OMG PR & AI’s LLMs WTF
My learnings and musings on how this update in LLMs could be the best thing to happen to PR in a looooong while
Instead of searching on Google, people are now asking ChatGPT questions. And instead of clicking links, they’re reading the AI summaries.
So if your client doesn’t show up in that summary? They’re basically invisible.
I’ve been piecing learnings together and musing on this for the last couple of weeks. A lot’s still unclear, and there’s definitely a fair bit of guesswork out there… but here’s what I’ve pulled together so far. A steer in … a … direction, with more information and clarity to come.
It’s not what you say about you, it’s what everyone else says (aka PR!)
AI tools (ChatGPT, Claude, Gemini, etc.) don’t care so much about what’s written on your/your client’s website. They care about what the rest of the internet is saying.
If people are mentioning the brand in articles, reviews, Reddit threads, forums, or listicles, that’s what shows up in LLM answers. It’s less about telling your story once and more about making sure others are repeating it.
Which, funnily enough, is exactly what PR is built to do.
What LLMs consider “credible”
Not all content is created equal in the eyes of AI. Here are the kinds of sources that help brands show up more often:
Editorial media (and well-known agencies like Reuters and AP)
User-generated platforms with strong topical authority (Reddit!)
Owned media with strong structure (company blogs, FAQs, press releases via wire)
Customer reviews/ratings websites
Analyst reports and research papers
Industry awards and “Top X” lists
Books and eBooks
Academic studies and journals
Social media, Substacks, podcasts
The classics, rephrased for the AI age
Some of our everyday PR language is taking on new meaning. Here’s how to think about it now:
Brand voice / Share of voice = Repeated phrasing across different platforms
Media hits = Structured citations (earned, owned, and social)
SEO = Schema markup and clean formatting
Thought leadership = Insight-led, answer-focused content (Q&As, listicles, explainers)
Social reach = Conversation and interaction, not just impressions
Press releases and wires are useful again!
Remember when press releases felt a bit... tired? And PR newswires felt… expensive and redundant? Yeah, well, they’re having a moment. LLMs are a big fan. AI tools like structure, clarity, and well-labelled info, so the more organised your release is (bullet points, quotes, links, headings), the easier it is to be found and referenced.
The same goes for well-formatted bios, Q&As, and listicles. If your content is clean and skimmable, it's more likely to be pulled into an answer.
SEO is out, LEO is in (sort of)
Google reportedly lost 30% of its search traffic last year. That’s a huge online behavioural shift, and one that puts AI-generated summaries front and centre.
So if your comms and content strategy is still built solely around search rankings, it might be time for a rethink. LLMs are now a serious discovery channel in their own right.
LEO = Language Engine Optimisation = making your brand easy for LLMs to “get.”
What matters most:
Trusted sources
Repeated mentions
Structured formatting
Third-party validation
This is a good thing if you do PR well. Especially digital PR.
Shout out to a recent Muck Rack webinar, ‘PR in the GEO Era: How PR Drives the Future of Search’, which is genuinely worth a listen. Featuring Sarah Evans, Partner & Head of PR at Zen Media, and Michael Brito, Global Head of Data + Intelligence at Zeno Group. Here’s the link to the episode.
Owned content is good, but earned is what matters
Only about 20% of what LLMs reference is owned (i.e. what brands are publishing themselves). The rest is out in the wild: media coverage, reviews, forum chatter, and user comments. If that’s where the influence is, then PR is already halfway there.
This shift just puts even more weight behind earned visibility. A good feature or mention in the right place now does more than bring in clicks, it literally helps shape how your brand shows up in the AI-fed internet.
Reddit is becoming a (questionable) source of truth
Reddit is currently the most cited domain in AI-generated answers.
So if your client’s being talked about in niche threads, that chat could be shaping how the brand shows up next time someone asks ChatGPT a question.
The same goes for long-form LinkedIn articles, comments, and community discussions. AI is scanning all of it for what it considers news, intel, and opinions.
A few tools worth nosing around
Not essential, but helpful if you want to see how a brand is showing up right now:
In The Mix (by CoverageBook) – shows which media sources LLMs are pulling from
Peec AI – lets you see your (or your client’s) share of AI answers
Prompt Cowboy – lets you build and test prompts to see what LLMs “say” about you
Shout out to Celia Harding, the founder of LEOPRD, and Stella Bayles, author, speaker and director of PR technologies 'CoverageBook' & 'AnswerThePublic' for sharing these tools and tips in the recent PR Resolution podcast. Tune in to the recent episode here.
So what does this mean for how we work?
It’s a reframe for us PR pros. A few tweaks and new considerations to how we think and what we prioritise:
Write clearly. Structure things well. Use subheads, bullet points, Q&As, bios
Pitch media that gets cited, scraped, and indexed
Use consistent brand messaging across everything
Format content so it’s skimmable (Q&As, lists, bios, FAQs)
Take Reddit and forums more seriously than you probably want to
Final thought: this is a win for PR
We’ve always been in the business of shaping perception. Of building trust, creating noise, and making people care. AI is just adding another layer… one that rewards the kind of visibility PR is brilliant at creating.
We don’t need to be tech experts to get this right. We just need to keep doing our jobs, just slightly more structured, slightly more intentional, and with one eye on what bots are picking up. Visible, consistent, and smart about where and how stories are told.
If anything? This is our moment.
Quick glossary (so we’re all on the same page)
LLM = Large Language Model – tools like ChatGPT, Claude, Copilot, Gemini, etc.
LEO = Language Engine Optimisation – making your brand easy for LLMs to understand, quote, and surface
GEO = Generative Engine Optimisation – same idea, but for visual AI (DALL·E, DeepAI, etc.)
NLP = Natural Language Processing – the tech that lets AI understand human-style language