A new question has crept into launch planning. It used to be "will they cover us?" Now it is "when someone asks ChatGPT what to buy, will it mention us?"
It's a fair question. The answer is less mysterious than the people selling "AI visibility" packages would like.
The machines are reading the press
Muck Rack has been tracking which links AI assistants cite since mid-2025. Its May 2026 edition analyzed more than 25 million links cited by ChatGPT, Claude and Gemini. About 84% came from what it classes as earned media, and about 27% were journalism. Paid and advertorial content made up 0.3%. Press releases were 1.1%.
Read the definitions before you quote the 84%. Muck Rack's "earned" bucket includes Wikipedia, academic papers, government sites, social platforms and other companies' content. So it is not saying 84% of answers come from reporters. The number that matters for PR is the 27%, and the 0.3%. Journalism is a big slice of what these systems read. Ads barely register.
The same study found that questions about industry trends pull in journalism citations at more than twice the rate of how-to questions. When people ask an assistant what's happening in a category, it goes looking for reporting.
A smaller slice, and a pickier one
Here's the less comfortable part. A Resonance study published in September 2026 looked at 1.3 million citations in answers to buyer-style questions about 116 B2B technology brands. Media titles were only about 4% of citations. Company websites made up most of the rest.
So coverage is not the whole answer. But look at which coverage gets used:
- On buying questions (shortlists, comparisons, vendor evaluations), about three in four media citations came from trade and specialist titles. The national press got about one in eight.
- On questions about a company's leaders and reputation, the national and business press supplied about two-thirds of media citations.
- ChatGPT and Gemini shared only six of their 20 most-cited media titles.
- A typical trade article was about nine months old when an assistant cited it. Tech news was cited at about 18 days.
That data is B2B, so treat it as a strong hint rather than a law for consumer products. The pattern is hard to ignore, though. The assistant wants the outlet that explains how products differ. That is usually the specialist writer who covers your category every week, not the national title that mentions you once a year.
Natan Edelsburg, Muck Rack's chief partnerships officer, told a PRSA session that, on average, only 2% of the journalists PR teams traditionally pitch are being cited by AI. Most media lists were built for a reader. They now have a second reader, and it has different taste.
Why the click matters less than the citation
Pew Research tracked the browsing of 900 US adults in March 2025. When Google showed an AI summary, people clicked a traditional result in 8% of visits, against 15% when there was no summary. They clicked a link inside the summary itself in 1% of visits.
Google disputes the methodology. Even so, if fewer people click through, the review can't do its job only by sending traffic. It has to do its job inside the answer. What the writer said about your product is now what the assistant says about your product.
That is the old earned-media argument with a new reader. Hype is borrowed. Earned media is yours to keep, and now it keeps getting quoted.
What to do about it
- Ask the assistants your buyers' questions. Do it before you build a media list. "Best wireless headphones for travel." "Is [your product] worth it?" Ask ChatGPT, Gemini and Perplexity, and write down which outlets they cite. That list will be longer, and more specialist, than your tier-one sheet.
- Move time toward the specialist writers. Audio reviewers, gear sites, category newsletters, the trade titles that cover your retail channel. They write the comparisons that assistants lean on, and that coverage keeps getting cited for months.
- Use national press for reputation. Profiles, founder stories and company news still matter, and the Resonance data suggests that's where assistants go when someone asks who you are.
- Give writers the detail. Assistants cite pieces that explain what a product does, what it works with and how it compares. A writer can only publish that if you supply specs, compatibility, pricing, availability and a person who can answer hard questions. A vague release produces a vague article, and the assistant will repeat it.
- Keep your facts the same everywhere. Your site, fact sheet, retailer listings and spokesperson should all say the same thing about price, ship date and specs. If they don't agree, the summary won't either.
- Don't pay for shortcuts. If paid and advertorial content is 0.3% of what assistants cite, a sponsored post is a poor way to get into the answer.
- Add one line to the coverage report. Keep reach and readership. Add a check of whether your coverage shows up in AI answers, for which questions, and on which assistant.
One story, two markets, more than one reader
The Resonance top-title lists include British outlets such as The Guardian, The Register, Computer Weekly and ITPro alongside US titles. If you sell in both markets, the assistants are reading both presses. A launch pitched only to New York will be missing part of what gets cited, so run the same story through London too, adapted for the desks there.
None of this is a new discipline. It's the old one, done properly: credible coverage, in the right outlets, with facts a writer can use. The audience just got bigger.
Want to know what the assistants say about your category? Request a briefing.