ChatGPT Cites Sources Most Often in Travel Answers, Similarweb Finds
Similarweb's 2026 Generative AI Landscape report found that 22.6% of ChatGPT travel answers included at least one citation in May 2026, compared with 6.8% across all measured topics and 4.8% for education. The U.S. desktop estimates show that the opportunity to earn a link from ChatGPT varies sharply by subject.
Similarweb's 2026 Generative AI Landscape report found a wide gap in how often ChatGPT answers linked to external sources in May 2026. In the company's U.S. desktop estimates, 22.6% of travel answers included at least one citation, the highest rate among nine measured sectors. Retail followed at 13.5%, sports at 10.7%, finance at 8.0%, and education ranked last at 4.8%. The average across all answers was 6.8%, up from roughly 1.3% in June 2025.
Citation opportunity depends on the question
The figures measure whether an answer contained at least one external link. They do not measure how many links appeared, whether a cited page received a click, or whether a citation stayed visible over time. That distinction matters: a higher citation rate creates more opportunities for publishers to appear, but it does not guarantee referral traffic.
Source selection also varied by topic. Similarweb reported that reviews and user-generated content made up 54.1% of citations in travel answers. Across the top 10,000 U.S. desktop citations in May, reviews and user-generated content accounted for 28.9%, while news and publisher sites accounted for 26.0%.
What the numbers mean for publishers
The practical lesson is not that every site should produce travel content. It is that AI visibility should be measured within the questions a site is qualified to answer. A publisher in education is competing inside a much smaller pool of linked answers than a travel publisher, so raw citation counts are not comparable across sectors.
LDS interpretation: teams should separate three metrics that are often blended together—answers containing a link, citations earned by a domain, and visits that arrive from those citations. Similarweb's report addresses the first two parts of that funnel; analytics on the publisher's own site are still needed to establish traffic and business value.
The estimates cover U.S. desktop use and exclude API activity, desktop applications, and AI features embedded in other apps. Similarweb also cautions that its figures are estimates extrapolated from third-party data rather than a complete record of ChatGPT activity.
Key Points
- 1Travel led nine measured sectors, with citations in 22.6% of ChatGPT answers in May 2026.
- 2The average citation rate was 6.8%, while education ranked lowest at 4.8%.
- 3Reviews and user-generated content supplied 54.1% of citations in travel answers.
- 4The estimates cover U.S. desktop use and do not measure clicks or downstream conversions.
Scoring Rationale
Similarweb's fresh cross-sector data gives publishers a useful benchmark for AI citation opportunity, but the estimates are limited to U.S. desktop use and do not establish referral or conversion impact.
Sources
Primary source and supporting public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems
