The same study defines "earned" broadly. It counts journalism, academic research, government sources, encyclopedic sites and third-party corporate content, and journalism alone accounts for 27% of all cited links, spread across more than 20,000 outlets. So the coverage that feeds AI answers is not limited to national press. Trade titles, local outlets and independent publishers all supply citations.
The mechanism is retrieval. An AI tool with web search looks up pages, reads them and cites some. Sources that many independent publishers repeat are more likely to be retrieved than a single page a company wrote about itself, which is why the company's own site works best as the place where reporters confirm facts, not as the only source.
Public opinion carries the same logic with people. Buyers, employees and investors give the benefit of the doubt to a company they have read well of, and that credit is what a steady run of positive stories builds. EPR's corporate reputation management guide covers how that credit is built, spent and rebuilt.
What kinds of positive publicity can a company earn?
A company can earn positive publicity from five kinds of source, and each one reaches readers and AI systems differently. Muck Rack's May 2026 study groups earned citations into journalism, academic research, government sources, encyclopedic sites and third-party corporate content, which gives a useful map of where favorable mentions come from.
| Source type | What a favorable mention looks like | What it adds |
|---|
| Journalism | A profile, feature, product review or trend story that names the company | Independent verification, and the largest share of citations at 27% |
| Academic research | A study that uses the company's data or product, or a researcher who cites its work | Credibility with analysts, buyers and other writers |
| Government and institutional sources | A grant, a contract award, a regulator's published finding or a public-agency partnership | A record that no one at the company wrote |
| Encyclopedic sites | An entry that summarizes the company with references to published coverage | A neutral summary that AI tools retrieve often |
| Third-party corporate content | A partner's case study, an industry association's member profile or an analyst note | A second company vouching in its own words |
The table shows why a single press-release distribution is not a positive publicity program. A distribution service sends the company's own words to outlets. Each row above is a third party deciding to write something, and that decision is the part a company has to earn.
It also shows where small companies can start. A partner case study or an association profile asks for a smaller favor than a national feature, and it still adds a named outside source that can be cited later.
What makes a company's news worth positive coverage?
A company's news earns positive coverage when it contains an action a reporter can verify, with numbers and named parties, rather than a claim about the company's values. The test is whether an outsider can check it without calling the company.
Patagonia's 2022 ownership announcement is a clean example. On September 14, 2022, Patagonia told employees at a global town hall that the Chouinard family had transferred ownership of the company, then posted a letter from founder Yvon Chouinard on its website. According to Patagonia's ownership page, the Patagonia Purpose Trust now holds 2% of the stock, all of it voting. The Holdfast Collective, a 501(c)(4) nonprofit, holds the other 98%, all nonvoting, and receives each year's profits left after reinvestment.
Every element of that announcement can be checked: the date, the two entity names, the percentages, the voting split and the dividend rule. Chouinard, Patagonia's founder and former owner, wrote that instead of going public, the company was "going purpose." The phrase gave reporters a headline, and the structure behind it gave them facts to confirm.
The announcement also answered the questions a reporter would otherwise have to ask. Patagonia's page states who runs the business (CEO Ryan Gellert and the leadership team), that the company remains a for-profit certified B Corporation, and that it will keep donating 1% of sales. A pitch that answers the obvious follow-ups lets a reporter publish accurately in one pass.
Before pitching, test any piece of news against three questions:
- Does it change something a reporter can measure, such as money, ownership, headcount, a product or a policy?
- Can an outsider confirm it from a filing, a dataset, a named person or a public page?
- Does it matter to people beyond the company's own customers?
News that fails these tests belongs on the company's own channels, not in a pitch.
How do you generate positive publicity, step by step?
