Earned Media Is What AI Reads. Most B2B Brands Stopped Earning It.

#marketing-leadership, Tuesday, 9:14 a.m.

CEO: Asked ChatGPT for the best tools in our category. We came up fourth. Behind a company that's half our size.

CEO: It linked to an Axios piece and some trade pub roundup. Why aren't we in those?

VP Marketing: We've published 140 blog posts this year. All optimized.

CEO: That's not what it linked to.

VP Marketing: ...We let the PR agency go in the 2024 budget cut. I'll pull a list.

I made that exchange up, but I'd bet some version of it has happened inside a lot of B2B companies this year. The details change. The ending usually doesn't: a marketing team with a big content library discovers that the AI tools their buyers use mostly ignore it, and quote journalists instead.

For CMOs, Content and Communications Leaders, and B2B Founders Deciding Where Next Year's Budget Goes

The number behind the Slack thread

In May 2026, Muck Rack published the third edition of its Generative Pulse research, which looks at what ChatGPT, Claude, and Gemini cite when they answer questions. The sample is big: more than 25 million cited links across 17 industries.

The headline result: 84% of AI citations came from earned media. Journalism alone made up 27% of all cited sources, spread across more than 20,000 distinct outlets. Paid and advertorial content accounted for 0.3%.

Read that last figure twice. Sponsored content barely registers.

Muck Rack's definition of "earned" is broad. Alongside journalism, it counts encyclopedic, academic, government, and social sources. Other studies use narrower buckets and get smaller numbers. Meltwater's analysis from spring 2026 put the earned and news share at 39.5%, and a research group that compared the major studies found the gap comes down to which engines were tested, how sources were categorized, and what counted as the denominator.

So I wouldn't put "84%" on a slide without the footnote. The direction holds across every study I looked at, though. AI answers lean heavily on what other people say about you, and lightly on what you say about yourself.

Why B2B teams should care more than most

B2B buyers have moved a big chunk of their early research into these tools. TrustRadius surveyed nearly 2,500 technology buyers and vendors in January 2026 for its B2B Buying Disconnect report. 63% of buyers used AI somewhere in their purchase journey.

They don't take the answers on faith, either. Of the buyers who used AI, 94% said they fact-check its responses at least some of the time. When they check, they go to sources they already trust: reviews, independent publications, peers.

Two more numbers from that same report explain why this gets expensive. 83% of buyers shortlisted three or fewer products. And only 13% used analyst reports, a 63% drop since 2022.

Put those together and the shape of the problem is clear:

• Buyers ask an AI tool for options early, before they ever talk to a vendor
• The AI builds its answer largely from third-party coverage
• The shortlist is short, so being fourth often means being out
• The old authority layer (analysts) matters less, which leaves the press and peers carrying more weight

Forrester's State of Business Buying 2026 adds one more wrinkle. A typical purchase involves 13 internal stakeholders and 9 external influencers. That's a lot of people, each running their own searches and their own prompts. Every one of them gets a slightly different answer, built from roughly the same pool of sources.

If you're not in that pool, you aren't in most of those answers.

The part nobody likes: coverage expires

For planning purposes, one finding in the Muck Rack research matters more than the headline. More than half of the journalism citations came from articles published within the previous 12 months, and citation volume dropped sharply once a piece was older than six months.

That changes how you should think about PR.

A lot of B2B companies treat press as a launch event. Big funding round, big product release, big analyst announcement. You get a burst of coverage, and then nothing for nine months. Under the old model that was fine, because a good article kept ranking in search for years.

AI tools seem to prefer fresh sources. A great feature story from 2024 is probably doing very little for you in a 2026 answer. Your competitor's modest trade-publication quote from last quarter might be doing more.

Earned media now behaves less like an asset and more like a subscription. If you stop paying in, the benefit fades within a couple of quarters.

Meanwhile, the newsroom got smaller

You might assume more brands chasing coverage means journalists are drowning in great pitches. They're drowning in pitches, all right. Most aren't great.

Cision's 2026 State of the Media report surveyed 1,899 journalists across 19 markets in January and February. A few results worth knowing:

• 66% rely on PR-provided content for story ideas
• 72% say fewer than a quarter of the pitches they receive are relevant
• 49% name shrinking budgets, staff cuts, and heavier workloads as obstacles
• 53% oppose AI-generated pitches, citing accuracy and personalization concerns

Muck Rack's own State of Journalism 2026 survey of 897 journalists points the same way. 86% say PR pitches inspire at least some of their stories. And 88% immediately throw out a pitch that misses their beat.

So the door is open. Reporters need sources, data, and story ideas, and they're working with fewer colleagues than they used to. But the bar for getting through is relevance, and generic outreach gets deleted in seconds.

There's an irony in the AI piece too. Muck Rack found 82% of journalists now use at least one AI tool, and Claude usage among them doubled in a year, from 6% to 12%. Reporters are using the same tools to research that your buyers are. They just don't want those tools writing the pitches they receive.

What a citation-grade PR program looks like

You don't need to rebuild your marketing org to fix this. You need to treat earned media as a steady input to how buyers (and their AI tools) understand your category. This is how I'd set it up.

1. Audit what the machines read about your category

Before you pitch anyone, find out where the answers come from today. Write 20 to 30 prompts a buyer in your category would plausibly use: "best [category] tools for mid-market," "[your brand] vs [competitor]," "how do companies solve [problem]." Run them in ChatGPT, Claude, Gemini, and Perplexity.

