How earned media affects your brand’s AI visibility

Earned media can influence what AI says about your business, even when the person asking never reads the original story. Reporting, reviews and expert commentary can supply information used in an answer about your category, your products or your competitors.

For a marketing team, that adds another question to the coverage report: is this story showing up in the answers your prospective customers see, and what are those answers saying about you?

A placement alone cannot answer that. The article might be cited without your brand being mentioned. Your business might appear without being recommended. Coverage can also surface criticism or outdated information. You need to inspect the answer to understand what the placement contributed.

What the citation research actually shows

In its May 2026 research, Muck Rack analyzed more than 25 million citation links from ChatGPT, Claude and Gemini. It classified 84% as earned media, with journalism accounting for 27% of all citations.

Those numbers need their context. Muck Rack’s methodology focuses on unbranded discovery questions: people exploring a problem or a category, rather than asking about a named business or completing a transaction. Its earned-media category is broader than news coverage.

Profound’s July 2026 analysis examined 11.84 billion citations across eight AI platforms and found that company-operated websites accounted for 57%. That category includes other companies’ websites, not just the website belonging to the brand being discussed.

The studies use different prompts, platforms and source classifications. You cannot combine their percentages into a universal rule for how much to spend on PR or your website. They give you reasons to investigate both, then check what appears for the questions relevant to your business.

Build your media list around customer questions

Start with questions your sales team already hears. A buyer asking “What does this agency offer?” is doing something different from a buyer asking “How can a regional retailer advertise on streaming TV?” The first names a business. The second leaves AI to decide which information and companies belong in the answer.

Run a manageable set of those questions, save the responses, and open the cited pages. Look for publications and specific articles that appear repeatedly. Check whether they discuss your competitors, the problem your customer is trying to solve, or the criteria used to compare providers.

Scrunch’s partnership with Stacker describes a similar workflow: identify gaps in AI answers and their sources, then use that information to guide earned-media opportunities. It gives your team a more specific starting point than pitching every publication with a high domain authority.

You still need a reason for an editor to cover the story and a publication whose audience fits your business. Frequent AI citations alone do not make a publisher a good target.

Give the coverage something specific to establish

A story about your company’s growth may be useful publicity without answering a buyer’s question about whether your product fits their needs. Before pitching, decide what the story could help a reader understand:

• The customer and problem your business serves.
• A capability, result or limitation that affects the buying decision.
• The evidence behind the claim.
• The correct company and product names.

Original research, a documented customer example or an expert contribution can give a journalist useful material. Include enough detail to assess it. For a survey, that means the sample, field dates and questions. For a customer result, explain the baseline and what changed. A number with no method leaves the reader guessing.

Keep press releases, distributed copies and independent coverage separate in your reporting. Meltwater’s April–May 2026 citation report classified earned media and news as 37.6% of its May citations, while press releases accounted for 0.2%. Those are findings from its sample, not a forecast for your next campaign. They also show why distributing a release and earning editorial coverage should not be counted as the same activity.

What an actual offsite program looked like

AirOps’ September 2026 case study with Vibe, a CTV advertising platform, describes an offsite program built around the sources appearing for priority topics. The team researched gaps and worked on publisher outreach rather than relying entirely on changes to Vibe’s own website.

AirOps reported a 13.3% relative lift in Vibe’s mention rate for the focus topic compared with the prelaunch period. It also reported that the gap with the leading competitor narrowed from 22 percentage points to 11.

This is a vendor-published before-and-after case study, not a controlled test proving how much of the change came from earned media. It is useful as a description of the work: research the sources, find gaps, provide relevant evidence and measure the resulting answers. It is not a result you should put into your forecast.

If your business sells connected TV advertising, for example, investigate the sources used in answers about buying, targeting and measurement. A broad company announcement may contribute less to those answers than a reported article explaining a relevant capability or customer problem.

Measure coverage, mentions and recommendations separately

Your report should make it possible to see what changed, without treating every appearance as a recommendation.

What to check	What it tells you
Source citation	Whether the answer cites the article or publication.
Brand mention	Whether your business appears in the answer.
Description accuracy	Whether your services, products and claims are described correctly.
Recommendation	Whether the answer suggests your business for the reader’s needs, and with what qualifications.
Website response	Whether identifiable visits or inquiries follow, where tracking makes that connection possible.

Record a baseline before the outreach begins. Repeat the same prompts on the same platforms, keeping language, location and other test conditions consistent where possible. Maintain a log of new coverage and website changes so you can interpret the results.

If your business appears in 12 of 40 tested answers, that is a 30% mention rate within that sample. It does not mean 30% of everyone using AI sees your brand. Repeat the checks on multiple dates and read the answers rather than relying only on a dashboard score.

Where an article is cited, check what information the answer actually uses. A citation beside a category explanation is different from a citation supporting a recommendation of your company. Neither, by itself, proves incremental sales.

Track identifiable AI referrals and qualified inquiries, while accepting that attribution will be incomplete. Someone may read an answer and visit your website directly later. That possibility is not a reason to assign every direct visit or new lead to AI visibility.

Keep your own website in the plan

Third-party coverage needs accurate information to work from, and an interested buyer needs somewhere to go next. Keep your service descriptions, product information and relevant pricing or availability details current. Make the next step clear for someone who arrives after researching your business elsewhere.

Also separate organic visibility from advertising. In Floodlight’s research on ChatGPT recommendations and advertisers, we examined organic answers and ad placements as distinct results. That study did not test whether buying ads causes more organic recommendations.

For your earned-media program, the useful question is whether relevant coverage helps your business appear accurately in the answers your buyers use. Work from those questions, pursue stories with something substantive to contribute, and inspect the answers after coverage appears.

Work with Floodlight

Floodlight provides AI research and visibility alongside campaigns across ChatGPT advertising, connected TV, streaming audio and podcasts, digital out-of-home and display advertising. We also handle programmatic media planning and buying. Talk with us about the questions your customers ask and how your brand appears in the answers.