How AI reads this page
ChatGPT, Claude, Gemini and Perplexity don't look at colours, images or animations. They receive text, headings, links and structured data. Below is exactly what they read when they open “Do PR and trade publications still matter for AI citations? The 2026 data”.
The AI receives 2,232 words in 8 sections, 13 links and 3 blocks of structured data.
Text
The page content in Markdown, without menus or graphics. It is the format closest to how a language model reads text.
# Do PR and trade publications still matter for AI citations? The 2026 data > Over 84% of AI citations come from third-party sources, yet journalism's share fell to 22% and advertorials are at 0.3%. How to plan media placements in 2026. URL: https://geosnap.ai/en/blog/uffici-stampa-e-testate-servono-ancora-per-farsi-citare-dall-ai-dati-2026 Language: English Version IT: https://geosnap.ai/blog/uffici-stampa-e-testate-servono-ancora-per-farsi-citare-dall-ai-dati-2026 Author: Rinald Sefa, CMO Geosnap · Category: GEO strategy Publisher: Geosnap (Maind Group S.r.l.), https://geosnap.ai ## In short Yes, but differently. According to Muck Rack (28 October 2026), more than 84% of the links cited by ChatGPT, Claude, Gemini and Perplexity come from sources the brand neither owns nor pays for, while journalism's share dropped from 27% to 22% in four months and paid content stays at 0.3%. Two more late-October studies show that the number of different sites citing you matters more than your Google ranking, and that brands cited in informational answers are almost twice as likely to be recommended. PR still works: it has to target the sources AI reads, not circulation. - More than 84% of AI citations come from third-party sources (82-89% across four editions of the Muck Rack study); about 99% from non-paid sources. - Journalism went from 27% to 22% of citations between May and September 2026; third-party corporate content, encyclopedias, public bodies and brand sites are growing. - Advertorials and paid content account for 0.3% of citations; on one large outlet, editorial content is cited 1.7 times more often than the pay-to-play programme. - Unique referring domains predict AI citations better than Google rankings: ten placements on ten different sites weigh more than ten on the same outlet. - 94% of brands cited in informational answers also appear in provider-choice answers, against 50% of the others: measure placements against that. Written by the Geosnap AI agent, reviewed and approved by Rinald Sefa. Translated from the Italian original. [Read the original](https://geosnap.ai/blog/uffici-stampa-e-testate-servono-ancora-per-farsi-citare-dall-ai-dati-2026) Anyone running communications for a brand has been asking the same question for months: does it still make sense to invest in PR, press releases and trade publications now that many people ask ChatGPT or Google AI Mode for advice instead of reading an article? Three studies released between 26 and 29 October 2026 give a measured answer: third-party sources are still behind more than 80% of AI citations, but journalism in the strict sense is losing share, paid editorial content is almost never cited, and variety of sources matters more than a single big name. Here are the numbers, the limits of each study and how to plan media work in 2026. ## How much do third-party sources weigh in AI answers? More than 84% of the links AI assistants cite come from sources the brand neither owns nor pays for. That is the headline figure of the fourth edition of Muck Rack's "What is AI Reading?" report, [published on 28 October 2026](https://www.globenewswire.com/news-release/2026/10/28/3390134/0/en/generative-pulse-earned-media-still-drives-more-than-84-of-ai-citations.html), built on more than 40 million AI responses and 207 million citations collected from January to September 2026, plus a September sample of 1,020 unbranded consumer prompts run on ChatGPT, Claude, Gemini and Perplexity. Across the four editions the share of third-party sources has ranged from 82% to 89%: it is a stable figure, not a spike. Inside that share, categories are shifting. Between May and September 2026 journalism fell from 27% to 22% of citations, while third-party corporate content grew (from 24% to 25.1%), as did aggregators and encyclopedic sites (from 17.4% to 18.3%), government and NGO sources (from 8.6% to 9.6%) and academic research (from 4% to 4.4%). Brands' own sites, the so-called owned content, went from 13.7% to 15.2%. The sharpest finding concerns advertising: paid and advertorial content accounts for 0.3% of citations, unchanged since May. About 99% of cited links come from non-paid sources. | Source type | May 2026 | September 2026 | | --- | --- | --- | | Journalism | 27% | 22% | | Third-party corporate content | 24% | 25.1% | | Aggregators and encyclopedic sites | 17.4% | 18.3% | | Brand sites (owned) | 13.7% | 15.2% | | Government and NGOs | 8.6% | 9.6% | | Academic research | 4% | 4.4% | | Paid and advertorial | 0.3% | 0.3% | One caveat: Muck Rack sells an AI citation monitoring tool and the report is a company release, not a peer-reviewed study. Definitions matter too: "third-party sources" includes journalism, encyclopedias, public bodies, universities and other companies' content, not just news outlets. ## Journalism is losing share: what changes for your PR team? The reason you seek coverage changes, not the fact that you seek it. The drop in journalism is the largest significant shift the authors record across four editions, and it shows up on every engine: on ChatGPT the share went from 28.1% in January to 22% in September, on Gemini from 27.6% to 23.6%, on Claude it moved between 28% and 24%. Several major publishers fell to near zero in ChatGPT citations. The same report notes that 81% of major publishers blocking Claude's search crawler receive almost no Claude citations, against 12% of those allowing it: part of the decline comes from publishers' own choices, not from the models. For communicators there are three practical consequences. First: an article on a publication that blocks AI crawlers serves human readers and Google, but will rarely make it into a ChatGPT answer. Before spending time on a placement, check the publication's robots.txt with the same method you use for your own site in our [guide to AI crawlers](https://geosnap.ai/en/blog/il-tuo-sito-blocca-chatgpt-guida-crawler-ai). Second: paid content is not a shortcut. Beyond the overall 0.3%, Muck Rack measured on one large business publication that its editorial articles are cited by ChatGPT about 1.7 times more often than content from the same outlet's pay-to-play programme. Third: with only 212 domains cited by all four engines, the same placement is worth different amounts on ChatGPT, Gemini and Perplexity, so coverage has to be measured engine by engine. ## Does one big outlet matter more than many different sources? Many different sources. A study by the Australian agency Earned Media, [summarised by its founder on 29 October 2026](https://www.startupdaily.net/advice/business-strategy/startups-dont-need-to-win-google-to-win-ai-search-but-they-need-to-do-this/), analysed 360 answers from ChatGPT, Gemini, Claude and Perplexity in Australia, covering more than 3,000 citation events and 137 domains. The number of unique referring domains linking to a site predicts AI citations better than both Google rankings and Ahrefs' Domain Rating, with less than a 0.1% chance that the relationship is a coincidence; a similar test on 141 US domains showed the same pattern. The share of keywords on Google's first page, by contrast, showed almost no relationship with citations in the Australian sample. The distinction matters when you plan placements. Ten articles on the same outlet count as ten backlinks but one domain; ten articles on ten different sites, even small ones, are ten domains. And AI citations are not concentrated only on big names: in the study the top 10% most-cited domains captured only around a quarter of total citations, with the rest spread across a long tail of specialist sites. Caution applies here too: Earned Media sells link building, the sample is small and the authors themselves stress that this is correlation, not causation. Still, the picture matches what [Google states in its documentation](https://developers.google.com/search/docs/appearance/ai-features) on AI Overviews and AI Mode: there are no additional requirements or special optimisations to appear, the usual SEO best practices apply, and one of them is being cited by diverse, trusted sites. ## Why does being cited in guides lead to being recommended? Because AI treats the research stage as a shortlist. The Q3 2026 report by Victorious, a US agency, [covered by La Mescolanza on 26 October 2026](https://www.lamescolanza.com/2026/10/26/intelligenza-artificiale-i-marchi-citati-nelle-risposte-hanno-quasi-il-doppio-delle-possibilita-di-essere-consigliati/), tracked 161 brands in five sectors across about 16,500 answers from nine AI platforms. 94% of brands cited in informational answers also appear in answers where the user asks which provider to choose, against 50% of brands that get no citations at the research stage. Of the 40 brands absent from choice answers on 1 July, 81% of those already cited in informational answers had entered at least once by October, against 21% of the others. The same report shows sources concentrating: in legal services the ten domains the platforms use most collect 50.7% of citations, up from 40.6% the previous quarter; in healthcare the share is 41.3%, in financial services 32.5%. And turnover is harsh: of the 42 brands present in Q2, 15 disappeared in Q3 and only 2 entered. The authors too speak of correlation, not causation. But the lesson for PR is clear: the goal of a placement is no longer visibility in itself, it is getting into the pages AI reads when someone asks an informational question about your sector, because that is where the recommendation comes from. ## How to plan media work in 2026 Start from the sources AI already cites in your sector, not from a list of outlets ranked by circulation. The method is the same you use to [choose the prompts to monitor](https://geosnap.ai/en/blog/quali-prompt-monitorare-risposte-ai-come-scegliere): take the ten questions a customer would ask before choosing a provider like you, ask them on ChatGPT, Gemini and Google AI Mode and note the cited domains. In Geosnap you get this list for every prompt and every engine, with the sources where your brand appears and those where it is missing. That list is your media plan. 1. **Prefer variety to size.** Ten placements on ten different sites, including vertical publications, consultants' blogs and trade associations, weigh more than ten articles in the same national daily. A travel agency in Manchester gets more from five industry portals and two associations than from a single national placement. 2. **Bring a data point, not an announcement.** AI citations reward pages that answer a question. A release with original research, a comparison or a market figure gets picked up and cited; a note about a new logo does not. 3. **Get your name written next to the data.** If the article carries your figure without your name, AI may cite the outlet and forget you: that is the problem of [ghost citations](https://geosnap.ai/en/blog/citazioni-fantasma-ai-cita-il-sito-ma-non-nomina-il-brand-cosa-fare). 4. **Do not buy advertorials for AI.** The 0.3% share of paid citations says that budget earns more on well-structured owned content or on research you can offer to publications. 5. **Check that the outlet is readable by AI crawlers.** A placement on a site that blocks GPTBot or Claude's crawler will not enter those answers: choose where it pays off. 6. **Measure engine by engine.** After each round of placements, repeat the same questions and check whether the domains you won appear among the sources and whether the brand enters choice answers. The brand's own site still matters, as [the SISTRIX study on cited pages](https://geosnap.ai/en/blog/quali-pagine-cita-ai-studio-sistrix-96000-url-cosa-conta-davvero) shows, but on its own it is not enough. PR work is not over: it has become measurable. For the first time you can see whether a placement entered the sources AI draws from when someone asks for advice in your sector, and plan the next ones accordingly. ## Sources - [Generative Pulse: Earned Media Still Drives More Than 84% of AI Citations](https://www.globenewswire.com/news-release/2026/10/28/3390134/0/en/generative-pulse-earned-media-still-drives-more-than-84-of-ai-citations.html) (Muck Rack (via GlobeNewswire), 2026-10-28) - [Startups don't need to win Google to win AI search, but they need to do this](https://www.startupdaily.net/advice/business-strategy/startups-dont-need-to-win-google-to-win-ai-search-but-they-need-to-do-this/) (Startup Daily (Earned Media), 2026-10-29) - [Intelligenza artificiale, i marchi citati nelle risposte hanno quasi il doppio delle possibilità di essere consigliati](https://www.lamescolanza.com/2026/10/26/intelligenza-artificiale-i-marchi-citati-nelle-risposte-hanno-quasi-il-doppio-delle-possibilita-di-essere-consigliati/) (La Mescolanza (Victorious Q3 2026 Search Report), 2026-10-26) - [AI features and your website](https://developers.google.com/search/docs/appearance/ai-features) (Google Search Central, 2025-12-10) ## Frequently asked questions ### Do press releases still matter for visibility in AI answers? Yes, if they carry something a publication will print and that answers a question: a figure, a piece of research, a comparison. More than 84% of AI citations come from third-party sources, but AI cites the resulting article, not the release itself. An announcement with no substance does not get picked up and does not enter the answers. ### Is it worth paying for an advertorial or sponsored article to get cited by AI? The data says no. In the Muck Rack report, paid and advertorial content is 0.3% of citations, unchanged since May 2026, and on one large business publication editorial content was cited by ChatGPT 1.7 times more often than the pay-to-play programme. That budget earns more on original research you can offer to publications. ### Is one placement in a national daily better than ten on industry sites? For AI citations, ten on different sites. The Earned Media study on 137 Australian and 141 US domains finds that the number of unique domains linking to a site predicts citations better than Google rankings and Domain Rating. The national daily is still useful for audiences and for Google, but it counts as a single domain. ### How do I know which publications to target? Ask the AI engines the questions a customer would ask in your sector and note the cited domains: those are the sources they draw from. In Geosnap this list is available for every prompt on ChatGPT, Gemini and Google AI Mode. Then check that the outlet does not block AI crawlers in its robots.txt, otherwise the placement will not enter those answers. ### Why is journalism losing share in AI citations? The Muck Rack report measures the drop from 27% to 22% but does not establish the cause. Part of it comes from publishers blocking crawlers: 81% of major publishers blocking Claude's search crawler receive almost no Claude citations, against 12% of those allowing it. Meanwhile encyclopedias, public bodies, third-party corporate content and brand sites are growing.
Structured data
Schema.org data (JSON-LD) embedded in the page code. It tells the AI explicitly what the page is about and who published it.
BreadcrumbList describes the path through the site
{
"@context": "https://schema.org",
"@type": "BreadcrumbList",
"itemListElement": [
{
"@type": "ListItem",
"position": 1,
"name": "Home",
"item": "https://geosnap.ai/en/"
},
{
"@type": "ListItem",
"position": 2,
"name": "Blog",
"item": "https://geosnap.ai/en/blog/"
},
{
"@type": "ListItem",
"position": 3,
"name": "GEO strategy",
"item": "https://geosnap.ai/en/blog/#strategia-geo"
},
{
"@type": "ListItem",
"position": 4,
"name": "Do PR and trade publications still matter for AI citations? The 2026 data",
"item": "https://geosnap.ai/en/blog/uffici-stampa-e-testate-servono-ancora-per-farsi-citare-dall-ai-dati-2026"
}
]
}
BlogPosting describes the article: author, dates and sources
{
"@context": "https://schema.org",
"@type": "BlogPosting",
"@id": "https://geosnap.ai/en/blog/uffici-stampa-e-testate-servono-ancora-per-farsi-citare-dall-ai-dati-2026#article",
"headline": "Do PR and trade publications still matter for AI citations? The 2026 data",
"description": "Over 84% of AI citations come from third-party sources, yet journalism's share fell to 22% and advertorials are at 0.3%. How to plan media placements in 2026.",
"inLanguage": "en",
"url": "https://geosnap.ai/en/blog/uffici-stampa-e-testate-servono-ancora-per-farsi-citare-dall-ai-dati-2026",
"mainEntityOfPage": "https://geosnap.ai/en/blog/uffici-stampa-e-testate-servono-ancora-per-farsi-citare-dall-ai-dati-2026",
"image": "https://geosnap.ai/img/blog/p2026-11-02-uffici-stampa-testate-citazioni-ai-en.webp",
"articleSection": "GEO strategy",
"wordCount": 1551,
"author": {
"@type": "Person",
"name": "Rinald Sefa",
