How Do You Measure the Impact of PR on AI Visibility?

Why does PR affect AI visibility?

Several studies point the same way:

• Earned sources dominate citations. Muck Rack's analysis of more than 25 million links cited in AI answers found about 84% came from earned sources and 0.3% from paid or advertorial content. Muck Rack's definition of earned is broad, covering journalism, academic and government sources, reference sites, social platforms and third-party company content. Journalism alone accounted for 27% of cited links.
• Question type matters. In the same research, questions about industry trends cited journalism more than twice as often as how-to questions.
• Mentions travel with visibility. Ahrefs' study of 75,000 brands found branded web mentions correlated strongly with visibility in AI Overviews, much more than backlinks did.

So PR can change AI answers in two ways. Articles can be cited directly when assistants search the web. And repeated coverage shapes what future models learn about the brand.

What should you measure?

Layer 1 is available within days. Layers 2 and 3 take weeks. Layer 4 takes a sales cycle.

How do you set up the measurement?

1. Define target questions before the campaign: 15–40 questions the campaign should affect, such as "Which companies lead in [topic]?" or "What's the latest research on [issue]?".
2. Add control questions on topics the campaign does not touch, to detect general shifts such as model updates.
3. Plant a marker. A campaign built on original data should have a distinctive figure or named finding, such as "the [Brand] 2026 Workplace Index found 41% of...". Its appearance in answers is clear evidence of pull-through.
4. Take a baseline: run target and control questions three times each in the assistants your audience uses, for two to four weeks before launch.
5. Map the citation landscape: list the outlets and sites currently cited for your target questions. These are your priority outlets.
6. Measure after launch at about 2, 6 and 12 weeks, the same way.

How do you read the results?

• Marker appears, citing campaign coverage: direct evidence the PR reached AI answers.
• Mention rate rises on target questions, not on controls: good evidence of a campaign effect.
• Mention rate rises on both: probably a general change, such as a model update or seasonal shift, rather than the campaign.
• Description changes: assistants start describing you with the campaign's positioning, such as "a leader in workplace research". That is a meaningful result even if the mention rate barely moves.
• Nothing changes after 12 weeks: check layer 1. Coverage in outlets that assistants never cite for your topics will rarely show up.

The approach mirrors a difference-in-differences test, comparing change on target questions with change on controls. Test design is covered in more depth in how to test whether GEO changes worked.

How do you choose outlets that matter for AI?

Before pitching, check which outlets AI answers cite for your target questions. They are often trade publications, specialist sites and reference sources, not only the biggest national titles. Coverage in a trade outlet that assistants cite for your category may matter more than a single national mention. Finding them is covered in how to find high-frequency AI-cited sources for outreach.

Which kinds of PR tend to show up in AI answers?

Some formats travel into AI answers better than others:

• Original data and research, because assistants cite specific figures and name who produced them.
• Expert commentary in trend pieces, since trend questions draw heavily on journalism.
• Inclusion in roundups and "best of" lists in trade outlets, which answer recommendation questions directly.
• Awards and rankings from recognised bodies, which assistants repeat as evidence.
• Profiles in reference sources, where the brand meets the source's own standards.

Announcements with no new information, such as a routine hire or a partnership with few details, rarely appear. Plan campaigns around something an answer would want to repeat.

What should a PR and AI visibility report include?

• placements, with the share in AI-cited outlets;
• citations of campaign coverage in AI answers, with examples;
• mention rate and position on target questions against baseline, alongside controls;
• how assistants now describe the brand, with before and after excerpts;
• marker pull-through: how often the campaign's figure or finding appears, and with which source;
• branded search trend, AI referral sessions and self-reported attribution for the period;
• method notes: engines, runs, dates and any known model changes.

What are the limits?

• Lag. Answers based on web search can change within weeks; answers drawn from a model's training reflect coverage only after later model releases.
• Volatility. Single runs vary, so use repeated runs and ranges.
• Overlapping activity. Content, SEO and product launches also affect answers. Controls help, but cannot remove every overlap.
• Paid placements are not the same. Sponsored and advertorial content made up a tiny share of citations in Muck Rack's analysis, so paid "coverage" should be measured separately.

What might an illustrative campaign show?

Consider an illustrative HR software company that publishes a survey on four-day working weeks, with a headline finding that 41% of mid-sized firms are piloting one. Coverage lands in six outlets, four of which the baseline showed were already cited for workplace trend questions. Six weeks later, the 41% finding appears in about a third of runs for "What are the trends in four-day work weeks?". Most of those runs cite two of the trade articles. The company's mention rate on its 20 target questions rises from 9% across three baseline weeks to 21% across three post-launch weeks, about 180 answers each, while control questions stay flat. With only one measurement of about 60 answers on each side, a change of less than about 13 points could be chance. Branded searches rise modestly over the same period. The report presents the marker pull-through as the clearest evidence, and the mention rate as supporting evidence.

What do people ask about PR and AI visibility?

Does a press release help AI visibility?

A release on its own is rarely cited. Coverage that journalists write from it, and original data it announces, are what assistants use.

How long before PR shows up in AI answers?

Weeks for answers that search the web; longer for answers drawn from model knowledge.

Should PR teams own AI visibility measurement?

They should share it with SEO or content teams, who usually run the question tracking. PR brings the outlet relationships and messages.

What is the single best indicator of PR impact?

A distinctive finding or phrase from the campaign appearing in AI answers, with campaign coverage as the cited source.