Does AI cite LinkedIn, Medium, Quora, or Wikipedia?

A citation link is one cited page in one AI answer. A buying question is an organic prompt in the public Parse index that asks which brand, product, or service to choose. We grouped 13 study platforms where a brand or a third party can publish without owning a website: LinkedIn, Medium, Quora, Substack, Facebook, Instagram, TikTok, X, Pinterest, GitHub, Stack Overflow, DEV Community, and Wikipedia.

Together the 13 study platforms supplied 309,049 of 7,433,783 citation links, or 4.2%. Reddit alone supplied 307,861, or 4.1%. YouTube supplied 372,819, or 5.0%. Every other website combined supplied the remaining 86.7%. The common advice to publish on many platforms describes a small slice of the evidence behind AI answers to buying questions.

LinkedIn led the 13 study platforms with 70,610 citation links, or 0.95% of all links. Medium followed with 65,138, Facebook with 60,675, and Instagram with 48,613. Those four platforms supplied 245,036 links, or 79.3% of the 309,049 platform links.

Quora supplied 18,062 links, GitHub 12,839, DEV Community 10,995, Substack 6,638, TikTok 6,567, and Wikipedia 6,017. Pinterest, Stack Overflow, and X each supplied fewer than 1,200. Reach differs from volume: Facebook appeared for 9,706 prompts, more than LinkedIn's 8,873, while supplying fewer links.

Google AI Mode drew 291,983 of its 5,662,995 citation links from the 13 study platforms, or 5.2%. ChatGPT Search drew 17,066 of its 1,770,788, or 1.0%. Google AI Mode's platform share was 5.4 times ChatGPT Search's. Answer coverage, the share of answers with at least one citation that include the platform group, was 40.7% on Google AI Mode and 3.1% on ChatGPT Search.

The gap is widest for visual and social platforms. Instagram's share of links was 475 times higher on Google AI Mode than on ChatGPT Search, Facebook's 102 times, and Medium's 74 times. Only two study platforms had a higher share on ChatGPT Search: GitHub, at 0.38% against 0.11%, and Wikipedia, at 0.17% against 0.05%. BrightEdge reported the same direction in April 2026: user-generated content made up 7% of Google AI Mode citations and 0.5% of ChatGPT citations.

Wikipedia supplied 6,017 of 7,433,783 citation links, or 0.08%. On ChatGPT Search it supplied 3,053 of 1,770,788, or 0.17%. On Google AI Mode it supplied 2,964 of 5,662,995, or 0.05%. Answer coverage was 0.65% on ChatGPT Search and 0.66% on Google AI Mode.

Published citation studies rank Wikipedia as ChatGPT's largest source. Profound measured Wikipedia at 7.8% of ChatGPT citations from August 2024 through June 2025, and at 5% of citations and 18% of cited conversations from October through December 2025. Semrush reported Wikipedia in more than half of ChatGPT responses in summer 2025, falling below 20% by mid-September. Those studies mix every kind of question. On buying questions, where the answer names brands, Wikipedia is close to absent.

English article pages supplied 5,283 of the 6,017 Wikipedia links, or 87.8%. Pages titled List of something supplied 650, or 10.8%, and pages titled Comparison of something supplied 44. The three most-cited pages were the list of largest video game companies by revenue, with 132 links, the major film studios article, with 106, and the media conglomerate article, with 75. Half of all Wikipedia links went to 334 of 2,492 cited pages.

A recommendation-backing citation link is one the answer attaches to a brand it ranks as a recommendation. The brands most often backed by a Wikipedia citation were large consumer and gaming companies: Nintendo with 92 links across 26 prompts, Sony with 33, and Tencent with 31.

On a matched panel of 14,990 buying questions with citation links in every month on both engines, Wikipedia's share of ChatGPT Search links fell from 0.31% in June to 0.15% in July and 0.06% in August 1 through 23. On Google AI Mode it stayed between 0.05% and 0.06% in all three months.

The matched panel removes movement caused by tracking more prompts over time. The decline is consistent with the reduction in Wikipedia reliance that Semrush observed on ChatGPT in September 2025, measured here on a different set of questions and a later period. LinkedIn moved the other way on ChatGPT Search over the same months, from 0.18% to 0.39%.

LinkedIn was the largest study platform, with 70,610 citation links, or 0.95% of all links. On Google AI Mode it supplied 65,718 links, or 1.16%, and appeared in 47,842 of 376,142 answers with citations, or 12.7%. On ChatGPT Search it supplied 4,892 links, or 0.28%, and appeared in 3,446 of 353,380 answers, or 1.0%.

Semrush reported LinkedIn in 13.5% of Google AI Mode responses and 14.3% of ChatGPT Search responses in January and February 2026, across 325,000 prompts of every kind. The Google AI Mode figure is close to this study's 12.7%. The ChatGPT Search figure is not: on buying questions, ChatGPT Search cited LinkedIn in 1.0% of answers.

On Google AI Mode, Pulse articles supplied 36,061 of 65,718 LinkedIn links, or 54.9%. Member posts supplied 18,089, or 27.5%, and collaborative articles 7,741, or 11.8%. Personal profiles supplied 1.4% and company pages 1.2%. Semrush found a similar mix in early 2026, with articles at 50% to 66% of cited LinkedIn content and posts at 15% to 28%.

