Which Sites Does AI Cite When Buyers Ask What to Buy? (October 2026)
The most cited site is youtube.com, in 23.3% of AI answers that cite sources, ahead of linkedin.com at 20.9% and reddit.com at 11%.
CatalystMost Cited Sites in AI Answers to Buying Questions
youtube.com23.3%
linkedin.com20.9%
reddit.com11%
facebook.com7.8%
instagram.com5.2%
medium.com3.9%
gartner.com2.7%
guideflow.com2.5%
tripadvisor.com2.4%
techradar.com2.4%
en.wikipedia.org1.9%
pmc.ncbi.nlm.nih.gov1.9%
g2.com1.8%
forbes.com1.8%
quora.com1.5%
prnewswire.com1.3%
ustechautomations.com1.1%
finance.yahoo.com1.1%
airwallex.com1.1%
indeed.com1.1%
When a buyer asks an LLM which company to pick, these are the sites it links to in its answer. Based on 3,485 answers with sources from ChatGPT and Google AI Overviews, July 17, 2026 to October 15, 2026.
youtube.com, linkedin.com and reddit.com lead
youtube.com is cited in 23.3% of answers with sources, followed by linkedin.com at 20.9% and reddit.com at 11%. facebook.com (7.8%), instagram.com (5.2%), medium.com (3.9%) and gartner.com (2.7%) come next.
Each LLM has its own favorite. The most cited site is techradar.com in ChatGPT (5.3%) and youtube.com in Google AI Overviews (39%).
Where to show up
- Publish short videos that answer buying questions by name, with a full description and transcript. YouTube is cited in 23.3% of answers with sources.
- Post from the company page and from your founder's profile, and keep both current. LinkedIn is cited in 20.9% of answers with sources.
- Answer the buying questions in your category's subreddits under a named account. Reddit is cited in 11% of answers with sources.
Which Sites Each LLM Prefers
instagram.comHow we collect this data
Every AI visibility audit we run writes buying questions for a company’s market and puts each one to every LLM through its API, with web search on. We record the answer and the pages it cites, and these reports add them up, one audit per company.
Written by the audit for each company's own market.
Every answer and every source it cites, from the last 90 days.
Each report keeps its earlier editions, so you can see what moved.
Added up across companies. No single company's results are shown.
Method notes
- Window
- July 17, 2026 to October 15, 2026, rebuilt October 15, 2026.
- Sample
- 517 companies across 199 categories, one audit per company, its latest in the window. Each audit writes buying questions for that company's market with an LLM and puts each one to every LLM through its API.
- Models
- ChatGPT means GPT-4o mini with web search, through the OpenAI API. Claude means Claude Haiku 4.5 with web search, through the Anthropic API. Gemini means Gemini 2.5 Flash with Google Search grounding, through the Gemini API. Perplexity means Perplexity Sonar, through the Perplexity API. Google AI Overviews means the AI Overview on a live Google results page, read through DataForSEO.
- How it is counted
- A site counts once per answer, however many of its pages the answer links. Subdomains stay separate, so en.wikipedia.org is its own row. Links to Google Maps, Google search and product viewers, and image servers are dropped first, because none of them is a page a reader can check.
- Citation figures
- Cover ChatGPT and Google AI Overviews, counting only the pages each answer cites. Claude, Gemini and Perplexity join once 300 of their answers are recorded with cited pages alone, which started on October 15, 2026. Until then their answers count toward the totals only.
Related data reports
See which sites AI cites in your category
The free audit puts your buyers’ questions to each LLM and shows who gets named, who gets cited, and where you are missing.