(Almost) Everything You Need to Know About GEO in PR

The Wildfire guide to GEO in PR

Welcome to our (almost) complete guide to Generative Engine Optimisation (GEO) and PR.

The word “almost” is doing some heavy lifting in this guide. There’s a huge amount of speculation (and a fair bit of misinformation) floating around about how AI search actually works. It’s also changing constantly, so anyone on LinkedIn claiming to have “five easy steps” to master GEO is, at best, getting ahead of themselves.

But despite all the noise, one thing is clear: AI search and tools like ChatGPT are changing how tech buyers research products, discover new platforms, and explore ideas — even in B2B.

So with that in mind, we’ve created this guide to answer the questions we’re hearing most from our clients and partners, as they work out how AI search and GEO fit into their comms strategies. 

Each month we’ll be adding new questions to this guide, bringing it up to date. So if you have any new questions you’d like us to answer, do get in touch using the form at the end of this guide.

1. The basics

  • GEO (Generative Engine Optimisation) is about making sure your brand shows up in the answers generated by AI tools like ChatGPT, Claude, Gemini, Google’s AI Overviews, Perplexity, and others. Instead of trying to rank a webpage at the top of a long list of search results, the goal of GEO is to be referenced and recommended when those tools pull together a response to a question.

    The key difference is that these platforms don’t point users to content. Instead, they scan everything from media coverage and forums to social platforms and your own website to build a single, summarised answer. GEO is about how you can shape that pool of information so that when AI tools go looking, your brand is part of the answers and recommendations they provide.

  • SEO is Search Engine Optimisation. It’s about getting your content to rank in traditional search engine results — those familiar pages of blue links on Google, Yahoo, or Bing. You optimise pages, target keywords, and build backlinks so your site appears as high as possible in the results when someone searches for a relevant term.

    GEO is Generative Engine Optimisation. It’s about getting your content recommended by large language models (LLMs) like ChatGPT, or in the AI-generated recommendations that display at the top of search results.

    Instead of competing for a position in a list, you’re competing to be referenced in a single, compiled answer. Whether your brand gets included is shaped by how often it appears online, how clearly it’s associated with a topic, and how consistently that story shows up across different sources.

  • Short answer: no one’s really agreed what to call it yet.

    You’ll hear AEO (Answer Engine Optimisation), GEO (Generative Engine Optimisation), and just AI search used almost interchangeably, depending on who you’re talking to. AEO leans towards getting your content pulled into direct answers (think snippets or AI responses), while GEO is broader and looks at how your brand shows up across AI-generated outputs more generally. “AI search” is the loosest of the three — it doesn’t really describe a discipline so much as the shift in how people are searching.

    There are also plenty of other terms floating around — LLMO, GAIO, and a few more besides — but for the sake of clarity, we’re sticking with GEO throughout this report.

  • AI search results are already shaping how people form opinions about your tech product or brand — often before they’ve visited your website or spoken to your team.

    And it’s not just consumers who are using these tools to find, buy, and evaluate tech. AI search is also increasingly influential in B2B. According to Forrester, 89% of B2B buyers now use generative AI at least once within the purchasing process.

    PR and comms teams are in a uniquely strong position here because AI models rely so heavily on earned mentions. Research shows that editorial media drives more than 60% of LLM responses about the world’s largest brands, while data from Muck Rack suggests more than 85% of citations in AI answers come from earned sources.

    But it’s not just earned media that ties GEO to PR. As AI tools build answers from multiple sources, they look for signals of trust and credibility, and consistent repetition of key messages. In other words, the very things PR teams have been building for decades!

  • AI search is already driving traffic, although volumes are still relatively small compared to Google. What’s interesting is the quality of that traffic. 

    Multiple studies show that visitors coming from AI tools are often further along in their decision-making process. For example, Semrush found that AI-referred visitors convert at around 4.4x the rate of traditional organic traffic. Other studies point to similar patterns, with LLM traffic delivering significantly higher value per visit — typically converting 3–6x more than those from organic search.

    The reason is fairly simple. By the time someone clicks through from an AI-generated answer, they’ve often already compared options, read summaries, and built a shortlist. In B2B, where research cycles are long and messy, that makes AI search a valuable source worth targeting.