"jobTitle": "CMO",
"url": "https://geosnap.ai/en/about",
"worksFor": {
"@id": "https://geosnap.ai/#organization"
}
},
"publisher": {
"@id": "https://geosnap.ai/#organization"
},
"isAccessibleForFree": true,
"abstract": "Yes, but differently. According to Muck Rack (28 October 2026), more than 84% of the links cited by ChatGPT, Claude, Gemini and Perplexity come from sources the brand neither owns nor pays for, while journalism's share dropped from 27% to 22% in four months and paid content stays at 0.3%. Two more late-October studies show that the number of different sites citing you matters more than your Google ranking, and that brands cited in informational answers are almost twice as likely to be recommended. PR still works: it has to target the sources AI reads, not circulation.",
"datePublished": "2026-11-02",
"dateModified": "2026-11-02",
"citation": [
"https://www.globenewswire.com/news-release/2026/10/28/3390134/0/en/generative-pulse-earned-media-still-drives-more-than-84-of-ai-citations.html",
"https://www.startupdaily.net/advice/business-strategy/startups-dont-need-to-win-google-to-win-ai-search-but-they-need-to-do-this/",
"https://www.lamescolanza.com/2026/10/26/intelligenza-artificiale-i-marchi-citati-nelle-risposte-hanno-quasi-il-doppio-delle-possibilita-di-essere-consigliati/",
"https://developers.google.com/search/docs/appearance/ai-features"
],
"translationOfWork": {
"@id": "https://geosnap.ai/blog/uffici-stampa-e-testate-servono-ancora-per-farsi-citare-dall-ai-dati-2026#article"
}
}
FAQPage describes the frequently asked questions
{
"@context": "https://schema.org",
"@type": "FAQPage",
"@id": "https://geosnap.ai/en/blog/uffici-stampa-e-testate-servono-ancora-per-farsi-citare-dall-ai-dati-2026#faq",
"mainEntity": [
{
"@type": "Question",
"name": "Do press releases still matter for visibility in AI answers?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Yes, if they carry something a publication will print and that answers a question: a figure, a piece of research, a comparison. More than 84% of AI citations come from third-party sources, but AI cites the resulting article, not the release itself. An announcement with no substance does not get picked up and does not enter the answers."
}
},
{
"@type": "Question",
"name": "Is it worth paying for an advertorial or sponsored article to get cited by AI?",
"acceptedAnswer": {
"@type": "Answer",
"text": "The data says no. In the Muck Rack report, paid and advertorial content is 0.3% of citations, unchanged since May 2026, and on one large business publication editorial content was cited by ChatGPT 1.7 times more often than the pay-to-play programme. That budget earns more on original research you can offer to publications."
}
},
{
"@type": "Question",
"name": "Is one placement in a national daily better than ten on industry sites?",
"acceptedAnswer": {
"@type": "Answer",
"text": "For AI citations, ten on different sites. The Earned Media study on 137 Australian and 141 US domains finds that the number of unique domains linking to a site predicts citations better than Google rankings and Domain Rating. The national daily is still useful for audiences and for Google, but it counts as a single domain."
}
},
{
"@type": "Question",
"name": "How do I know which publications to target?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Ask the AI engines the questions a customer would ask in your sector and note the cited domains: those are the sources they draw from. In Geosnap this list is available for every prompt on ChatGPT, Gemini and Google AI Mode. Then check that the outlet does not block AI crawlers in its robots.txt, otherwise the placement will not enter those answers."
}
},
{
"@type": "Question",
"name": "Why is journalism losing share in AI citations?",
"acceptedAnswer": {
"@type": "Answer",
"text": "The Muck Rack report measures the drop from 27% to 22% but does not establish the cause. Part of it comes from publishers blocking crawlers: 81% of major publishers blocking Claude's search crawler receive almost no Claude citations, against 12% of those allowing it. Meanwhile encyclopedias, public bodies, third-party corporate content and brand sites are growing."
}
}
]
}
Metadata
The information in the page header, which crawlers and engines read before the content.