nga.gov
success
Claude Code
intent

What does nga.gov do and who is it for? Explain it back to me.

13steps
70.4sduration
$0.9731cost
165,915tokens
  • The site refused the agent's first request with HTTP 403
21 steps8 reasoning steps7 searches
home
docs
/careers
search
search
/about-us
/collection.html
/pressroom.html
search
search
search
search
search
83%
on-site discovery
0%
reliability
17%
link following
path origin
  • previous resource17%
  • web search17%
  • prior knowledge67%
insight

The agent successfully understood nga.gov as the National Gallery of Art by working around a Cloudflare-blocked website. All direct fetches to nga.gov returned 403 errors, forcing the agent to rely entirely on web search results and prior knowledge to assemble a comprehensive answer about the museum's mission, free admission model, audience, and competitive positioning. The site itself was not machine-readable to the agent, but search engines and cached/indexed content provided sufficient fragments to construct a coherent explanation.

  • ›All seven direct fetch attempts to nga.gov URLs returned Cloudflare 403 blocks with no usable HTML content, making the site completely inaccessible to agent crawling.
  • ›The agent discovered the correct identity of nga.gov (National Gallery of Art, not National Geospatial-Intelligence Agency) through web search results [5][7], which surfaced the official homepage and Wikipedia entries in snippet form.
  • ›Web search results [13][14][16][17][19] provided concrete details about admission (free), collection size (160,000+ works), membership tiers ($75–$20,000+), comparative positioning vs. Smithsonian museums, and specific programs (Jazz in the Garden, Sculpture Garden). None of these came from direct site access.
  • ›The agent cited search result URLs in its final Sources section ([13][14][16][17][19]), treating search snippets and linked pages as authoritative even though it could not fetch them directly.
  • ›Prior knowledge contributed 67% of the answer according to the run metadata, indicating the agent relied on its training data to fill gaps that the site's Cloudflare protection prevented it from answering directly.
  • ›The site's Cloudflare protection created a 'heavy bridge' friction pattern: the agent had to abandon direct navigation and pivot entirely to search-based discovery, yet still assembled a credible, well-sourced answer.
  • ›Critical gaps remained unfilled despite the comprehensive answer: exact parking costs, detailed membership tier benefits, suggested donation amounts, and funding model transparency—information that would likely be findable on the live site but was not accessible through search or prior knowledge.

See every agent that visits nga.gov

Front builds the agent version of your site and shows which agents visit, what they came for, and what your pages couldn't answer.

or talk to us about agent readiness →