A positive publicity program runs in six steps: choose the fact, choose the audience, build the pitch, time it, supply proof and record the result. Getting coverage and building an image takes planning, strategy and sustained execution. The internet made research and reach faster, but the core moves are the same.
| Step | Action | Output |
|---|
| 1. Choose the fact | List what changed in the company in the last 90 days: ownership, product, hiring, data, a partnership or a policy | One lead fact with a number and a date |
| 2. Choose the audience | Pick the journalists, creators and institutions that covered a similar change in the last 12 months | A target list with the story each person last wrote |
| 3. Build the pitch | Open with the fact, attach the proof, name who can speak on the record | A pitch under 200 words that links to a page holding the data |
| 4. Time it | Pair the fact with a news event, report or season already in the coverage calendar | A send date, with an embargo if needed |
| 5. Supply proof | Publish a page with the figures, quotes, photos and a named spokesperson | An owned page a reporter can cite |
| 6. Record the result | Log each placement, the angle the outlet took and the follow-up questions | A tracker that feeds the next pitch |
Step one decides most of the outcome, because a weak fact cannot be rescued by a good pitch. The public relations plan guide and the PR plan template show how to schedule these steps across a year, and launch public relations covers the case where the fact is a new product.
Illustrative scenario: a regional bakery chain with 12 stores wants coverage when it starts paying every baker a set hourly wage above the local minimum. Step one makes the fact specific: the wage figure, the number of bakers covered and the start date. Step two builds a list of the local business reporter, a food and drink editor and the state restaurant association's newsletter, each of whom wrote about wages or bakeries in the past year.
Step three puts the wage figure in the first sentence of the pitch, names the owner as the spokesperson and links to a page with the payroll numbers. Step four ties the send to a local minimum-wage vote already on the calendar. Step five is that page. Step six is a tracker that records which outlet ran which angle and what each reporter asked next. None of the numbers here are real, and the point is the order of the work: fact first, proof attached, timing chosen, result logged.
Which kind of story do AI systems cite most often?
Industry trend stories draw journalism citations at more than double the rate of how-to questions, according to Muck Rack's May 2026 report. The same report found that press releases appear almost exclusively in industry trend responses, at a rate 3.5 times higher than in best-of queries.
The practical reading is that a trend story with the company's own data attached has a better chance of being cited than a how-to article that restates common advice. Muck Rack's December 2025 analysis added a format detail: it reported that the majority of cited press releases carried twice as many statistics and 2.5 times as many bullet points on average, plus a higher share of objective sentences. A release that reads like a dataset is easier to quote than one that reads like a slogan.
Who should receive the pitch?
A company should pitch the reporters AI systems already cite for its category, not only the reporters it already knows. Muck Rack's December 2025 report found that the overlap between the journalists most pitched on its platform and those most cited by AI engines for a given brand was only 2% on average.
The fix is a lookup, not a guess. Ask ChatGPT, Claude and Gemini the questions a buyer would ask about the category, list the cited articles, and note the bylines. Those reporters have already written the stories that AI tools retrieve, so a pitch with a fact their last article lacked gives them a reason to update the record. For the wider playbook, see how to get press coverage.
Why do specific stories outperform generic advice?
Specific stories outperform generic advice because they give a reader, a reporter and a retrieval system something concrete to match. EPR's own Search Console data from June 2025 to October 2026 shows the gap. A page built around named publicity stunts earned 201 clicks from about 12,800 impressions at an average position near 15. A generic page of tips on getting free publicity earned no clicks from about 2,200 impressions, at an average position near 65.
The same rule applies to pitches. "Our brand is passionate about sustainability" cannot be checked, so a reporter cannot print it. "Ownership moved to a trust and a nonprofit on September 14" can be checked, so a reporter can. Bold, specific moves belong to the publicity stunt playbook, and everyday consumer stories are covered in consumer PR tactics. Storytelling frameworks that hold up over time are collected in the marketing books that changed corporate America.
How often should a company earn new coverage?
A company should refresh each priority topic with new coverage at least every six months, because journalism citations fall sharply after that point. Muck Rack's May 2026 report found that more than half of journalism citations come from articles under 12 months old, and its December 2025 report found that about half of all citations come from articles published within the previous 11 months.
The planning rule that follows is simple. List the three or four topics the company wants to be known for, and check the date of the newest independent article on each one. Any topic whose newest article is older than six months goes to the top of the pitch calendar, with a new fact attached.
The same data argues against chasing daily news. Muck Rack's December 2025 report found that only about 4% of citations come from articles published in the prior week. Coverage that arrives steadily across the quarter does more for AI answers than a burst timed to a single day.
