On 29 September, Charlie Marchant from Exposure Ninja walked us through how to check whether ChatGPT, Google, and Perplexity are recommending your brand, and what to do if they aren’t. A lot of you had questions, from how often to re-run your prompts to whether ChatGPT ads are worth testing.
Below are the answers to 40 of them, in five sections. Some answers mention Exposure Ninja’s own client work, because that’s where Charlie’s examples come from 🙂
Check out the webinar on demand right here.
Tracking and measuring AI visibility
Is there an AI equivalent of Google Search Console or Google for prompts, or Google Trends for what people are prompting for?
Asked by James Proctor, Montse Pe
Not in the way we’d all love. There’s no Google Search Console equivalent from OpenAI yet, and nobody publishes first-party prompt data the way Google Trends does for searches.
The closest thing is Bing Webmaster Tools. Its AI Performance report (in public preview since February 2026) shows how often your pages are cited in Copilot and Bing’s AI summaries, plus a sample of the “grounding queries” it used to find you. Google Search Console does include AI Overviews and AI Mode in your performance data, but it doesn’t split them out.
For prompt demand, tools like Semrush, SE Ranking, Peec AI and Profound estimate prompt volumes. I’d treat those as directional rather than gospel. Don’t forget your own data: the long, question-style queries in Search Console are a decent starting point to how people phrase things to AI.
How do you check your brand in AI?
Asked by Chloe Griffin
The quick version of what I walked through:
- Write 20 prompts from your sales calls, mostly unbranded (“best [category] for [customer type]”).
- Run them in ChatGPT, Google AI Overviews and AI Mode, Gemini, Perplexity and Copilot, signed out or in a temporary chat.
- For each one, record whether you’re mentioned, whether your site is cited, and how you’re described.
- Put your results next to three competitors.
If you’d rather we did it for you, that’s what Exposure Ninja’s free AI Visibility Review covers. Please do drop us a message.
How often would you carry out the AI research exercise?
Asked by Lindsey Morham
Get a baseline now, then I’d re-run your manual set monthly. If you’re using a tracking tool, it’ll run prompts daily or weekly, which is better because it smooths out the randomness.
Look at trends month to month rather than reacting to one run. I’d also re-run after anything big, like a site launch, a rebrand or a PR campaign, so you can see the effect.
Review the prompt set itself every quarter. The questions your customers ask can evolve over time, so your prompts should too.
To benchmark against competitors are we doing the same prompts but puttin in competitor name?
Asked by Lauren Paine
The most useful benchmark is running the same unbranded prompts (“best [category] for [customer type]”) and recording every brand that appears. That shows you and your competitors side by side in the same answers, and you can work out a share of voice.
Then add a few comparison prompts (“[you] vs [competitor]”) to see how you’re described against each other. Prompts like “alternatives to [competitor]” are also worth tracking because they’re very high intent, and you want to be in that answer.
What’s the best information source to use for building the trackable prompts? Calls with prospects/clients?
Asked by Érico Mafra
Sales and prospect calls are the best source, because you hear the exact words people use.
I’d add customer support tickets, answers to “how did you hear about us”, your on-site search, long question-style queries in Search Console, People Also Ask boxes, and Reddit or forum threads in your space.
The key is writing prompts in your customers’ language, not your internal language. Customers rarely describe your product the way your marketing team does.
If you run the same prompt in the same LLM more than once, will you get the same results each time? I thought they didn’t (i.e. just because your cited once, doesn’t mean you are in the next identical prompt on the same LLM)
AI answers are probabilistic, so the same prompt can give a different list each time.
SparkToro ran nearly 3,000 tests earlier this year and found the chance of getting the same list of brands twice was under 1 in 100, and the same list in the same order was closer to 1 in 1,000.
What did hold up in that research was how often a brand appears across lots of runs. Some brands showed up in 60 to 90% of answers for a given intent. That’s why I’d always report AI visibility as a percentage across repeated runs and different versions of prompts on the same topics/search intents.
how do we make sure we can stand by our finds, especially given research that shows you can run the same prompt multiple times and get very different results. It’s feels more like a lottery than a ranking. plenty of people in businesses still applying SEO type thinking…
Asked by Dave Heywood
Run each prompt multiple times, across several engines, over weeks, and report how often you appear as a percentage – or use a tool to do this for you. You want to get to: “We appeared in 7 out of 10 runs” which is something clear, rather than “We ranked third sometimes.”
