Yaser Ayub, founder of Rayze. Yaser started out as a web designer and general marketer, then moved into SEO and, more recently, GEO and AEO. Much of his background is in fintech and financial services, so he has worked in some of the most competitive search spaces there are. With 18 years of SEO experience, he helps companies understand their share of voice in AI and improve it.
Hannah George-Forbes, Global Head of Content Marketing at Onclusive. Onclusive is a media intelligence company. Hannah leads its content team and looks after everything from global content campaigns to the on-site blog across multiple markets. She has spent about four and a half years in the role, after starting out as a freelance and radio journalist and then working in PR. That mix gives her a clear view of the problems marketers, PR teams and communicators face.
Every boardroom is asking the same uncomfortable question. When someone asks an AI tool for a recommendation, does our brand come up? Are our competitors being recommended and cited? Are there brands we have never heard of taking ground from us? We hear this question every week, and it is why we have spent so long testing how AI visibility can be measured. Share of voice is the metric that needed the most clarity.
It helps to keep the scale in view. A SparkToro study shows Google is still by far the most used search platform, even when you add up Amazon, YouTube and the rest. AI matters, but its share of overall searching is still small, even though the behaviour is very different. So if leadership asks why you are not showing up in LLMs, you can point out that people are still searching in Google. The caveat is that AI is now built into Google itself, through AI Overviews at the top of the results.
Google's "People also ask" boxes used to be a place brands fought to appear. AI Overviews have now replaced them almost entirely. They also push organic results further down the page, so the question is less about where you rank and more about how far someone has to scroll to find you. It is also a zero-click world: the clicks are not attributed to anything, which makes return on investment harder to measure.
Hannah's team at Onclusive saw this shift coming. With limited resources, they chose to prioritise AI Overviews first, because it was prime real estate they could influence quickly. To find the questions people really ask, we use a tool called AlsoAsked, which pulls those questions in real time. Yaser and Hannah have been building them into their content briefs for a while.
When we say AI, we mean ChatGPT, Gemini, Perplexity, Claude and Google's AI Overviews, all treated as one. People use them to ask longer, more detailed questions, to compare providers and to build shortlists. Yaser's wife even used one to plan a holiday itinerary. Whatever the task, buyers are making decisions inside these answers.
The effect on lead quality is what caught our attention. Before a recent Similarweb study came out, several of our clients told us their leads were getting noticeably better. We suspected the AI work was behind it, but we could not be sure. The study then found that when a brand is recommended in AI, people are 2.5x times more likely to visit its homepage directly, and they spend twice as long on the site.
Yaser says it feels a little like advertising on billboards ten years ago, with people going straight to the brand. We are now measuring the impact on direct traffic alongside our AI optimisation work. Hannah has seen the same rise in direct visits at Onclusive. The whole ecosystem is connected, even when we cannot yet connect every dot.
Rankings and clicks no longer tell the whole story. Four measures give a fuller picture, and each has its own quirks.
Prompt tracking. A prompt is the question someone types into an AI tool. You need a library of the prompts that matter to your buyers, and you need to track them over time. We keep ours open-ended, such as "recommend the best SEO agency in London" or "recommend a mortgage provider", because tracking every individual prompt is almost impossible. Deciding which prompts to track is the most important job.
Share of voice. This is how much your brand appears in AI answers. At Onclusive, Hannah's team weight it by overall visibility and by rank position within the answer, so a mention at the top counts for more than one at the bottom. Different tools calculate it differently, and some call it presence rate. The simplest version is the number of tracked prompts where your brand appears, divided by the total number of prompts, times 100. Another version divides your brand mentions by all brand mentions across everything.
Sentiment. Onclusive calls this reputation in its GEO Analytics tool. It is a weighted sentiment score drawn from AI responses about your brand. When someone asks an AI tool to compare one brand with another, the answer can be quite honest, based on sources such as customer review platforms like G2. If your share of voice is large but the sentiment is negative, fixing the sentiment is the first job.
AI session data. Tracking visits from AI is getting harder, and AI traffic is still a small share of the total. For some of our clients, though, we can trace a visit back to a marketing-qualified lead. It is not a perfect science and you cannot capture every lead that comes from AI. The ones you can see are well worth having.
Invest in a tool that measures share of voice, then pick one and stick with it. Because every tool calculates the number differently, switching will give you a very different result. Use your first reading as a baseline and treat the figure as a direction of travel. Are we generally trending up? That is what senior leadership wants to know: where are we appearing, and are we getting better at it?
Two habits keep the number honest. First, do not measure yourself only against your direct competitors. You do want to know how you compare, but a competitor-only view is not a true picture of what the AI tools are actually saying, so your share of voice will not reflect the real landscape. Second, measure across every platform, not just one. The tools draw on very different sources, so you might do well on Gemini or Perplexity and not show up at all on ChatGPT. Your share of voice there will be far lower, and you need to know which tools your buyers actually use.