Log every cited domain in a spreadsheet. Note which outlets show up repeatedly and which of your competitors they mention.

Expect each platform to behave differently. In Muck Rack's data, ChatGPT included citations in 96% of responses with about five per answer, and its most-cited domain was Wikipedia. Gemini cited in 82% of responses, around eight per answer, with Reddit on top. Claude cited in 55% of responses but averaged 13 citations when it did, and its most-cited domain was PubMed Central. One audit won't cover all of them with the same list.

2. Build your media list from the citations

Most PR target lists start with prestige: the outlets the CEO reads on a plane. Start with the domains that actually show up in your audit instead.

Sometimes those overlap. Axios, for example, showed up in ChatGPT's top three cited domains in 13 of the 17 industries Muck Rack studied, the only journalism outlet to manage that on any platform. But in a lot of B2B niches, the outlets AI keeps citing are trade publications, specialist newsletters, and industry associations with modest audiences and very high relevance. Those are often easier to land, too.

3. Give reporters what they asked for

The Cision report is unusually specific about what journalists want from PR people: original research, access to experts, embargoed information, relevant and credible data, and assets they can use without extra work.

Notice what isn't on that list. Product announcements dressed up as news. Thought leadership that doesn't say anything. Quotes from executives that could have come from any company.

The most reliable thing a B2B company can offer is data the reporter can't get anywhere else. If you have product usage data, a customer base you can survey, or benchmark numbers from your platform, that's your raw material. One credible dataset can feed a year of pitches if you slice it by industry, company size, or region.

4. Set a cadence that matches the decay curve

If citations fade after about six months, a once-a-year news push leaves you invisible half the time. Plan for something pitchable every quarter at a minimum.

That doesn't mean four press releases. It might look like this:

• One original research release a year, broken into four or five follow-up angles
• A standing roster of two or three internal experts who are trained, available, and allowed to say something interesting
• Quarterly commentary on news in your space, offered to reporters already covering it
• Customer stories pitched to the trade outlets those customers read

The goal is a steady presence in recent coverage, which is what the AI tools seem to reward.

5. Pitch fewer people, better

With 88% of journalists discarding pitches outside their beat and most saying the bulk of what they get is irrelevant, volume is the wrong lever. Ten pitches to reporters who have covered your exact topic in the last three months will beat 300 sent to a purchased list.

I'd also be careful with AI-written outreach here. Over half of journalists in the Cision survey oppose AI-generated pitches. Use AI to research a reporter's recent work if you want. Write the pitch yourself, and keep it short.

6. Make your owned content the source behind the story

None of this means your blog is useless. It means its job changed. The useful role for owned content now is to be the thing a journalist links to: the methodology page for your research, the full dataset, the detailed explainer a reporter can cite instead of rewriting.

When a reporter quotes your survey and links to your methodology, you get two shots at being cited. The article itself, and the page it points to.

7. Measure share of citation

Traditional PR reporting counts placements and impressions. Add one more metric: how often your brand appears in AI answers for your audit prompts, and which sources those answers cite.

Rerun the same prompt set every quarter. Track your appearance rate, your position in the answer, and which of your placements get cited. It's imperfect, because the models change and the answers vary from run to run. It's still far closer to what your buyers see than a media-value number.

What not to do

A few tempting shortcuts are worth avoiding.

Buying placements is the first one. Paid and advertorial content made up 0.3% of citations in the Muck Rack data. Sponsored posts on news sites may still have a role in brand campaigns, but they won't do much for AI visibility.

Blasting press releases is the second. A wire release with no news in it doesn't become earned coverage just because it's hosted on a news-looking domain. Reporters ignore it, and the people you wanted to reach never see it.

The third is treating this as something you can game. The firms selling "AI visibility" services will keep finding tricks, and the models will keep closing them. The durable version of this work is boring: say something true and useful often enough that credible people repeat it.

The budget conversation

This is the uncomfortable part. A lot of B2B marketing teams cut PR first when budgets tightened, because it was hard to attribute. Content kept its budget because it was easier to measure in search traffic.

That trade looks different now. Most buyers have started doing at least some of their research inside AI tools, and those tools prefer exactly the kind of coverage many teams stopped paying for.

I'm not arguing for blowing up your content program. I'm arguing that a share of what goes into publishing your 141st blog post would probably do more as one solid piece of original research, a quarterly pitching rhythm, and an expert your PR lead can actually put in front of a reporter.

A note on the sources

Most of the citation research comes from companies that sell PR and media tools. Muck Rack, Cision, and Meltwater all benefit if you decide earned media matters. That doesn't make the data wrong, and the surveys of journalists are large and well documented. But the earned-media percentages vary a lot depending on who's counting and how. Use them as direction, not gospel, and run your own audit. Your category's numbers are the only ones that really matter for your budget.

Where to start this month

1. Write 25 buyer prompts for your category and run them across four AI tools. Log every citation.
2. Rank the outlets that show up most and compare them against your current media list.
3. Find one dataset you already own that a reporter in your space would find interesting.
4. Name two internal experts and block time on their calendars for press.
5. Schedule the prompt audit again for next quarter, so you can see whether any of this moved.

If your CEO sends you that Slack message next month, you'll at least have a list, and a plan for getting onto it.