The Trump administration Tuesday launched America.gov, an AI chatbot intended to become the single starting point for Americans looking for federal services, FedScoop reported. The site uses Google's Gemini and xAI's Grok models to answer questions drawn from roughly 29,000 government websites.
What Happened
In its current form, America.gov answers questions and directs users to the right service, for tasks such as getting a passport, replacing a Social Security card or booking a campsite at a national park. FedScoop reported that the system draws on content from about 29,000 federal sites and picks up updates as soon as they are published.
The administration plans to go further in 2027, letting people submit documents without leaving the platform and track the progress of applications. Officials used a name change after marriage as an example of a process that would run through the site.
President Trump signed an executive order directing the General Services Administration, the National Design Studio and the Office of Management and Budget to make America.gov "the single point of entry" to federal services, with Login.gov handling identity verification. TechCrunch reported that Trump described the site as a front door that replaces a maze of tens of thousands of government websites.
U.S. Chief Design Officer Joe Gebbia, a co-founder of Airbnb, led the design, and DOGE member Edward Coristine is the lead engineer. Peter Arnell is chief brand architect, and FedScoop said Elon Musk was credited with the initial vision.
Why It Matters
The site moves generative AI from pilot projects into the main public entrance for federal government, and it does so with commercial models from two companies. Google confirmed its involvement with Gemini, according to TechCrunch. The use of Grok links the project to xAI, the AI company founded by Musk, who led the DOGE effort earlier in the administration.
The privacy notice offers some detail. It says the AI providers do not retain prompts or responses, and that America.gov caches answers for up to two hours using hashed versions of prompts rather than the text itself. Users are advised to share only what is needed. Beyond that, the White House declined to describe procurement, contract terms, the technology agreements with Google and xAI, or specific privacy protections.
The Accuracy Problem
Large language models still produce confident but incorrect answers. TechCrunch highlighted the risk for people using the chatbot to navigate food assistance, visa renewals or tax filing, where a wrong instruction could lead to a missed deadline, a denied benefit or a penalty. Today that risk is limited because the site mostly routes people to existing agency pages. The 2027 plan to accept documents and track applications would put the chatbot directly inside transactions that carry legal and financial consequences.
Integrating Login.gov adds a separate set of questions. Tying identity verification to an AI interface means the site will handle more personal information over time, which makes the undisclosed contract and data-handling terms more significant.
What to Watch
Key questions are what the government is paying Google and xAI, how the two models divide the work, and whether independent testing or an inspector general review checks the accuracy of answers on high-stakes topics. Agencies will also have to show how their own services connect to America.gov under the executive order before the planned 2027 transactional features arrive.
TECHNOLOGY · AI · CYBERSECURITY · THE SYSTEMS SHAPING BUSINESS