Charlie Marchant is CEO of Exposure Ninja, an independent, B Corp certified agency that has been running AI search campaigns for clients since 2023, back when Google’s AI Overviews were still called SGE. In this session, she talked through AI search visibility: how to find out whether AI tools mention your brand, why your competitors keep getting recommended instead of you, and what to do about it.
She started with a message most of the room recognised. It usually arrives around 4pm on a Tuesday, from the CEO, the board, or the sales team: “I’ve just asked ChatGPT about businesses like ours and we’re not there. Screenshots attached. Can you look at this by Friday?” Someone in the chat said they’d had one on a Friday night. If that’s you, you’re in good company.
Table of Contents
- Ranking on Google doesn't mean showing up in AI
- What being invisible in AI is costing you
- The four places your brand can sit in AI answers
- How to check your AI search visibility in an afternoon
- Why your competitors get named and you don't
- A three-part framework for AI search visibility
- Where to start: 30, 60, and 90 days
- Questions from the Q&A
- Making it work for you
Ranking on Google doesn’t mean showing up in AI
It’s easy to assume that if you’re winning on Google, you’ll win in AI answers too. Charlie’s research suggests otherwise. In a recent study of financial brands, Monzo had far more organic visits per month in the UK than Starling Bank. In AI answers, the picture flipped, and Starling was cited much more often.
“Google ranks pages. AI names brands,” Charlie said. Google gives you a list of pages to work through yourself. AI gives you recommendations and tries to solve your problem. SEO still matters, and there’s a lot of crossover, but they’re two different systems doing two different jobs, and marketers are now expected to manage both.
What being invisible in AI is costing you
A lot of the early research people used to do on blogs and websites now happens in AI tools. Charlie shared SimilarWeb research showing people prefer AI tools to search engines at every stage of buying, from discovery through to comparing options and deciding where to buy.
Visitors who’ve used AI before arriving on your website also tend to engage more. In the data Charlie shared, they viewed around twice as many pages and stayed much longer, roughly 12 minutes compared with five. That gives your own site far more chance to do its job.
The revenue can follow. One Exposure Ninja client in finance and education was getting around $100,000 a month in revenue from AI referrals in January. After focused AI search work, that was around $300,000 a month by August. The conversion rate for AI visitors was 4%, compared with 2.5% elsewhere.
The problem is that most reporting can’t see much of this. If someone reads an AI answer and doesn’t click, there’s no record of it, even though the answer may have shaped their decision. People often copy a brand name from an AI answer into Google, or type the URL straight in, so the credit lands in branded or direct traffic instead. If your branded search has been creeping up, AI might be part of the reason.
That’s why Charlie’s first measure isn’t traffic. It’s how often you’re mentioned in the AI conversations that matter to your business.
The four places your brand can sit in AI answers
Charlie described four positions, and you might find yourself in more than one:
- Invisible. You’re not mentioned in answers or cited as a source. As far as AI is concerned, you don’t exist yet.
- Cited. AI uses your pages as sources, but doesn’t recommend your business.
- Named, but described badly. You appear, but the description is out of date, thin, confused, or unflattering. This happens when AI pieces together your story from lots of different websites and nobody has tried to shape it.
- Recommended. You’re on the shortlist for the searches that matter to you, and described accurately, with your strengths clear.
The last one is where you want to be.
How to check your AI search visibility in an afternoon
This is Charlie’s four-step process, and it doesn’t need a big strategy behind it.
Step 1: Build a set of 20 prompts
Twenty is enough for an initial picture. “It’s better to start small and have something tracked,” Charlie said. Planning three months of prompts and tracking nothing helps nobody. She suggested four categories.
The first is unbranded, high intent prompts, such as “the best cybersecurity software for B2B businesses in the UK”. These are the searches your future customers make before they know you exist, so they matter most commercially. The second is comparison prompts, like “Monzo vs Starling, which is better for business banking?”, which show how you’re described against competitors.
The third is problem-led prompts, from people who know they have a problem but not what the solution is. If you’re not sure what your customers’ problems are, Charlie recommended listening to sales calls or reading your reviews. The fourth is branded prompts, such as “what does [your company] do, and who is it for?”. You should appear in these. The check is whether you’re described the way you’d like to be.
Step 2: Run them where your buyers search
If your buyers work in the Microsoft ecosystem, include Copilot. For consumer brands, ChatGPT and Google’s AI Overviews are likely to matter most. Whatever you use, run the prompts signed out, ideally in an incognito window. Otherwise you’re mostly measuring your own search history.
Step 3: Check three things for each prompt
- Mentioned: are you in the answer at all?
- Cited: is your website one of the sources? Most platforms list them in a side panel, and in ChatGPT you click “Sources” to see them.
- Described: when you do show up, is it accurate, flattering, and aimed at the right audience?
Charlie showed an example from ZUGU, an iPad case brand and Exposure Ninja client. In ChatGPT, ZUGU was recommended, with many of the sources coming from its own website. In Google’s AI Overview, it was recommended again, but the sources were third-party publications like ZDNet and Wired. Each platform works differently, which is why it’s worth looking at the sources as well as the answer.
Step 4: Compare yourself with competitors
Finally, compare how often you come up with how often competitors do. Charlie shared an example from CRM software, where Salesforce had around 45% of the AI share of voice. If you’re a smaller player in your market, don’t be put off. It can all still change.
Exposure Ninja has a free calculator on its website that uses your 20-prompt results to give a conservative estimate of what AI invisibility might be costing you. You don’t need to give an email address to use it.
Why your competitors get named and you don’t
Charlie gave five reasons.
