AI Search Cites News Coverage 80% of the Time, So Earned Media Now Beats Every Other Backlink

AI search engines cite earned media, meaning third-party news coverage, about 80% of the time, while Google splits its citations roughly evenly between brand sites and earned sources. That is the core finding @heyemmettdotcom pulls from a new University of Toronto paper on Generative Engine Optimization. The practical consequence is that press coverage from outlets like Forbes or TechCrunch now matters more for AI visibility than it ever did for traditional SEO.

The paper behind the claim

The video walks through "Generative Engine Optimization: How to Dominate AI Search," a September 2025 paper by Mahe Chen, Xiaoxuan Wang, Kaiwen Chen, and Nick Koudas at the University of Toronto. The creator holds the paper on screen for most of the video, starting with the title page and author list before zooming into a results figure.

First page of the Generative Engine Optimization paper with the creator's caption asking which backlinks are important

The study compares which sources Google and AI search engines cite when answering the same queries, across countries and verticals.

Three backlink categories

The paper sorts every citation into three buckets. Branded means the domain belongs to a company or app, whether that is geico.com or a small personal blog. Earned means third-party news coverage, with Forbes and TechCrunch given as examples. Social means community platforms like Reddit and Quora.

Figure 2 of the paper showing the distribution of Brand, Earned, and Social sources across Google and GPT in Canada and the USA

This taxonomy is the useful part. Most link-building advice treats all backlinks as one pile. The paper measures how differently each engine weighs the three types.

Google vs. AI search: the numbers

For the US automotive vertical, the paper reports that Google's citations broke down as 39.5% brand, 15.4% social, and 45.1% earned. The creator rounds these to 40% brand, 15% social, and 45% earned, and notes that less than half of Google's citations are news coverage.

Highlighted passage in the paper listing the percentage breakdown of Brand, Social, and Earned sources for Google and AI search

AI search flips the mix. For the same US automotive queries, GPT's citations were 81.9% earned and 18.1% brand, with social sources absent. The creator's takeaway is direct: getting third-party news coverage is now more important than it was in the Google era.

What about Reddit and Quora?

The creator anticipates the obvious objection, since people see AI answers referencing Reddit and Quora constantly. His explanation is that this happens through the web search toggle. When a user turns on web search, the AI pulls brand mentions from Reddit and blends them into the answer text rather than showing Reddit as a cited link. So social platforms still influence what the AI says about a brand, but they do not show up in the citation data the paper measures.

Key Takeaways

A University of Toronto paper categorizes backlinks as branded (company domains), earned (news coverage), and social (Reddit, Quora).
In the US automotive data, Google cited 39.5% brand, 15.4% social, and 45.1% earned sources, a fairly balanced mix.
AI search cited 81.9% earned and 18.1% brand for the same queries, with no social citations.
Third-party press coverage is the highest-leverage backlink type for AI search visibility.
Reddit and Quora still shape AI answers through web search blending, but as background material rather than cited links.

Resources

Generative Engine Optimization: How to Dominate AI Search (arXiv:2509.08919) — The University of Toronto paper shown in the video, by Mahe Chen, Xiaoxuan Wang, Kaiwen Chen, and Nick Koudas.