Studies have shown that brands with real AI presence appeared in 60 to 90% of runs, so strong visibility does show through the noise. Use a decent sample of prompts, look at trends rather than snapshots, and be open about your method when you report it internally.
It is definitely not a lottery though, as many strategies and tactics influence AI visibility.
Will us searching for prompts be affected by our own AI use, so it’ll be making recommendations to us based on our history rather than representing what others might see?
Asked by Kathryn Borthwick
Yes, it can. ChatGPT and Gemini both use memory and your past conversations, so if you’ve been asking about your own brand for months, you’re not seeing what a new customer sees.
Run your checks signed out, in a temporary chat, in an incognito browser, or with memory switched off. Also phrase prompts the way a customer would, and don’t mention that you work there.
The AI visibility tools run prompts in clean sessions, which gives you a consistent baseline to track over time. It won’t match any one real person’s personalised answer, but it’s the fairest benchmark you can get.
I use LLM for work, would using prompts in the same LLM skew result? Would the LLM give me the result I want rather than the true picture?
Asked by Louise Major
Yes, it can skew things. Memory, custom instructions and past chats about your brand or industry all feed into the answer you get. Run checks in a temporary chat or signed out, or use a separate account with memory off.
Should we try these prompts across different ai tools when researching or just pick one?
Asked by Kathryn Leadbeater
Across different tools, because results vary between them. You might be recommended in Perplexity and invisible in ChatGPT.
At a minimum I’d cover ChatGPT and Google (AI Overviews and AI Mode), since that’s where most people are. Add Perplexity because it’s the easiest place to see sources, Gemini, Claude and Copilot if you sell into Microsoft-heavy organisations.
Prioritise the ones your customers use most.
Can you offer any further tips on how to find out which AI your customers are more likely to use?
Asked by Susie Lober
The simplest way is to ask them: in your enquiry form, onboarding survey or on sales calls. Even one question like “Which AI tools do you use when researching?” is useful.
GA4 helps too. Look at referral traffic in the AI Assistant traffic channel.
Beyond that, ChatGPT has the largest share of standalone AI assistant use, and Google’s AI Overviews reach almost everyone who uses Google.
I launched a new website 2 weeks ago. Is it too soon to run queries on AI to diagnose?
Asked by Kathryn Leadbeater
It’s not too soon to run a baseline, but it’s too soon to judge results.
Use this stage to check the foundations. Are your pages indexed in Google and Bing? Are AI crawlers allowed in your robots.txt? Are your old URLs redirected properly? Does your structured data validate?
Then re-run the same prompts monthly. You’ll usually see things start to move over the following weeks and months, especially as you build third-party mentions.
If we had endless budget…Do you think there would be focus groups where people (AKA customer personas) will be testing prompts in their own accounts to test and perfect the brand’s AI exposure?
Asked by Eva De Luis
I love this idea, and yes, with endless budget I’d do that. Large-scale tracking tells you how often you appear. Real customers testing in their own accounts would show you why, and how answers change once memory and personalisation kick in.
How can you track mentions reliably when a user’s relationship with their LLM is so personalised and contextual?
Asked by Paul Mortimer
You can’t track every personalised version, and to be fair we never could in SEO either. Google search results have been personalised for years.
What you can track is the baseline: how visible you are in clean sessions across a broad set of prompts. Personalisation shifts answers away from that baseline, but brands with strong baseline visibility and a strong third-party reputation tend to show up across far more contexts. Use the topical areas.
Then pair that leading indicator with outcome measures: self-reported attribution (“how did you hear about us”), branded search volume and AI referral traffic. Together they give you a picture you can make decisions with, even if it’s not perfect.
is AI revenue incremental, or just revenue coming from previous channels
Asked by Mark Grech
It’s a mix.
Some of it is substitution: people who would have Googled are now asking ChatGPT. However, the visitors who do click from AI answers tend to convert well. In the data I shared, it was 2.5 to 4 times the conversion rate of Google search visitors.