AI answers are built mostly from other people’s websites. In the figures Charlie shared, more than 90% of AI answers that mention a brand draw on third-party sources, and only around 9% on the brand’s own site. So a lot of the most valuable work happens off your website, in the articles and reviews AI is already reading.
AI asks far more questions than you do. This is called query fan-out. You type “best CRM for a small team”, and the AI runs tens or hundreds of related searches behind the scenes. It checks the cost, how hard it is to migrate, who it’s for, and what real users say. A page that ranks for one keyword isn’t enough, and your content needs to cover the wider set of questions.
Your site is hard to read. Content hidden behind JavaScript, missing schema, or poor structure can all get in the way. Some very large brands have accidentally blocked AI crawlers in their robots.txt file or through Cloudflare settings, sometimes for months. There’s no shame in it, but it’s worth checking.
It’s not clear who you are. Inconsistent naming confuses AI. Charlie described a client that had acquired several products and renamed them, while the team used their own nicknames for them. AI couldn’t tell whether it was looking at one product or three.
There isn’t enough proof. AI checks what your website says against reviews, Reddit, Quora, press, trade publications, accreditations, and even your team’s LinkedIn profiles. If your reviews are old or you’re missing from the “best of” roundups for your category, it has less reason to trust you.
A three-part framework for AI search visibility
Charlie’s framework has three parts:
- Technical foundations. A clean, crawlable site, with clear content and schema, and nothing blocking AI crawlers.
- Positioning and content. Clear statements about what you do and who you do it for, plus content that answers your customers’ questions. Pricing helps. If your pricing is bespoke, you can still give ranges or describe the size of business you usually work with.
- Citations and reputation. Reviews, comparison pieces, roundups, and press, so AI finds a consistent picture of you elsewhere.
She shared two examples of it working. For The Ordinary, a campaign around its hyaluronic acid serum improved the copy, imagery, and user experience, and sales of that product went up 14%. For Elite Renewables, a regional installer up against much bigger energy companies, the work included repositioning, a site rebuild, location pages, local case studies, and targeted digital PR. Leads went up four times, organic users by 1,000%, and organic search, including AI, became its biggest acquisition channel.
Where to start: 30, 60, and 90 days
Charlie works in 30-day chunks, because everyone has other work to do.
First, diagnose. Run your 20 prompts across the main platforms, and use what you find to start a proper conversation with whoever sent that screenshot.
In the first 30 days, remove blockers. Check that AI crawlers aren’t blocked, and fix the schema on your money pages, meaning the 10 to 20 pages where people buy, book, or get in touch. Make sure your site states clearly what you do and who for. Positioning is ultimately a leadership decision, but marketing is well placed to start that conversation. Then rebuild one page to cover the wider set of questions your customers ask.
In days 30 to 90, build the proof. Plan how you’ll get into roundups and comparison sites, work out who in the business should ask customers for reviews, and start tracking AI referrals in Google Analytics.
If that feels like a lot, Charlie had a simpler version: “Just run the 20 prompts.”
Questions from the Q&A
Is AI search the same in B2B as B2C?
People use AI for B2B buying too, especially for purchases they know nothing about. Charlie recently used AI to compare password managers for her team. The difference comes in very niche markets. If you make a specific part for a specific machine, there’s so little information online that AI leans more heavily on your own site, including technical specifications. A couple of attendees in the chat turned out to be in exactly that kind of business.
Is AI revenue new, or just moved from other channels?
A bit of both. Buyers touch lots of channels before they buy. “It’s a pinball machine,” Charlie said, not a straight line. But across Exposure Ninja’s clients, AI revenue keeps growing without eating into organic or paid search, and all channels tend to grow as AI visibility improves.
Are ChatGPT ads worth it?
Exposure Ninja has been testing them, first on its own brand. Charlie’s early view is that the platform is much less mature than Google, LinkedIn, or Microsoft ads, and hard to get a return from. Placements aren’t always relevant either: she once saw an ad in a conversation about British politics that had nothing to do with it. She expects this to improve. For now, they’re worth testing if you have budget to spare and brand awareness is part of your remit, but not if you need to prove ROI.
Does using your work AI tool skew the results?
Yes, especially if memory is turned on. There’s no perfect method, but logging out, using an incognito window, and using the free version gets you closest to what most people see.
What if another brand has the same name as yours?
One attendee’s fashion brand shares its name with another clothing brand, one for luxury womenswear and the other for Gen Z gamers. Because the audiences are so different, people’s searches will usually make it clear which one they mean. The risk is when one brand is much better known.
Charlie’s advice was to make your positioning very clear, and to frame your content, PR, and reviews around your customer. Ask review questions that draw out context. “Five stars, love it” tells AI very little. A review that mentions buying a suit for a corporate gala tells it exactly who you’re for.
Should you aim for big publications or smaller, relevant ones?
“It’s not beyond you,” Charlie told the attendee who asked, who had just started her business. PR takes resilience: expect a reply to around one in 10 pitches, and coverage from around one in 20. Relevance matters more than size. A niche recipe blog is more useful to a sugar brand than a general placement in a national paper. Many of ZUGU’s citations come from tech publications, some dating back to 2024, and AI still cites them because they’re so relevant.
Making it work for you
The point Charlie kept returning to was having a baseline. A single screenshot from your CEO is an anecdote. Twenty prompts, checked across the platforms your buyers use, gives you something you can plan around and report on. It also gives you a much better answer for the next Tuesday afternoon message.
Improving your AI search visibility won’t happen by Friday, but checking where you stand can. After that, it’s 30 days at a time.
Thank you to Charlie for sharing so much, and for taking on a Q&A that reached more than 40 questions. Exposure Ninja also offers a free AI visibility review if you’d like help getting started š