ChatGPT Search used LinkedIn differently. LinkedIn product pages, such as the homepage, LinkedIn Jobs, LinkedIn Learning, and LinkedIn Recruiter, supplied 3,929 of its 4,892 LinkedIn links, or 80.3%. Those links appear when LinkedIn itself is the product being recommended. Member and company content supplied 846 links, or 17.3%. Among the 17,716 titled Pulse articles, 3,091, or 17.5%, carried a roundup-style title such as best, top 10, or alternatives.

Medium supplied 65,138 citation links, or 0.88% of all links. Google AI Mode supplied 64,863 of them and ChatGPT Search 275. Medium's share of links was 1.15% on Google AI Mode and 0.016% on ChatGPT Search, a 74-fold gap.

On Google AI Mode, individual author pages at medium.com/@name supplied 34,421 links, or 53.1%. Publications and tag pages on medium.com supplied 17,266, or 26.6%, and author or publication subdomains 13,176, or 20.3%. The largest publications were spread thin: the leading one, medium.com/data-science, supplied 588 links across 304 pages, under 1% of Medium's total.

Facebook supplied 60,675 citation links and Instagram 48,613. Google AI Mode supplied 60,489 of the Facebook links and 48,581 of the Instagram links. ChatGPT Search supplied 186 and 32. On Google AI Mode, Facebook appeared in 12.3% of answers with citations, Instagram in 8.8%, and TikTok in 1.5%.

Facebook groups supplied 31,140 of the Facebook links, or 51.3%. Posts, videos, and photos supplied 27,493, or 45.3%. Pages and profiles supplied 3.0%. Instagram posts and reels supplied 47,622 of the Instagram links, or 98.0%, and account profiles 2.0%. TikTok videos supplied 6,067 of 6,567 TikTok links, or 92.4%.

Quora supplied 18,062 citation links, or 0.24%. Google AI Mode cited Quora for 4,114 of its 18,200 prompts, or 22.6%. ChatGPT Search cited Quora for 27 of its 18,115 prompts, or 0.15%. DEV Community reached 11.5% of Google AI Mode prompts and 0.5% of ChatGPT Search prompts. Substack reached 13.1% and 0.3%.

GitHub was the exception among developer platforms. It reached 7.0% of ChatGPT Search prompts and 8.5% of Google AI Mode prompts, and repository root pages supplied 8,624 of its 12,839 links, or 67.2%. Stack Overflow, X, and Pinterest each reached under 3.5% of prompts on either engine.

Among the 16 industry groups with at least 100,000 citation links, the 13 study platforms' share ranged from 6.0% in Internet Services and Artificial Intelligence to 2.4% in Financial Services. Medium was the leading study platform in six industries, including Software at 2.1% and Artificial Intelligence at 2.9%. Facebook led in Commerce and Shopping, Food and Beverage, Media and Entertainment, and Real Estate. Instagram led in Consumer Goods at 2.2% and Health Care. LinkedIn led in Professional Services, Sales and Marketing, Financial Services, and Education.

The split is sharp. LinkedIn and Medium together supplied under 0.15% of links in Consumer Goods and Food and Beverage. Facebook and Instagram together supplied under 0.5% in Artificial Intelligence, Software, and Data and Analytics.

Not every citation supports a brand. Using the recommendation-backing citation link defined above, across the 13 study platforms, the recommendation-backing rate ranged from 55.5% for GitHub and 48.2% for Wikipedia to 38.0% for TikTok. LinkedIn's rate was 42.3%, Facebook's 40.7%, and Medium's 40.0%. Reddit's rate was 41.4%. YouTube's was 28.6%.

The rate was higher on ChatGPT Search, where 65.2% of the 17,066 platform links backed a ranked brand, than on Google AI Mode, where 40.6% of 291,983 did. Google AI Mode cites platforms more often, but a smaller share of those citations is tied to a named recommendation.

We began with 7,814,836 citation links in answers to organic buying questions on ChatGPT Search and Google AI Mode. We excluded 218,332 links to Google search result pages, 133,355 links to Google redirect addresses that do not identify a publisher, and 29,366 links to Parse's own websites. That left 7,433,783 links. A second derivation from answer-level records with the same exclusions produced the same per-engine totals, 1,770,788 and 5,662,995, and the same count for every platform.

Each platform is a registered website, so regional and author subdomains count toward it. Publications that moved to their own domains, such as Medium or Substack newsletters on custom domains, are not counted as platform links. The window ends on August 23, 2026, before a change in how Google AI Mode reports its sources. Page types are assigned from the address, not from the page content.

The 13 study platforms combined supplied 4.2% of citation links on buying questions, about the same as Reddit alone. Google AI Mode supplied 94.5% of those platform links. Wikipedia supplied 0.08% of all links.

Decide by engine first. If Google AI Mode answers matter, publish Pulse articles and substantive posts on LinkedIn, articles on Medium for software topics, and take part in Facebook groups and Instagram for consumer categories. If ChatGPT Search answers matter, platform work adds little beyond GitHub for developer tools; the evidence there comes from other websites. Do not fund a Wikipedia effort to change buying-question answers. Measure by answer coverage on the prompts that matter, and recheck after each engine change.