"Which industries get cited most?" sounds like one question. In the public data it is at least three: in which industries AI answers appear most often, which industries receive the most traffic from AI engines, and which kinds of sources get cited within each industry. There is published data on all three, almost all of it from companies that sell search or AI-visibility software, and the studies rarely share industry definitions, query sets or time periods. This piece compiles the main findings with their samples and dates, then explains where and why they diverge.

Where AI answers appear most often

Health leads every dataset that reports it, but by very different margins.

  • BrightEdge, tracking its own keyword set from May 2024 to December 2025, found AI Overviews on 88% of healthcare queries in December 2025 and on 21% of finance queries. It attributes the gap to query mix: about 70% of the finance queries it tracks are stock-ticker lookups, where AI Overviews appeared 8% of the time, while educational finance queries reached 67%.
  • Conductor, analyzing 21.9 million US Google searches between September 15 and October 12, 2025, found AI Overviews on 48.75% of Health Care searches, 25.79% in Financials and 25.40% in Utilities, against 6.82% in Consumer Staples and 4.48% in Real Estate. Its overall rate was 25.11%.
  • Semrush, tracking more than 10 million keywords through 2025, found the highest November 2025 rates in Science (25.96%), Computers & Electronics (17.92%) and People & Society (17.29%), with Real Estate, Shopping and Arts & Entertainment below 3%.
  • A later Semrush study of more than 600,000 keywords in ten industries, from November 2025 to April 2026, found the share of commercial-intent searches showing an AI Overview grew 71% overall and 231% in finance, while the share of transactional searches fell 5%.

The disagreement on health, 88% in one dataset and 48.75% in another, is largely about what is counted. BrightEdge and Conductor each use their own tracked keywords, and Conductor maps domains to the GICS industry standard; Semrush uses topical categories across a much larger database. Semrush's write-up also describes Health both as one of the largest decreases since March 2025 and as a category with a large rise, so we do not rely on it for that category. The point of agreement is at the bottom: Real Estate is among the least-covered categories in both Conductor's and Semrush's data.

Intent inside a category matters as much as the category. Seer Interactive found AI Overviews on 95.4% of the comparison queries in its 2026 dataset but on 5% of transactional ones.

Which industries receive AI referral traffic

Two large datasets measure this, in different units.

  • Conductor, using traffic from 1,215 enterprise customer domains in the US between May and September 2025, reports AI referrals as a share of all visits: 2.80% in Information Technology, 1.91% in Consumer Staples, 1.08% in Financials, 0.98% in Health Care, 0.86% in Industrials, 0.75% in Materials, 0.68% in Consumer Discretionary, 0.41% in Real Estate, 0.35% in Utilities and 0.25% in Communication Services.
  • Similarweb, using worldwide desktop and mobile web data from June 2025 to May 2026, reports absolute volume. Marketplaces led with 46.8 million average monthly AI referral visits, followed by news and travel at 44.5 million each, finance at 19.1 million and consumer electronics at 17.1 million. Growth was fastest in beauty (up 312.5%) and fashion (up 278.2%), from small bases.

These do not contradict each other. One measures a slice of each site's own traffic, the other counts visits across the web, and their category boundaries differ. Similarweb's data also suggests why some categories receive more: in May 2026, 22.6% of ChatGPT answers about travel included web citations (US, desktop), against 6.8% across all conversations, and 13.5% in retail. Earlier Adobe data pointed the same way: between July 2024 and February 2025, generative AI referrals grew 1,700% for US travel sites and 1,200% for banking and retail sites.

What gets cited within an industry

Here the studies reach opposite headline conclusions, for reasons that can be traced.

  • Yext analyzed 6.8 million citations from ChatGPT, Gemini and Perplexity in July and August 2025, generated from location-based queries across its client and prospect locations in four industries. Listings accounted for 52.6% of healthcare citations, with directories such as WebMD and Vitals prominent; first-party websites for 47.6% in retail; brand-owned websites for 48.2% in financial services; and reviews and social content for 13.3% in food service, the highest of the four. Overall, Yext attributes 86% of citations to websites and listings that brands manage.
  • Muck Rack analyzed more than 25 million links from ChatGPT, Claude and Gemini across 17 industries, attributing 84% of citations to earned media and 27% to journalism. Axios was among ChatGPT's three most-cited domains in 13 of the 17 industries.
  • Conductor's citation data, drawn from 17 million AI responses between May and September 2025, names category leaders: Amazon took 17.99% of citations in Consumer Staples, NerdWallet 6.73% in Financials, ahead of traditional banks, and Zillow 7.36% of brand mentions in Real Estate.
  • BrightEdge found the share of AI Overview citations that also rank in Google's organic top 10 varies by industry: about 24% in healthcare, 22.6% in B2B tech, 11.3% in finance and 9.3% in restaurants.
  • An academic audit of mental health queries by Nguyen, Torous and colleagues, covering ChatGPT, Perplexity and Google AI Overviews, found the ten most-cited domains accounted for 43.6% of English-language citations, with government, commercial health and academic sources each making up roughly 22%.

Yext's 86% brand-managed and Muck Rack's 84% earned media answer different questions. Yext's prompts ask where to find a branch, clinic or restaurant, questions that directories and location pages answer; Muck Rack's cover broader consumer and industry questions, which editorial coverage answers. Yext also counts listings as brand-managed, where others might treat them as third-party. Both companies sell products tied to the source types their studies favor, as do most firms cited here. An academic preprint from the University of Toronto (Chen et al.) found earned media dominated AI answers to automotive, consumer electronics and software ranking prompts, which supports Muck Rack's pattern for research-style questions.

What an ideal dataset would look like

No public dataset yet answers the industry question cleanly. One that did would need:

  • A shared industry taxonomy, published with the mapping of every query and domain.
  • Queries sampled from real usage, or weighted by it, and labeled by intent, since intent explains much of the variation above.
  • Repeated runs of each prompt reported as distributions; SparkToro found fewer than one in 100 repeated runs returned the same list of brands.
  • Separate metrics for answer presence, citation, brand mention and click, which current studies often merge.
  • Coverage of consumer interfaces as well as APIs, and of more than one country and language.
  • Disclosure of whose sites are in the sample, because vendor client bases skew toward particular industries and company sizes.
  • Open data, or enough detail to reproduce the result.

Until such a dataset exists, the safest reading of the public data is directional. AI answers appear most on educational and comparison questions, which leaves health and personal-finance research heavily exposed. AI referral traffic concentrates in marketplaces, travel, news and IT. And within most industries, which sources get cited depends on whether people are researching a decision or looking for a place to go. For how the engines themselves differ, see our review of cross-engine comparisons; for the mechanisms behind citation, see how AI answer engines choose which sources to cite.