If AI answers the question and names a competitor, that customer is gone, whether you call it incremental or not.
Third-party sources and digital PR
Backlinks and digital PR campaigns have always been an important pillar of an SEO/visibility strategy – where reputation is so crucial even now, are you finding this is the case for AI visibility too?
Asked by Clare Dyckhoff
Yes, if anything reputation matters more. AI answers are mostly built from other people’s websites. Search Engine Land reported that around 82%+ of citations come from third-party sites, which is the stat I shared in the webinar deck.
What changes is how you judge a PR win. The mention and the context around it carry a lot of weight, so a good unlinked mention in the right publication still helps. I’d pick target publications based on which ones AI is already citing for your prompts, rather than on domain authority alone.
With Elite Renewables, the Digital PR coverage in the Mirror and Ideal Home was a big part of why their links in Google AI Overviews went from 67 to 260.
how do we find out what 3rd party sources and articles ai is using?
Asked by Kathryn Leadbeater
Start manually. Perplexity lists its sources for every answer, ChatGPT shows sources when it searches the web, and Google AI Overviews and AI Mode show link icons next to each part of the answer. Log the domains for each prompt in a spreadsheet and you’ll quickly see the same sites coming up again and again.
If you want to do it at scale, AI visibility tools like Peec AI, Semrush and Profound report the domains and URLs being cited across your prompt set.
Are there any sources we specifically need to target that AI consistently cites eg YouTube, Reddit etc?
Asked by Kate Strawson
There isn’t one universal list. It depends on your category and on the engine. Reddit, YouTube, Wikipedia etc come up a lot across industries, but in your market the most-cited source might be a trade publication, a review site or a comparison blog you’ve never thought about.
Different engines have different habits too. Google’s AI features pull heavily from its own ecosystem, including YouTube, while Perplexity leans on a wide range of web sources.
So I’d run your prompt set, log the sources, and let that tell you where to focus. It’s much more efficient than trying to be everywhere.
In terms of third party sources, should we be looking to get onto high traffic sites (obviously ideal) or do lower traffic ones still offer benefit? I’ve just started my business and trying to pitch to big name publications feels a bit beyond me currently
Asked by Eve Davitt
Lower-traffic sites absolutely still help. What matters to AI is whether the website is relevant and whether it gets pulled in for your customers’ prompts.
Niche blogs, local sites, trade newsletters and specialist roundups get cited frequently. Start with the ones already appearing in your prompt sources. They’re often much more approachable than national publications. For our client ZUGU, who sell iPad cases, we published in Wired and ZDNET, two well-known publications specifically in the tech sector and these influence many AI answers years later.
Build up from there. The big names become much easier to pitch once you’ve got a track record and a story to tell.
How do we get other websites to mention our brand and website? Apart from PR, what else can be done? I’m in the legal sector
Asked by Jodie Walsh
There’s quite a lot beyond PR in legal. Directories carry real weight: Legal 500 and Chambers if they’re relevant to your practice, the Law Society’s Find a Solicitor, and review platforms like ReviewSolicitors, Trustpilot and Google.
Expert commentary for trade press, guest articles in the publications your clients read (not just legal media), and webinars or talks for industry bodies all create mentions. Referral partners like accountants, estate agents and IFAs can feature you on their sites. Podcasts are good too.
Does social media, podcasts, youtubes help a lot?
Asked by Piotr Bytnar
Yes, they help a lot, especially YouTube, which shows up regularly as a source in AI answers, particularly in Google’s AI features. Podcasts help because transcripts and show notes turn your expertise into text AI can read, and every guest spot is another mention.
Social is more mixed. LinkedIn and Reddit get cited directly. Platforms like Instagram and TikTok are cited less, but they drive branded search and awareness, which feeds into everything else.
It’s a big part of why we invest so heavily in our podcast and YouTube channel at Exposure Ninja.
How much do you tend to spend on digital PR? What’s realistic for a small business?
Asked by Hilary Rayner
I’d be wary of giving a number, because it varies so much by sector, how aggressive you want to be (or need to be in your market) and whether you’re working with an agency or doing it yourself.
For a small business, I’d put time in before money. Respond to journalist requests (#journorequest on social, or services like ResponseSource and Qwoted), pitch local and trade press, and put yourself forward for podcasts in your industry.
If you do hire help, ask how they’ll decide which publications to target. The right answer now includes “the ones AI is already citing for your prompts.”
We are an events venue – does longform from other websites contribute to being shown in AI answers like SEO?
Asked by Rebecca Gallally
Yes, very much so. For venues, AI leans heavily on third-party content: “best wedding venues in [area]” round-ups, local guides, wedding and events blogs, venue directories like Hitched and Guides for Brides, and of course Google reviews.
Run your prompts and look at which of those sites are being cited. That’s your outreach list.
Then make sure your own site answers the fan-out questions planners actually ask: capacity for different layouts, rough pricing, catering options, accessibility, parking, transport links and availability. If AI can’t find those facts on your site, it will take them from somewhere else, or skip you.
For B2B spaces where classic recommendations articles etc. are less common, what are some other options for increasing AI-visible reputation?
Asked by Kat Valentine
B2B has more options than people think. Review platforms matter a lot (G2, Capterra, Clutch or Trustpilot, depending on your sector), along with industry associations, awards, partner and integration directories, and analyst or trade reports.
Podcasts, webinars and speaking slots all create content AI can read, especially where there are transcripts or show notes. Co-published case studies with clients carry weight. Original research that trade press go on to quote is one of the best long-term plays.
I’d also create honest comparison and “alternatives” content on your own site, since those are high-intent prompts in B2B.
Your website and content
With prompt fan out — does that mean single page, single Q/A is still best (but many of them)? Or have fewer pages that answer many many Q/As?
Asked by Sophie Bryant
Neither extreme, really. I’d aim for one strong page per core topic or intent that answers the main question and the natural follow-ups together: what it costs, who it’s for, how it compares, what the downsides are. Then add supporting pages only where a question deserves real depth.
Hundreds of thin pages each answering one tiny question isn’t a good idea. It’s a poor experience, and it risks falling foul of Google’s scaled content policies.
A good test: would a customer want these answers on the same page? If they’d naturally ask them in the same conversation, they probably belong together.
If you’re in the Saas space or similar, where you have complex FAQs specifically about your products, you may build out a knowledge hub / FAQ section on your website.
Tips on rebuilding the fan-out page without it looking like it’s built for search?
Write it for the customer, as if you’re answering their questions on a sales call. Take the actual questions from sales and support, use natural headings, and give the answer first before the detail.
Include the things only you can say: real prices or ranges, examples or case studies, your opinion, and who it’s not right for. That’s what makes a page useful to people, and it’s also what AI finds most worth citing.
Tables and comparisons help both readers and AI. My test is simple: if you’d be happy to send the page to a prospect, you’re fine.
If we’re re-designing a website or re-positioning the brand, where do you start when you know you won’t have any meaningful stats from the prompts research?
Asked by Rachel Carr
Start with the positioning, because AI needs that anyway. Write a plain statement of what you do and who for, and make sure the new site says it clearly.
Then run a baseline on unbranded prompts for the category you’re moving into. Even if you don’t appear yet, that tells you which competitors AI recommends and which sources it’s using. That becomes your target list.
Build your prompts from sales calls and customer research rather than stats. On the technical side, make sure content is crawlable without heavy scripts, structured data is in place, redirects are mapped, and you’re not blocking AI crawlers. When you launch, update your third-party profiles in the same week so the old positioning isn’t hanging around.
We’re about to rebrand and therefore have a whole new website. How can we build this in right now?
Asked by Laura Le Maire
Great timing, because it’s much easier to build in now than to retrofit.
On the build: make sure content is in the HTML rather than hidden behind scripts, add Organization structured data with sameAs links to all your official profiles, write a clear about page, and map every redirect. Check your robots.txt allows the AI crawlers you want.
On the rebrand itself, AI needs to connect the old name with the new one. Say “formerly [old name]” on your about page, update every third-party profile and directory in the same week, and use your launch PR to get coverage that mentions both names. Otherwise AI can end up describing the old brand for months.
Is it important to have Bing Webmasters set up as well as Google Search Console?
