Can a PR Agency Do GEO?
Yes, and the best GEO programs need one. LLMs build their answers mostly from what other people say about a brand: press, reviews, Reddit threads, YouTube videos and industry roundups. Earning those mentions is PR. What a PR agency has to add is the GEO layer: measuring the answers, knowing which sources each model trusts, and aiming the work at them.
Key takeaways
- A PR agency can do GEO well if it adds measurement: testing the answers, tracking which sources each model cites, and aiming the work at them.
- Across eight brands Wild Signal tracks, third-party sites made up 75% to 97% of the sources LLMs cited on non-branded questions; the brand's own website made up 0.1% to 11%.
- For every source a brand controls, LLMs cite anywhere from about eight to about a thousand that it doesn't.
- Coverage only helps AI visibility when it lands in sources the models actually cite, and those sources differ by model.
- Of the 30 roundup pages LLMs cite most for hiring an AI visibility agency, two thirds were written by agencies ranking themselves first.
Most of the market still treats GEO (generative engine optimization) as a branch of SEO. When you ask LLMs which agencies help brands show up in AI answers, the names that come back are mostly SEO firms. But the data on where AI answers come from points the other way, and that is the opening for PR.
Where do AI answers come from?
Wild Signal measured where LLM answers come from across eight brands we track, in travel, consumer tech, health, food and B2B services. We looked only at non-branded questions (ones that don't name the brand) from June 2026 onward, as a share of all sources cited.
| Where cited sources came from | Share of citations, by brand |
|---|---|
| The brand's own website | 0.1% to 11% (under 3% for six of the eight) |
| Social and community platforms | 2% to 14% |
| Other third-party sites: press, reviews, guides, directories and other companies' sites | 75% to 97% |
In other words, for every source a brand controls, LLMs cite anywhere from about eight to about a thousand that it doesn't. That is the territory PR has always worked in. The difference now is that the audience reading those sources includes the models that write the answers.
Why does GEO without PR stall?
You can make every page on your site cleaner, faster and easier to quote, and you should. But when someone asks an LLM which brand to choose, the answer leans on outside voices. An agency that only touches your site is working on about a tenth of what gets cited at most, and often far less.
We see the same pattern in our own category. When we looked at the agencies LLMs name most for "who should I hire for AI visibility," the leader wasn't winning through its own website. It appeared on roughly a dozen of the 30 roundup pages LLMs cite most for those questions, and when it was named, it was named first 84% of the time. Repetition across third-party lists made it the safe answer. That is an earned-media outcome, whether or not anyone called it PR. We wrote about the shift in how PR must evolve for AI discovery.
Why does PR without GEO miss?
The reverse is just as true. Traditional PR measures coverage by reach and impressions. But a hit in an outlet LLMs never cite does little for the answer, while a trade piece, a niche review site or a well-answered Reddit thread can matter more than a big headline.
The models also disagree with each other. Across the same eight brands, Google's AI Overviews cited YouTube in about half of the answers that cited any source, Perplexity in about one in seven, and ChatGPT in under 1%. A PR plan that ignores those differences will win in one model and stay invisible in another. Without measuring which sources each model uses, a PR team can win coverage and still be missing from AI answers. More in why different LLMs trust different sources.
What does a PR agency need to do GEO well?
- Measure the answers. Test the questions that matter in the category across ChatGPT, Claude, Gemini, Perplexity and Google, and track who is named, in what order, and which sources are cited.
- Build media lists from citation data. Rank outlets by how often models cite them for your category, not only by audience size.
- Shape the message for extraction. Use clear claims, specific facts and the same language everywhere, so a model can lift the point accurately and attribute it to you.
- Work community as part of earned. Reddit, YouTube and LinkedIn show up in answers, and leaders answering questions under their own names count. See why Reddit is the most cited thing on the web.
- Get on the lists. Many "best of" roundups that LLMs cite are open to pitches or submissions. Of the 30 we reviewed in our own category, two thirds were written by agencies ranking themselves first, so the independent ones are worth real effort.
- Plan by model. In our data, video matters most for Google's AI answers and barely registers in ChatGPT, which leans on written sources. Measure each model separately and plan for the gaps.
- Give it time. In Wild Signal's measurement, visibility from new content usually peaks about 12 weeks after it is published.
What should you ask a PR agency that says it does GEO?
Ask them to show you the AI answers in your category today and which of their past placements those answers cite. Ask how they choose outlets, and whether citation data is part of that choice. Ask how they will report results: share of answers that name you, how often you come first, and which sources carry you. If the answer is impressions and clip counts alone, they are doing PR next to GEO, not PR as GEO.
How did we measure this?
The source-mix and YouTube figures come from Wild Signal's own tracking of eight brands across travel, consumer tech, health, food and B2B services, using non-branded questions only, from June 2026 onward. Source-mix figures are a share of all sources cited, shown as a range across the brands; YouTube figures are a share of answers that cited at least one source. Claude answers are not included. The roundup figures come from Wild Signal's own non-branded prompt set about hiring for AI visibility, June to September 2026, and a review of the 30 roundup pages LLMs cited most for those questions in November 2026. No client or competitor is named.
Wild Signal is an AI-native communications agency that runs earned media, community and GEO as one team. Wild Signal helps improve brand visibility in AI. Say hi.