How AI is reshaping US government websites—and where it’s failing

President Donald Trump’s 2025 initiative to overhaul every federal .gov website using AI has exposed critical flaws in the government’s digital transformation strategy. The plan, launched under the banner of the National Design Studio (NDS), sought to replace outdated designs with a unified, AI-generated aesthetic within three years. Instead, the rollout has delivered mixed results, with many sites now displaying inconsistent branding, confusing navigation, and accessibility barriers that undermine usability.

A rushed mandate with unrealistic timelines

The NDS was established via executive order in August 2025, bypassing traditional procurement and agency oversight to centralize control under a temporary entity directly accountable to the White House. Its core mission was twofold: modernize the U.S. Web Design System (USWDS) and migrate 27,000 government websites to a new AI-driven framework by 2028. The urgency stemmed from complaints about outdated interfaces and the need for a cohesive digital identity across federal agencies.

However, the initiative’s ambitious scope collided with severe staffing and resource constraints. The Trump administration dismantled key government units such as 18F—a digital services agency—and restructured the U.S. Digital Service into DOGE, further reducing institutional expertise. With fewer personnel and no dedicated funding, the NDS relied heavily on automated tools, including generative AI, to generate designs without sufficient human oversight or iterative testing.

The flaws in AI-generated government websites

Critics say the AI-generated designs prioritize aesthetics over functionality, creating websites that look polished but function poorly. Common issues include:

• - Inconsistent branding: Agencies using the same AI templates produced divergent visual styles, violating federal design guidelines.
• - Poor navigation: AI-generated menus often lack logical hierarchy, burying critical services under convoluted pathways.
• - Accessibility gaps: Many sites failed to meet WCAG 2.1 standards, excluding users with disabilities due to missing alt text, low contrast, or unlabelled form fields.
• - Content duplication: AI repurposed boilerplate text without adapting it to agency-specific needs, leading to generic, uninformative pages.

One federal employee, who spoke on condition of anonymity, described the new designs as "visually striking but fundamentally broken." They noted that internal feedback loops were ignored in favor of meeting arbitrary deadlines. "The AI doesn’t understand context," they said. "It churns out templates, but it can’t replace human judgment—or common sense."

The human cost of automation in public service

The failures underscore a broader tension between speed and quality in government digital initiatives. The NDS’s reliance on AI reflects a broader trend in public sector tech, where cost-cutting and rapid deployment often trump long-term sustainability. Yet, the consequences extend beyond aesthetics: citizens depend on these sites for essential services like tax filings, disaster relief, and healthcare navigation.

For example, the redesigned Internal Revenue Service (IRS) portal, one of the high-profile launches under the initiative, reportedly saw a 20% increase in user complaints within weeks of its rollout. Users struggled to locate tax forms and payment tools, forcing many to revert to older, less secure versions of the site.

What’s next for government website design?

The NDS’s struggles raise questions about the viability of AI-driven government overhauls without robust oversight. While generative AI can accelerate design prototyping, it cannot replace the iterative refinement and domain expertise required for public-facing systems. Industry analysts suggest that future efforts must balance automation with human-centered design principles and invest in upskilling government teams.

The Biden administration, which has signaled skepticism toward DOGE’s initiatives, may revisit the NDS’s mandate after the 2026 elections. For now, millions of Americans navigating federal websites will continue to encounter a patchwork of outdated and experimental interfaces—each bearing the unintended consequences of AI’s growing role in public service.