遇见数据集

Houston AI Citation Study, October 2026

收藏
Zenodo2026-10-30 更新2026-11-05 收录
官方服务:

资源简介:

What Google AI Mode and Perplexity cite and recommend when people ask about local businesses, and whether AI-ready websites get cited more. On October 30, 2026, 232 questions about 15 local industries in 8 Greater Houston cities (a recommendation question and a price question per industry and city) were sent to Google AI Mode (via SerpApi, searched from Houston) and Perplexity Sonar (API), with no system prompt and no brand named. Every cited URL was coded by source type, every recommended business was extracted with the sources attached to it, and cited local business websites were compared with the 3,014-site census of the Houston AI Visibility Index (doi:10.5281/zenodo.23130268). Headline findings. 95.5% of the businesses AI Mode recommended were Google Business Profile entities (median 4.9 stars, 323 reviews); Perplexity cited business websites for 49.4% and directories for 45.0% of its citations on recommendation questions. The engines named the same top business for 8.3% of questions. Census sites passing four AI-ready basics on October 2, 2026 were cited by October 30 at 5.9% against 1.4% (adjusted odds ratio 4.6, 95% CI 2.3 to 9.2); cited local business sites passed all four basics 65.2% of the time against 32.1% for the census (adjusted OR 3.6), and the gap held across Common Crawl link-authority quartiles. Files. questions (232), answers (696), citations (5,422), recommendations (1,460), summary tables with bootstrap 95% intervals, results by industry, website-check comparisons with adjusted odds ratios, a methods note (PDF), the report chart, and the scripts used for collection, coding and analysis. Answer text and per-site check results are not included. Full report: https://marketingandai.com/research/houston-ai-citation-study/ Disclosure: Marketing & AI is a Houston SEO and AEO agency; SEO agencies were one of the 15 industries. It was not cited or named in any answer. Version 1.1: website factor test. 40 commonly recommended website factors were tested on 880 Houston-area business homepages crawled on October 30, 2026: 492 cited by Google AI Mode or Perplexity and 388 census sites in the same 12 industries that were not. Six factors held up in every test (per-factor logistic regression with industry and Common Crawl authority controls and Benjamini-Hochberg correction, a regularized all-factor model with bootstrap sign stability, gradient-boosted trees with permutation importance, and a census-only cohort check): a meta description (96% of cited sites vs 79%; adjusted odds ratio 5.9), reviews or testimonials on the homepage (3.4), a specific schema business type (3.0), prices on the homepage (2.5), years in business stated (1.8) and city pages in the sitemap (1.6 per SD). 27 factors, including business and FAQ schema, question-style subheadings, llms.txt and freshness signals, were linked to citations alone but explained by the others. Link authority was mixed; page speed, domain age and AI-crawler firewall blocking showed no clear link. New files: factors.csv (40 factors x 4 outcomes), factor-models.csv, the factor chart, updated scripts and methods note. Per-site values are not published. Report: https://marketingandai.com/research/houston-ai-citation-study/ai-citation-factors/

提供机构:
Zenodo
创建时间:
2026-10-30
二维码
社区交流群
二维码
科研交流群
商业服务