major legislative institution is integrating generative AI into its daily operations, providing a useful case study for researchers and technologists studying institutional AI adoption more broadly.