top of page

Generative Engine Optimisation for PR Agencies: How Earned Media and Citations Influence AI Recommendations

Oct 12
7 min read

Summary

Two studies of press releases in AI search can look like opposites, one reporting that 99.3% of wire releases were cited by an LLM (large language model) and another putting press releases at about 1% of the links AI answers cite, and explaining that gap is the first job of generative engine optimisation for PR agencies. The research agrees on the broad shape: engines lean on third-party coverage, and a release does its work through the editorial pickup, trade coverage and newsroom page that follow it. The sections below cover which releases get picked up, how big-name outlets and trade press divide the work and a six-step routine for turning coverage into citations.


Cover for generative engine optimisation for PR agencies with three figures: 99.3% of 8,000+ GlobeNewswire releases cited by ChatGPT or Claude, 38% of releases with a top-50 outlet pickup surfacing in AI citations against 6% without, and 75.8% of news citations in 16 US industries going to trade media.

PR agencies are judged by coverage, and LLMs now read that coverage before deciding which brands to name. The IPRN PR Business Survey 2026 of member agencies in 22+ countries has every agency using AI, 15% generating business from generative engine optimisation (GEO) and 41% building knowledge and starting tests. A Muck Rack survey of PR professionals, fielded in May and June 2026, puts the share not measuring GEO success at 39%. The companion breakdown on answer engine optimisation for PR agencies covers page structure and extraction. This article stays with the citation economy: which placements engines read, which they show and what a PR team can do about both.


Generative engine optimisation starts with earned media

Generative engine optimisation starts with earned media because AI search favours independent coverage over what brands publish about themselves. A September 2025 study by Chen, Wang, Chen and Koudas compared AI search engines with Google and documented a "systematic and overwhelming bias towards Earned media" over brand-owned and social content. A January 2026 paper by the same four authors at the University of Toronto put numbers on it for 300 consumer-electronics queries: Google's results split 41% earned, 34% social and 26% brand, while Claude drew 65% of its sources from earned media and GPT-4o 57%.


Do press releases influence AI recommendations?

Press releases influence AI answers mainly through the coverage, trade pickups and newsroom pages they produce, and LLMs pick up the releases themselves hours after distribution. Notified, the owner of GlobeNewswire, reported in July 2026 that 99.3% of more than 8,000 releases from a one-week period in May were cited by ChatGPT or Claude, with eight hours as the average time to first citation; the report does not define "cited". Muck Rack's Generative Pulse tracked press-release citations growing fivefold between July and December 2025.


Counts of the links shown in answers put the release itself at a small share. Muck Rack's May 2026 edition, covering more than 25 million links, has press releases on any site at 1.1% of citations, and BuzzStream's January 2026 week of 4 million citations has syndicated releases at 0.04% and newswire domains at 0.21%. EPR Research, the research desk of the PR news site Everything-PR, tested 250 PR Newswire and Business Wire releases and traced where the citations went: ChatGPT's citations for the releases that surfaced included 2 wire URLs and 26 editorial pages, and the citations from Gemini, Claude and Google AI Overviews went to editorial pages. The release sets the chain in motion, and the pickup collects the citation.


Which releases get picked up

Release type, editorial pickup and a citable number in the opening paragraph decide most of the difference. In EPR Research's 250-release test, 47 releases (19%) surfaced in at least one engine's citations within 90 days. The rate was 40% for acquisition releases and 30% for regulatory filings, ahead of 14% for product launches and 10% for executive appointments. None of the 50 partnership, award or other announcements surfaced, which makes the newsroom the better home for those.


A pickup by a top-50 outlet lifted the rate to 38% from 6%, and a citable number in the first 100 words lifted it to 28% from 9%. In Muck Rack's December 2025 edition, cited releases carried about twice as many statistics and 2.5 times as many bullet points. Notified's analysis shows 2.6 times more citations for well-structured releases. These figures come from vendors and a PR news site, and most rest on small samples. The multiples show direction, and the exact values will move.


The newsroom page

A structured newsroom page gives LLMs a first-party source to cite for brand questions. BuzzStream's January 2026 data put newsroom content at 3% of all citations and 18% of ChatGPT's citations, with about 2% on Google's platforms. In Notified's data, newsroom, blog or campaign pages were twice as likely to be cited in the first 30 days when a press release supported them.


Big-name outlets and trade press divide the work

Big-name journalism leads for news and recency questions, and trade press leads inside most verticals.


Big-name outlets and licensing deals

Journalism supplies about 27% of the links AI answers cite, and earned media as a whole about 84%, in Muck Rack's May 2026 edition of Generative Pulse, which covered more than 25 million links across 17 industries and three engines. Those citations spread over more than 20,000 outlets, with Axios in ChatGPT's top three domains in 13 of 17 industries, and 57% of journalism citations with a known date were under 12 months old. A June 2026 study of 129.3 million citations by the press-release distributor Press Ranger and an AI search monitoring vendor showed publishers with OpenAI licensing deals averaging 10.2 citations per page on ChatGPT, where publishers without one averaged 6.9. The data covers one month and two commercial parties. It shows a correlation and leaves cause open.


