What does AI Read?...and Why It Matters in Public Relations

AI tools like ChatGPT, Perplexity, Gemini, and Claude are rapidly becoming a new gateway to information. Simply, the new Google. For PR pros, this sh

ift raises an important question: what sources does AI rely on when generating responses?

Recently, one of my clients noticed a sharp drop in page visits from Google search. Historically, that single metric had been one of their primary indicators of SEO success, and owned media gold. A decline suggested something was wrong.

But when they called, we decided to dig deeper into their traffic analytics. Instead of focusing only on Google queries (not the best metric), we analyzed all referring traffic sources. What we discovered is percent-for-percent, searchers are shifting from Google queries to referral links coming from tools like ChatGPT and Claude.

In other words, traffic to the site hadn’t actually dropped. It had simply changed.

This most certainly a shift happening across the information ecosystem. As AI tools become a new trusted place for search, traditional SEO metrics alone (or traffic reports) may no longer tell the full story. A dip in Google traffic might not mean visibility is declining. It may mean that large language models (LLMS) are referencing your content directly when answering user queries.

Research supports this trend. A recent report from Muck Rack’s Generative Pulse analyzed more than one million links cited in AI-generated responses and found that 82% of those links came from earned media sources, such as journalism and third-party publications. Yes, good old fashioned press. Overall, 94% of citations came from non-paid media (Muck Rack, 2025). These results reinforce the continued importance of credible, and public relations efforts.

For PR professionals, this helps explain why many industry watchers believe communications is becoming a major influence in AI-mediated discovery. As one analysis puts it, large language models may be positioning PR as a new “power player” in search, because they bring together information from trusted third-party sources rather than relying solely on traditional search rankings or old school methods of pay for rank.

This shift is fueling discussion around Generative Engine Optimization (GEO) and the practice of improving how brands appear in AI-generated answers. Some communicators are exploring how traditional strategies such as media pitching, authoritative coverage, and well-structured information may influence the sources LLMs rely upon (Sword and the Script; Headrick, 2025). Others have suggested that GEO may not be entirely new at all, but rather “PR by another name,” doubling down on the long-standing importance of credibility, relationships, and earned media in shaping visibility (PRSA).

For practitioners, the takeaway is simple: AI systems are becoming a new audience for information. Understanding what they read, trust, and cite may soon be just as important as understanding the audiences we traditionally aim to reach.

By: Joshua J. Smith, PhD

Co-Founder, Chair

The Media + AI Initiative

Ref.

Headrick, P. (2025, May 23). AI and large language models (LLMs) are changing how information is discovered and synthesized [LinkedIn post]. LinkedIn.https://www.linkedin.com/posts/parryheadrick_ai-and-large-language-models-llms-are-activity-7331402667606327296-n7tr/

Muck Rack. (2025). What is AI reading? Generative Pulse report. https://generativepulse.ai/report/

Sword and the Script. (2025, February). How can communicators pitch an LLM? https://www.swordandthescript.com/2025/02/how-can-communicators-pitch-an-llm/

PRSA. (2025, June 18). LLMs just made PR the new power player in search. https://prsay.prsa.org/2025/06/18/llms-just-made-pr-the-new-power-player-in-search/

PRSA. (2025, December 10). Why GEO is just PR by another name. https://prsay.prsa.org/2025/12/10/why-geo-is-just-pr-by-another-name/

AI Disclosure

Portions of this post were developed with the assistance of generative AI (Gemini) to support drafting and editing. The author reviewed, revised, and verified the content, analysis, and sources to ensure accuracy and alignment with the cited materials.

Image credit: Google Gemini