How PR Agencies Can Adapt Their Services for AI Search

Why does PR matter to AI answers?

Muck Rack's continuing analysis of links cited in AI answers found that earned media accounted for 84% of citations in its May 2026 update. Paid and advertorial content accounted for 0.3%. Journalism alone was 27% of cited links. Muck Rack defines earned media broadly, including academic, government and third-party content as well as journalism. Even so, the finding is clear: in Muck Rack's prompts, assistants cited what others published about companies far more than anything companies paid for. For local questions the picture differs: Yext found most citations came from brands' own websites and listings.

Press releases themselves are a different matter. A study of cited sources for branded questions found press pages among the least cited content types. Reviews, listicles, forums and other social proof made up 57% of citations. So the value lies in the coverage and validation that releases lead to, not the release pages.

What changes in each service?

How should media targeting change?

Build a list of the outlets AI answers cite for each client's category and buyer questions. Run the questions several times across assistants, record every cited domain, and rank outlets by citation frequency. Add that to your usual criteria.

Then check access. A Reuters Institute study found that by the end of 2023, 48% of the most widely used news websites across ten countries were blocking OpenAI's crawlers, and 24% Google's AI crawler. Rules often differ between training crawlers and the crawlers assistants use for live search. So check each target outlet's robots.txt for the search crawlers too. A feature in a publication that assistants cannot read will still reach human readers. It is less likely to shape AI answers, and clients should know that before judging it.

How should releases and newsrooms change?

• One factual lead sentence that can stand alone: who, what, when, and a number if there is one.
• Dated figures with their source, so assistants can repeat them accurately.
• Consistent boilerplate, updated whenever the company changes, and the same on the site, releases and profiles. How outdated boilerplate spreads is covered in how to test whether AI assistants use your approved company boilerplate.
• HTML newsroom pages, not PDFs, with text that loads without scripts.
• Quotes that say something specific. "We expect demand to double" is quotable. "We're thrilled" is not.

How to make coverage easier for AI systems to read and quote is covered in how PR teams can make brand coverage easier for AI systems to crawl and quote.

What new services can agencies offer?

• AI visibility audits: what assistants say about the client, which sources they cite, and where competitors appear instead.
• Third-party validation programmes: pursuing inclusion in the reviews, rankings and roundups cited for the client's category.
• Expert visibility: building clear expert profiles and credentials that assistants can cite when asked who the experts are.
• Correction services: monitoring for wrong or harmful AI answers, tracing their sources, and running correction and reporting processes. How to diagnose where errors come from is covered in how reputation teams can separate source errors from model synthesis errors.
• Launch readiness: checking before a launch or rebrand that the new facts will be findable, and retesting afterwards.

What should agencies stop promising?

• Guaranteed placement in ChatGPT, Gemini or any assistant. No one controls answers.
• Suppression of negative answers. Accurate criticism generally cannot be removed on request, and no agency can make assistants stop repeating it. False claims can be corrected, but not on a schedule.
• Single-run screenshots as results. Answers vary between runs, so report rates across repeated runs.

Clients are increasingly checking these claims. The questions they ask are covered in how to audit a reputation agency's promise to suppress negative AI answers.

How should reporting change?

Keep coverage reporting, and add:

• Mention rate in a fixed set of category and buyer questions, per assistant, across repeated runs.
• Accuracy: whether answers describe the client correctly.
• Citations of coverage: which of your placements AI answers cite, and how often.
• Competitor comparison on the same questions.

Report these quarterly against a baseline. The link between specific coverage and cited sources is the evidence clients will value most.

What skills and tools does an agency need?

Less than many agencies fear, but more discipline than most have. Someone must own a fixed question set for each client and run it on a schedule. That means several runs per question, in several assistants, recorded the same way each time. The team also needs to read robots.txt files, check whether pages load without scripts, and tell a source error from a synthesis error. Monitoring tools can automate the runs, but they do not replace judgement about which questions matter to buyers. Work closely with the client's SEO and web teams. Many fixes, such as newsroom pages, boilerplate and profile updates, sit on the client's own site. An agency that can hand over a precise change list is more useful than one that only reports the problem.

What might a retained programme look like?

Consider an illustrative agency retained by a cybersecurity software client. In the first month, it ran 40 buyer questions across four assistants. The client appeared in 15% of answers, and the most cited outlets for the category were two trade publications, one analyst blog and a review platform. The client had no coverage in either publication. Over two quarters, the agency secured features in both and placed two expert bylines. It supported a review programme and rewrote the newsroom's boilerplate and release template. The client's mention rate rose to 32%, and the agency showed that 60% of answers mentioning the client cited one of the two features. Its reports now lead with that link.

What do people ask about PR and AI search?

Is PR becoming less important because of AI?

No. Earned media is one of the main sources AI answers draw on. What changes is how it is targeted and measured.

Should agencies charge separately for AI visibility work?

Audits and correction services are distinct pieces of work and are often priced separately. Targeting and reporting changes fit within existing retainers.

Do press release distribution services help?

Syndicated copies spread a release's text widely, which can help facts stay consistent. The coverage a release earns matters more to AI answers than the release itself.