Specialist trade press

Trade press wins most news citations inside a vertical. Press Ranger's June 2026 study gives trade and niche media 75.8% of news citations in 16 US industries, against 24.2% for mainstream outlets, and the lead in 15 of the 16. Finance and insurance was the exception, with mainstream outlets at 52.2%.


Syndicated copies and the original story

Original editorial coverage collects the citations that syndicated copies rarely earn. BuzzStream's January 2026 week put original editorial content at 81% of news citations and syndicated releases on Yahoo and MSN at 0.32%. A pickup that lands on a syndication page adds less than the original story, and agencies should record which URL an engine cites.


Buyer questions draw on different sources

Brand, transactional, category and news questions draw on different sources, and PR work reaches each through a different placement.


Buyer question

Sources engines lean on

Where PR work applies

Brand: "What is [agency name]?"

Wikipedia, press coverage of its releases and the brand's own site and newsroom

Release wording, boilerplate and a structured newsroom.

Transactional: "PR agency retainer cost"

Brand-owned pages, which drew 52% to 68% of citations in the Toronto consumer-electronics test

Service, pricing and proof pages.

Category: "best PR agency for fintech"

Editorial best-of lists, buying guides and reviews

Rankings, review profiles and expert quotes inside those lists.

News: "latest PR industry trends"

Journalism, cited in 46% of industry-trend responses in Muck Rack's data, plus trade press; press releases appear about 3.5 times more often here than in best-of answers

Data-led releases, exclusives and pitch timing.


Category questions deserve the most attention, since a buyer asking one has no brand in mind yet.


A six-step routine for turning coverage into citations

Six steps move a PR agency from counting clips to counting citations.


  1. Build the media list from the outlets and trade titles LLMs cite for the client's category prompts.

  2. Choose release topics with news value, such as acquisitions, regulatory filings and original data.

  3. Open each release with a citable number and name the companies, products and people involved.

  4. Pitch the formats engines cite: best-of lists, buying guides, reviews, comparisons and original data.

  5. Keep a structured newsroom page for every release and key fact.

  6. Log which URL an engine cites (the wire copy, the original pickup or a syndicated copy) and report citation share per prompt cluster and engine.


Step 3 follows the openings data above, and step 6 shows when a syndicated copy collects the citation meant for the original story. Rerunning the same prompts every few weeks tracks the movement per engine. The post on AI visibility metrics: mentions, citations, rank, sentiment and share of voice covers how each one is measured.


Agencies earn recommendations through the same third-party layer

A PR agency earns recommendations in AI answers through the same third-party layer it builds for clients: rankings, directory profiles, trade coverage and expert quotes in best-of lists. SEO specialist Lily Ray ran 100 B2B "best software" queries through Google AI Overviews on three dates between April and June 2026, as reported by Search Engine Land. Of the 100 queries, 80 triggered an AI Overview, and self-promotional listicles were cited 323 times. In 224 of those cases (69%) the AI Overview left the publishing brand out of its recommendations. A "best PR agencies" list on an agency's own site is likely to behave the same way, and a place in an independent ranking is the stronger route.


AI, TELL ME!, a full-cycle GEO and AEO agency, prioritises third-party placements by AI search evidence and relevance to the client's vertical. Its GEO Monitoring shows how a brand appears in LLM answers, and GEO Readiness explains why it appears that way, from its own site and content to third-party sources.


FAQ

What is generative engine optimisation for PR agencies?

Generative engine optimisation for PR agencies is the work of earning coverage and placements that LLMs cite and repeat when buyers ask for recommendations. It adds a measurement layer to media relations: which outlets and URLs get cited and how the answer describes the client.


Do press releases help with AI search visibility?

Press releases help most when they lead to coverage. In EPR Research's test of 250 releases, those with a top-50 outlet pickup surfaced in AI citations 38% of the time, and those without one 6%. The release itself is rarely the link shown, so the pickup, the trade article and the newsroom page carry the citation.


Which press releases are most likely to be cited?

Acquisition and regulatory releases did best in EPR Research's sample, at 40% and 30% within 90 days. Product launches reached 14% and executive appointments 10%, and none of the 50 partnership or award announcements surfaced. A citable number in the opening paragraph raised the rate for every type.


Are big-name outlets or trade press more useful?

Big-name journalism serves news and recency questions, and trade press serves category and industry questions. Press Ranger's June 2026 study gives trade media 75.8% of news citations in 16 industries, and journalism appears in 46% of industry-trend responses in Muck Rack's data.


Is earned media enough for AI recommendations?

Earned media carries heavy weight in AI search, at about 84% of citations in Muck Rack's May 2026 edition, and it shares the source mix with reference pages and community platforms. In Muck Rack's data Wikipedia is ChatGPT's most-cited domain and Reddit is Gemini's, so both need tracking alongside coverage.

See which sources LLMs cite when buyers ask about your agency, starting with the free AEO Monitoring scan from AI, TELL ME!.



bottom of page