Which mistakes make positive publicity fail?
Positive publicity programs fail for five repeatable reasons, and each one has a direct fix. The first is pitching a claim instead of a fact. A reporter cannot verify "industry-leading service," so the pitch ends in the delete folder, while a dated figure with a named source gets a reply.
The second is relying on a single launch. Because journalism citations fall off sharply after six months, according to Muck Rack, a company that earns coverage once a year spends most of the year with aging sources. A quarterly run of smaller, factual stories keeps newer articles in circulation.
The third is pitching only familiar reporters. The 2% overlap that Muck Rack reported between pitched and AI-cited journalists means most teams are writing to people whose stories AI tools rarely retrieve. The fix is the lookup described earlier: ask the AI tools the buyer's questions and pitch the bylines they cite.
The fourth is leaving the proof off the page. A pitch that says "data available on request" costs the reporter an extra email, and many will not send it. Publish the figures, the method and a named contact on a page the pitch links to, so a story can be written in one sitting.
The fifth is never checking what the AI tools say. A company can earn a strong run of coverage and still be described with a two-year-old fact in AI answers, because the answer drew on an older article. The monthly log in the measurement section catches that gap while it is still cheap to fix.
What should a company do when publicity turns negative?
A company facing negative publicity should publish a documented fact, name one accountable spokesperson and announce a dated corrective step, because reporters and AI systems can quote all three. PRSA lists crisis communications and protecting an organization's reputation within the scope of public relations, which is why the same program that earns positive stories also has to answer negative ones.
| Move | What it contains | Why a reporter can use it |
|---|
| Documented fact | What happened, when, and how many people or dollars were affected | It can be checked and quoted without a second call |
| Named spokesperson | One executive with a title, available on the record | It gives the story a person who answers, not a statement from no one |
| Dated corrective step | What the company will change and by what date | It lets later coverage report whether the step happened |
Silence leaves the story to whoever speaks. A company that has already built a record of verifiable positive coverage has an advantage here, because reporters already hold accurate facts about it and have less reason to rely on a single critic's account.
EPR's guide to five reputation management cases documents Tylenol in 1982, Domino's in 2009, Old Spice in 2010, Patagonia and NVIDIA. The operating steps live in the online reputation management guide, which includes a quick-reference card for the first hours of a response.
How do you measure positive publicity in press and in AI answers?
Positive publicity is measured on three layers: the coverage earned, the sources AI systems cite and what AI answers say about the company. A count of clippings alone misses the second and third layers, which is where buyers now ask their first question.
| Layer | How to check | What it shows |
|---|
| Coverage earned | Count placements per quarter by outlet, date and angle | Whether the program produces a steady run of stories |
| Sources AI systems cite | Ask the same 20 buyer questions each month in ChatGPT, Claude and Gemini and log every cited URL | Which outlets and pages supply the answers |
| What the answers say | Record whether the company is named, in what position and with what description | Whether coverage became a favorable description |
Each platform needs its own tally, because they behave differently. According to Muck Rack's May 2026 report, ChatGPT cites sources in 96% of responses but averages five citations. Gemini cites in 82% of responses with an average of eight, and Claude cites in 55% of responses but averages 13 when it does. A company can be absent from one tool's five links and present in another's thirteen.
Read the log for gaps, not totals. If ChatGPT cites a competitor's profile when a user asks which companies lead the category, find the article, find the reporter and pitch a fact that article lacks. If the company is named but described with outdated facts, the cited page is probably old, and the fix is fresh coverage rather than a rewrite of the company's own site.
Illustrative scenario: a software company runs its 20 buyer questions in October and finds that ChatGPT cites a trade-press comparison from the previous year, names the company fourth in a list of five and describes its pricing with an outdated model. The log now holds a specific cause (an old comparison), a specific outlet and a specific reporter. The next pitch is a pricing update with the new figures, sent to that reporter with a link to the company's pricing page. In November the same 20 questions are run again and the log shows whether the cited article, the company's position or the description changed.
Treat every number as a baseline. Muck Rack states that citation behavior may shift as models are updated or retrained, so the monthly log matters more than any single study. The full methodology is on the What Is AI Reading? report page.