Hannah's team at Onclusive produces reports on brands and industries, and on who dominates each one across all LLMs. Along the way they noticed something odd. Their own blog content was being cited in AI answers, but the brand was not being recommended. The reason is that the trust signals around a company do not come only from its own media.
The same pattern showed up in another way. Onclusive's high-value informational content began appearing at the top of Google's AI Overviews within days, which was a great result. But once the prompt turned commercial, for example "tell me who the best media monitoring providers are and who I should pick", the brand slipped away. They saw the same thing in ChatGPT, which most people were using by the start of this year. For highly commercial prompts, the answer was sometimes no.
Once you have the basics, three further layers matter:
Yaser describes an LLM as a giant octopus. The head sends its tentacles out to find the most relevant information from the most relevant destinations, then brings it back and pulls it together into one answer. We can see which sources those tentacles prefer in each niche and industry, and that picture keeps changing.
For a while now, organic LinkedIn content has been a trusted source for LLMs in B2B. Before that, listicles worked well: pages where a company named itself the best software in its industry and then listed its competitors. ChatGPT liked them, and then the tools devalued them. Reddit followed the same path. Everyone scrambled to build a Reddit strategy, ChatGPT's use of the platform climbed, and then it fell off a cliff. We expect LinkedIn to go the same way eventually.
The advice is simple, though not quick. Be as authentic as you can, create great quality content and get on the most trusted sources.
Traditional SEO is keyword-led. AI share of voice is prompt-led, and that needs a different mindset. Buyers type full sentences and long, detailed questions into AI tools, and they ask similar questions in forums such as Reddit and Quora, which is probably why Reddit did so well. We saw the gap first-hand when a prospect asked us to track 50 keywords in LLMs. We explained that this is not how it works, and he went with someone else.
The better starting point is your ideal customer profile (ICP): the kind of client or customer you most want to win. Dig into their pain points and needs. It takes a lot of research and grunt work, but it is everything we should already know, and it pays off. It also improves your landing pages, because you end up speaking better to your audience.
There are practical ways to do the research. We built an AI agent that plugs into customer sales calls, asks a set of questions and pulls out the most recent topics customers raise, which we then write content for. Listening to recorded sales and discovery calls is just as useful. At Onclusive, Hannah's team used topic clusters built from their SEO keywords to arrive at their commercial and informational prompts.
Nobody can tell you exactly what everyone is searching for, so be wary of anyone who claims they can. You have to do the work. Optimising for LLMs is a topic for another webinar.
Yaser believes the value PR brings to an organisation has roughly doubled. Many of the sources that shape what LLMs recommend are heavily influenced by earned media, which is coverage you win rather than pay for. Hannah points out that PR and comms teams have always struggled to prove their worth beyond vanity metrics. This is the moment that changes. PR now has a direct influence on brand awareness, on how your brand is perceived and on its reputation, all of which AI now feeds back to buyers.
Even coverage in trade press, which may not be tier-one media, directly influences the category prompts people ask. For years SEOs and PR teams have not always got on, because SEOs wanted links. Links still matter, but mentions now matter just as much, and so does where you get them.
Think of it as layers. Your own media, such as your website, blog and YouTube channel, is the bare minimum, and it is mostly what we should have been doing for SEO anyway. Third-party sources then complete the picture. Tier-one and trade media create the commercial recommendations that owned content cannot. Link PR to the work you are already doing and you become invaluable.
SEO sits underneath all of it. Yaser recently told a friend who is building a site for a well-known person that SEO is still fundamental to doing well in LLMs, because the LLMs crawl the site. The whole ecosystem is SEO, and it always has been. It has just become more important. Someone with a big YouTube and LinkedIn presence already has a good head start.
Onclusive has published two AI visibility playbooks. The first covered the basics and how the source ecosystem works. The second goes further and shows how much that ecosystem changes once you break an industry down by geography and by sub-category. Telecoms in the UK looks different from telecoms in the US, and in automotive, SUVs and electric cars draw on different sources. The research found four things:
Banking is a good example. In the UK it is led by journalism, with The Guardian, the Financial Times and The Telegraph among the most cited sources. They are hard to get into, but once you know how to tell a good story and make the right connections, PR can have a huge impact. Split banking into sub-categories and the picture shifts. In the US, digital banking and online payment services both cite CNBC as their top source, but they diverge straight after that. Digital banking leans on personal finance sites such as NerdWallet, while online payments leans on tech and innovation coverage. Even within one industry, you need to get granular.