2. How AI search works

  • When you ask an AI tool a question, it pulls from a mix of sources and assembles a response. That typically includes older training data (what it already knows), live web results (what it can find now), and the context of your query. It also expands that query into related prompts — called fan-out queries — to build a fuller picture of the topic before generating an answer.

    Typically this process draws on editorial coverage, comparison content, forums like Reddit, social platforms, and other sources that help answer the question clearly. The model then looks for patterns, asking what’s repeated, what aligns across sources, and what seems most useful, before compiling that into a single response.

    For comms teams, the key point is that these systems are cross-referencing multiple sources and forming a view based on credibility, authority, and message consistency across the web. These are all things PR has been owning and building for years.

  • We’ve all been there. You ask ChatGPT something like “what are the best vendors in this space?” and get a neat list of your competitors… with your brand either right at the bottom or missing entirely. These answers are not only shaped by what the model can find across the web, but also by what it considers most helpful. AI tools tend to favour content that clearly answers the question being asked. Right now, that often includes list-style content, comparisons, and pages that directly spell out options, pros and cons, or rankings. They also favour content that mirrors how people actually ask questions, rather than dense product-led copy or waffly thought leadership.

    What tends to work best is focusing on the questions your audience is actually asking. With that in mind, many marketing teams are now mapping out long lists of queries based on search data, social conversations, and the questions coming into sales and support teams. Creating content that answers those questions directly gives AI tools something clear to reference — but it’s also just good for your audience. If you can help answer the questions people have about your industry, product, or brand, why wouldn’t you?

  • Not in any way that will last.

    There’s a temptation to treat GEO like the early days of SEO. Spot patterns, figure out how the algorithm decides what to recommend, and try to get ahead of it. 

    At the moment, you’ll see certain formats showing up more often, like ‘listicles’. It would be very easy for PR and content teams to see that and think, “Great, let’s turn everything into lists.” But that approach has a short shelf life. As soon as platforms like ChatGPT catch on, they adjust. Just like Google did with SEO shortcuts, or LinkedIn does when people try to game its feed.

    Instead of trying to outsmart the system, you’re better off working with what these tools are trying to do. They’re designed to be useful. So if you want to ‘win’ at GEO, focus on creating content that gives your target audience genuinely useful advice. Ideally, written in a clear and easy-to-understand format.

3. Earned and social

  • Not for GEO.

    Backlinks don’t carry the same weight in AI search as they do in traditional SEO. Chasing journalists for links — or measuring coverage purely on whether it includes one — doesn’t really mean anything for AI.

    The good news is that this means the work your PR team is already doing — building visibility, starting conversations, and landing media mentions — has a direct impact on your GEO, even without a hyperlink included.

    Of course, backlinks still matter in a broader sense. They help with traditional search and they make it easier for people to trace information back to a source and explore further. So if a link adds value to the reader’s experience, then it’s still worth having in earned coverage.

  • Personally, we’ve never been huge fans of overly formal ‘corporate’ press releases. While official releases can be important to share information with the market, effective press releases should also act as a cheat sheet that helps journalists quickly see what’s worth covering.

    That’s why tailored pitching around a press release (adapting headlines, shaping angles, and writing directly to what each journalist cares about) is far more effective than sending one standard version and expecting them to dig out the story themselves.

    That said, LLMs may be reviving the need for a more structured, ‘formal’ version of releases to sit on your website. 

    At the moment (and this may change), AI tools tend to treat press releases as a clear, factual source that comes directly from the company. So having a clean, well-written version hosted on your site helps them understand and reference your announcements. The best approach is to do both: create something useful and tailored for journalists, and make sure there’s an ‘official’ version that’s easy for AI tools to read, extract from, and cross-check.

  • This one’s open to debate. Wire services can help get your news in more places, which in theory increases the chances of your messaging being picked up and repeated across the web. That kind of distribution can reinforce visibility and give AI tools more sources to draw from.

    They also serve as a more ‘official’ source — so when LLMs are cross-checking what they’ve found elsewhere, a wire-hosted version can help validate that the information is accurate and endorsed by the company itself.

