The End of PR as We Knew It: How AI Search Rewrote Visibility, Influence, and Reputation

Organic CTR for queries that trigger Google AI Overviews fell 61% between June 2024 and September 2025, dropping from 1.76% to 0.61%. For brands that built awareness through traditional earned media, that is a structural rupture: the article still ranks, the press hit still went live, and nobody clicks through.

The collapse happened in plain view. Google traffic dipped for publishers and brands alike, SEO dashboards showed stable rankings, but sessions fell week over week. AI Overviews had arrived to answer questions directly at the top of the page, absorbing the click before it reached any domain, while the ranked content itself remained untouched. PR teams running on pre-2024 measurement frameworks are tracking a funnel that no longer converts at the rate their dashboards imply.

Press Coverage No Longer Drives Discovery

More than half of all U.S. Google searches now end without a single click to any external website. Add AI chat platforms, where there are no links to click at all, and the share of brand discovery that never touches a website is substantially higher. A Forbes feature, an Inc. profile, a TechCrunch announcement: all still carry cultural weight. But the traffic they once reliably delivered has been intercepted by systems that summarize, extract, and serve information without a referral.

Press coverage earns credibility and reinforces brand authority. The additional question for AI-era PR is whether that coverage reaches the language model training surfaces that determine brand visibility inside AI responses. When someone asks Perplexity “What are the best clean skincare brands?” the system doesn’t check who got featured in Allure last month. It checks which brands appear consistently across structured sources: FAQs, schema-marked product pages, forum discussions, and citation trails from sources the model was trained on. A single press win is an input. It only becomes sustained visibility if it’s reinforced across the surfaces AI systems draw from.

Entity Recognition Determines AI Brand Visibility

Generative AI systems build answers from entity graphs: structured webs of relationships connecting brands, people, products, and claims. Media prestige is not a ranking signal in that construction. A brand that appears consistently across multiple trusted surfaces, with the same name, the same core description, and the same positioning repeated across its website, press profiles, schema markup, and community content, gets recognized as a reliable entity. Inconsistency across those surfaces gets resolved by the model defaulting to whatever version appeared most often, which may not be the one the company intended.

Entity recognition explains why two brands with identical press coverage can have completely different AI visibility. If your Schema.org markup describes your company one way, your founder’s byline bio describes it another way, and your Crunchbase entry uses a third version of your tagline, the model has competing claims with no stable ground. It picks the most-repeated version, which may not be yours. This is how brands end up described incorrectly in AI responses despite having strong press coverage. Canonical messaging, consistent across every machine-readable surface, is the foundation AI citation is built on.

The Inscribe campaign shows what this looks like when it works.

Reddit Outranks Forbes Inside LLM Training Data

Reddit is the most-cited domain across AI-generated responses from multiple platforms, according to a citation analysis of 150,000 citations across 5,000 keywords. It outpaced Wikipedia, YouTube, and major news outlets across the keyword categories the study covered. AI systems are trained disproportionately on conversational, user-generated content: the kind that sounds like real people discussing actual experiences with products and brands. That training bias is why community platforms rank above prestige publications in LLM citation data.

A Reddit thread where users discuss your product can train more AI models than a full-page profile in a national publication. A Substack essay with 1,500 subscribers, if it uses consistent language and answers a specific question well, often gets cited more reliably than a TechCrunch feature. Press placements still carry real weight with buyers and in sales conversations. The AI visibility question is whether that coverage reaches the communities and platforms where language models draw their training signal, which often have far less prestige than the publications in a typical pitch deck.

The PR + Search + LLM Flywheel

The brands sustaining AI visibility have built a flywheel where each earned media win compounds into the next citation cycle. A press placement gets indexed; that article shows up referenced in a Reddit thread using the same phrasing; the thread gets scraped into a training dataset; the brand’s FAQ uses matching language with schema markup attached. When someone asks Gemini a relevant question three months later, the brand surfaces because the accumulated pattern of consistent, structured mentions across trusted surfaces is what the model learned to recognize.

This is the core strategic change: from chasing individual placements to building a system where each placement generates derivative citations. PR teams executing this well don’t stop at the article. They embed the coverage phrasing into their FAQ copy. They answer Reddit threads using language consistent with the press hit. They update schema markup to reflect the framing the coverage established. Each earned media placement is raw material for that flywheel, worth considerably more than its initial publication metrics suggest.

Canva is the most studied example of this at scale. Years of publishing high-volume, structured design education, under consistent author names, with deep schema markup and internal linking, turned their content into a training reference point. Ask Perplexity or Gemini about free pitch deck tools or visual branding best practices today, and Canva surfaces because the model was trained on enough consistent, trusted mentions to treat them as a default authority in that space. That outcome didn’t come from any single press win. It came from a flywheel built over years.

Hallucination-Proofing Your Brand

AI hallucination is a consistency problem: when multiple sources describe a brand differently and no single version repeats often enough to establish ground truth, the model synthesizes its own interpretation. If your homepage describes your company one way, your press releases use different language, and your founder’s LinkedIn bio was written in 2019 and never updated, the model has competing claims with no stable ground. It picks the most-repeated version, or synthesizes a middle ground that matches none of your intended positioning. For well-funded brands with strong coverage, this can mean being described in AI responses as doing something that resembles your work without accurately representing what your company does.

The fix is canonical messaging: one specific, citable description of what your company does, written in the same language everywhere it appears. One specific, searchable, citable sentence, precise enough to appear in an AI response verbatim. It goes in your Organization schema (name, description, and sameAs fields). It opens your About page. It’s the first sentence of every executive bio. Press release boilerplate uses it verbatim. When AI encounters that phrase across enough different sources, it stops generating its own interpretation and starts echoing yours. That transition, for brands that commit to it, typically takes six to eight weeks to show up in AI responses.