The pattern changes again in other sectors. For B2C and e-commerce companies, the key sources are very different from the LinkedIn focus common in B2B. In software, review sites such as G2 are a major source, and Onclusive had already invested in customer advocacy there.
The playbook landing page lists 12 industries and the markets analysed, and you can filter to see the top 20 sources for each combination.
Our clients have seen real movement from this work, and Onclusive is a good example. In five months, their AI referrals rose considerably, and Gemini sessions went up too. Their share of voice grew from 1.8% to 4.2%. That is measured on open-ended prompts built around their ICP, which they have not changed in a while, so it is a pure baseline. Hannah is quick to put the number in context. Their biggest competitor sits at around 11%, thanks to long-standing strong SEO foundations, so a move from 1.8% to 4.2% is meaningful. It also shows the optimisation work is having an effect. Non-branded searches and clicks are up considerably as well, because much of the core work was AI-focused but SEO-led.
The work also turned up problems. Onclusive was not appearing in Claude at all, which only measuring across every platform would have revealed.
We looked at what AI knew about Onclusive and its competitors, and identified clear topic gaps for each entity they wanted to optimise. We sharpened the messaging and the ICP across the whole website so it was very clear who it was for.
Internal linking made the biggest difference. Hannah, who implements much of it, says showing LLMs how everything connects has probably moved the needle more than anything else, and it was not something they had prioritised before. Search engines do not read a page as words. They turn the words into numbers, called an embedding. We built our own tool that pulls those embeddings, matches how similar pages are to each other using cosine similarity and builds the internal links from that. Yaser is happy to run it on a couple of pages for anyone who is interested.
Ongoing technical work mattered too. Make sure the site is free of errors and fast, so crawlers can reach the content easily. We also use entity maps, a file a little like a sitemap that tells LLMs plainly what your website is about and what your core products and offerings are.
Hannah's central team is small, and one person looks after the website alongside digital campaigns. A full technical audit felt overwhelming, so having us sweep in and manage the fixes helped them get on with it, and page load speeds improved. We do not hand over a spreadsheet and leave. We give the remedy too, because a list of problems does not go anywhere on its own.
Read the case study here: How We Grew AI Visibility by 364% and Nearly Doubled Non-Branded Search Impressions
Hannah closed the main session with three principles. Each comes back to authenticity, and none of them is a quick fix.
Topical authority. Build real authority around what your company does. This is harder in larger organisations, where not everyone agrees on what the brand wants to be known for. If you have experts in-house, use them: put them in your content, create author bios on your website and have them write for your blog. Approval processes make this hard, but it builds authority on your own media. LinkedIn articles are another place to build authority around specific topics.
Content that isn't a commodity. Use a content gap analysis to see where you need to fill gaps, and fill them with high-value content that brings something to your audience, not generic AI filler. Structure it so LLMs can read it. That does not mean simply adding a special file for LLMs (lllms.txt file), which does not work. It means structuring the content itself in the right way. This feeds directly into your topic authority.
Third-party validation. This is the hardest part, and it is what completes the picture. It includes being featured in "best of" lists on impartial domains, industry coverage and PR. Great PR is the foundation you already know: storytelling, pitching and using data in press releases. Make sure you also have a newsroom on your website that LLMs can read. A good newsroom makes it easier for journalists to cover your story, and it helps AI tools pick up the right information about your brand the next time they train on new data.
You can't yet. Someone could ask an AI tool a question, then see an ad, then click through, and there is no proper way to measure how those pieces connect.
Start with AlsoAsked, which pulls out the long-tail questions people search for. Build prompts around your ICP, and get your prompt tracking right. Yaser flagged an agency that wrote the exact prompt string into an article to influence its results, which tells you nothing, because nobody else searches that exact phrase.
Bottom of the funnel is where buyers compare reviews and customer feedback, so customer advocacy campaigns matter. In a recent LLM gap analysis, we found that a competitor appeared more often because its first paragraph called out some very clear statistics. Our fix was to reword the first paragraph and cut the content back, since the tools want the most important answer straight away. Adding case studies and improving trust helps too. Anything that makes your website better and speaks your customers' language helps, and it is probably why lead quality is improving.
It can, but it also drops off faster, which is where content freshness comes in. Some models crawl the web, so a press release can be picked up straight away. Others rely on training data, so the release will not appear until that data is updated, and each tool does it differently. If you do not keep content up to date, it will probably disappear within three to six months. That is why our SEO work has always involved reviewing and optimising content and removing what no longer drives value.
Being seen for high-intent, bottom-of-funnel prompts, such as "recommend me software for my niche", is worth far more than being seen for a broad prompt.
We can run the same audit on your brand: what AI already knows about you, where the gaps are, and what it will take to close them.
Book a 1-2-1 with our team. Talk to a senior consultant.