    The trade-off is quality and context. Syndicated content isn’t always the most influential or trusted, and it often lacks the nuance or depth that stronger editorial coverage provides. Used well, it’s a useful layer in the mix, particularly for reach and consistency. Used in isolation, it’s unlikely to make much difference.

  • Reddit is consistently one of the most cited sources in AI-generated answers, largely because it’s full of direct, experience-based responses to real questions (asked in a very AI-friendly Q&A format). AI tools seem to value that kind of content, especially when it reflects genuine discussion rather than polished brand messaging.

    But keep in mind that this is permanently changing. Recent updates mean that reddit is not being scrapped in the same way it used to, reducing the number of new citations coming from the site.

    LinkedIn articles on the other hand remain influential, particularly for B2B topics. As more professional conversations, opinions, and content are shared there, it’s increasingly being picked up as a source.

    Using LinkedIn articles to share useful, relevant content (not sales messages or AI-generated cringe) can help shape how your brand appears in the wider conversation that AI tools are drawing from.

    There’s also a practical balance to strike with visibility. Keeping profiles completely locked down limits how much of that content can be found and referenced by AI. At the same time, you probably don’t want everything public — especially things like your follower list. It’s worth reviewing your LinkedIn privacy settings so your content and insights are visible, while still protecting the parts you’d rather keep private.

4. Owned content

  • Research shows that owned content plays a significant role, with around 44% of LLM responses being drawn from owned sources.

    And while earned media helps shape how your brand is talked about, your website acts as the place AI tools go to verify that information. When a model finds mentions of your brand elsewhere, it will often cross-check your site to confirm those claims. If the messaging doesn’t match — you’re less likely to be included.

    When developing content for your own website, remember that AI tools favour content that is easy to interpret. That means pages and blogs that clearly answer questions, use straightforward language, and are organised in a logical way. Typical examples would include using clear headings, tables, concise explanations, FAQ sections, and even ‘TLDR’ summaries at the top of the page.

    As with traditional SEO, there are also technical factors to take into account. If AI crawlers can’t access or process your content, it won’t be considered — no matter how good it is. That includes pages blocked by robots.txt or heavy JavaScript that bots struggle to render. Even site performance plays a role. If a page takes too long to load, AI bots may simply give up and move on.

    It’s well worth speaking to your web team to make sure none of these issues impact your site.

  • PDFs still have a place in PR, especially if you need something printable for events, sales meetings, or trade shows. But there’s no denying they’re harder for LLMs to read and extract from.

    So if you’re producing long-form content like reports or whitepapers, it’s a good idea to also have an on-site version. That could be the full content, or a series of pages that highlight the key findings and answer the main questions coming out of it.

    Gating is a bit more clear-cut. Hard gates are a ‘no’ if your goal is visibility in AI search. If a model can’t access the content, it can’t use it — and that applies to both your own content and coverage sitting behind paywalls. Gates still have a role in lead generation, but there’s a trade-off to consider.

    A more balanced approach is to keep the insight open and gate the next step. For example, share the thinking, the data, and the answers publicly, then offer something more interactive at the end — like a tool, audit, or deeper analysis — in exchange for sign-up. That way you stay visible, while still giving your sales team something to work with.

  • On-site newsrooms provide a central place for announcements, updates, and key company information, which makes it easier for AI tools to verify what’s being said about you elsewhere — and sense-check it against an ‘official’ version from your company.

    It also helps over time. If your coverage is saying one thing, but your site doesn’t reflect that, it weakens your chances of being included. A well-maintained newsroom reinforces what’s being said externally and gives AI tools a clear reference point to check against. 

    If you do have a newsroom, it’s worth making sure your releases are published as HTML pages (like blog posts), rather than a series of linked PDFs, so they’re readable and usable for AI tools.

More questions?

The GEO space is constantly changing, and the way AI search works today won’t look exactly the same in a few months’ time. We’ll update this guide regularly to reflect what’s changing, add new questions, and share what we’re seeing in practice.

If there’s something you’re curious about — or something we’ve missed — let us know. Leave your questions below and we’ll do our best to answer them in the next update. Include your email address and we’ll also send you the answers directly.