Building a Machine-Readable Press Presence

The PDF press kit is a dead end for AI discoverability. PDFs bypass the structured formats AI systems prefer, they do not update when information changes, and they carry no connection to the schema layer that makes brand information machine-readable at scale. Brands appearing reliably in AI summaries have replaced static press kits with structured media hubs: live web pages using JSON-LD schema types (Organization, Product, Person, FAQPage) to declare who they are, what they do, and what third-party sources confirm about them.

The $250 million licensing deal News Corp signed with OpenAI in May 2024 makes the structural point clearly. The agreement covered WSJ, Barron’s, MarketWatch, the New York Post, and UK and Australian outlets. Its terms weren’t just about article access. They were about structured, branded data being ingested into GPT models so that News Corp entities (journalists, publications, topics) would be recognized and cited with attribution. The same principle applies at every scale. Brands getting cited in AI responses aren’t the ones with the best-designed media pages. They’re the ones whose information is structured well enough for a language model to extract and trust.

Reactive PR: Publish Before the Coverage Cycle Starts

AI search tools prioritize freshness. When a major news event breaks, the brands that publish a structured, schema-marked answer first tend to appear in AI-generated responses for days afterward, not the ones with the highest domain authority. Brands that publish a structured answer within hours of a relevant news event become the cited source that subsequent AI responses reference. Speed to structure determines that outcome more than speed to publish.

Three quarters of PR professionals now use generative AI in their workflow, nearly tripling the adoption from 2023. The most sophisticated teams are using AI to move faster on reactive content: monitoring for query surges, drafting structured FAQs within hours of a relevant news event, and deploying them with proper markup before the major outlets finish their explainer pieces. A 300-word FAQ published within hours of a major announcement, with schema, clear H2 structure, and internal links to relevant service pages, consistently outperforms long-form reactive content published the following day.

Three Metrics That Replace Impressions

When I review reporting dashboards with PR clients, impressions lead the deck almost every time. I understand why, because it is what clients and boards have accepted as the standard for years. The problem is that impressions measure reach in a world where people click through to read things, and AI search changed that. The impression now often happens inside the AI response itself, and no pixel tracks it. PR teams reporting primarily on impressions are measuring a layer of the funnel that has lost most of its commercial significance. The metrics that reflect actual influence now are AI citation frequency, share of voice in generative answers, and sentiment tone in AI summaries.

Citation frequency tracks how often your brand appears in AI-generated responses when relevant questions are asked, across ChatGPT, Perplexity, Gemini, and AI search modes. Share of voice compares that frequency against competitors in the same category. Sentiment tracks whether AI summaries frame your brand positively, neutrally, or with qualifications. Monitoring platforms dedicated to AI visibility now track all three across ChatGPT, Perplexity, Gemini, and AI search modes. Building a reporting structure around these metrics is the only way to see where actual brand visibility lives in 2025.

Your 90-Day AI PR Action Plan

Start with what AI currently says about your brand across ChatGPT, Perplexity, and Gemini. Write it down, screenshot it, and treat it as your baseline. The gap between those descriptions and the positioning you’ve built over years of earned media is almost always more significant than teams expect, and it is the first thing to address. The 30-60-90 sequence below follows from that diagnosis.

Related: Build a B2B PR Strategy: 12 Tactics That Generate Coverage and Pipeline

Related: B2B AI Visibility: The 2026 Playbook for Getting Cited by ChatGPT, Perplexity, and Gemini

For brands that want to move faster, reach out to Zen Media to start with an AI visibility audit covering your current citation rate, entity recognition gaps, and structured content opportunities.

Frequently Asked Questions: PR in the Age of AI Search

How has AI search changed PR strategy?

AI search has moved the primary visibility layer off the website and into AI-generated answers. Organic click-through rates on queries that trigger Google AI Overviews dropped sharply between 2024 and 2025, and that drop reflects a structural change in how search pages are built. PR strategy now requires structuring content, messaging, and citations to appear inside AI responses, with traditional search rankings as one component of a broader visibility system.

What is AI citation frequency and why does it matter for brands?

AI citation frequency measures how often a brand appears in responses generated by tools like ChatGPT, Perplexity, and Gemini when users ask relevant questions. It matters because the majority of U.S. Google searches now end without a click to any website, meaning AI-generated responses are often the only touchpoint a potential buyer has with a brand before forming an opinion.

How do you get your brand cited in AI-generated answers?

Brands get cited in AI answers by publishing structured, machine-readable content on platforms AI systems index heavily: Reddit, Wikipedia, niche publications, and structured FAQs. Schema markup, consistent entity information across all web properties, and repeated mentions in contextually relevant sources all increase citation frequency in AI-generated responses.

What is zero-click reputation?

Zero-click reputation is the brand perception formed entirely within AI-generated summaries and search snippets, before a user ever clicks to a website. Because most searches now end without a click, the two or three sentences an AI generates in response to a brand-related query often constitute a buyer’s complete first impression of the brand.

How do PR teams measure success in AI-first search?

The three metrics replacing traditional impressions in AI-first PR are AI citation frequency (how often the brand appears in relevant AI responses), share of voice in generative answers (brand mentions relative to competitors in the same category), and sentiment tone (whether AI summaries frame the brand positively, neutrally, or negatively). Monitoring platforms now offer dashboards for all three metrics across major AI systems.

About the author: Sarah Evans is Partner and Head of PR at Zen Media, a global B2B PR and marketing agency. With 23+ years in communications, she architects PR strategy, drives earned media initiatives, and helps brands navigate AI-driven visibility. She is a regular contributor to Entrepreneur and has been recognized as a top writer on business and tech.