Asked by Jennie Hawkins
I’d set it up. It’s free and takes a few minutes, as you can import your sites straight from Search Console.
Bing’s index powers Copilot, and Bing Webmaster Tools now has an AI Performance report showing how often your pages are cited in Copilot and Bing’s AI summaries, along with some of the queries that triggered those citations. It’s the only first-party AI citation data any of the big platforms give you right now.
If accuracy is wrong in AI how do you amend it?
Asked by Chloe Griffin
You can’t edit the answer directly, so you fix the sources it’s drawing from.
Start by finding where the wrong information is coming from. Perplexity is the easiest place to see this because it lists its sources, and ChatGPT shows sources when it searches. Nine times out of ten it’s an old page on your own site, an out-of-date directory listing, an old press article, or a profile like LinkedIn, Crunchbase or Wikipedia.
Fix your own site first with a clear, current statement of the facts. Then update the third-party listings and ask publishers to correct old articles. You can use the feedback buttons in ChatGPT too, but I wouldn’t rely on them.
Answers that pull from live search update faster, often within weeks. What a model “knows” from training takes longer to update.
I work for a brand with the same name as another brand in the same sector (fashion), which means AI can confuse the two. What approach would you take to tackle this? Both brands sell clothing but for different customers (high-end ladies’ clothing vs. clothing for Gen Z gamers).
Asked by Ellie Drinkwater
This is an entity problem, so the aim is to make it very easy for AI to tell you apart.
Use a consistent descriptor alongside your name everywhere, for example “[Brand], the luxury womenswear label from [place]”. Add Organization structured data to all your official profiles, claim your Google Knowledge Panel if you have one, and make your about page very clear on founding, location, founders and who you serve.
Then run the prompts where confusion happens and look at which sources are causing it. Ask those sites to use your descriptor. Real customers usually add context to their questions (“luxury dresses”), so focus on owning those prompts rather than the bare brand name.
Advice by type of business
I wonder if there is anything media companies specifically should keep in mind for optimising AI search? SEO looks a bit different for us, so I wonder if AEO is different too
Asked by Zara Puckrin
Media is a really interesting case, because you’re usually the source AI cites rather than the brand it recommends. I’d measure citation share on the topics you cover, and whether AI credits you by name when it uses your reporting.
A few things I’d focus on. Make your original reporting, data and expert commentary clearly attributable, with proper bylines, author pages and article structured data. Watch for syndicated or scraped copies being cited instead of your original. Freshness matters a lot for news-style prompts.
The big topic for media companies is making a deliberate decision on AI crawlers. Blocking training bots and blocking the bots that fetch pages for live answers are two different decisions and lots of publishers have blocked both without realising they are trading off visibility as a result.
Is AI visibility really achievable for a small / micro business with minimal budget / time?
Asked by Simon Hague
Yes, and in some ways it’s easier for small businesses, because the specific prompts your customers use (“best accountant for freelancers in Leeds”) are far less crowded than the big generic ones.
You can run the twenty-prompt check, then the basics go a long way: a plain statement on your homepage and about page of what you do, who for and where; a well-kept Google Business Profile; a steady flow of reviews; and getting yourself listed on the directories and round-ups that keep showing up as sources in your prompt results.
As a younger company in our industry, can we make good ground in AI search by having the right content for the right search queries or like SEO, do longstanding competitor websites carry a lot of weight simply due to longevity
Asked by Nicola Cook
You can absolutely make good ground. Established brands do have an advantage, but it comes from the years of mentions, reviews and coverage they’ve built up rather than from the age of their website.
AI is looking for the best answer to the question and for evidence that others rate you. A younger company can win on specific prompts, especially the long, detailed ones, where incumbents often have generic content that doesn’t answer the question properly.
I’d pick the niche or customer type where you’re the best option, build content that answers every part of that question, and put effort into third-party proof early. The broad category prompts will take longer, and that’s fine.
For a small start up brand, how do you build the proof?
Asked by Vicky Williams
Start collecting reviews from your very first customers on Google and the review platforms that matter in your sector. Turn happy customers into case studies with real names and results, with their permission.
Founder visibility works well for start-ups: podcasts, LinkedIn, speaking at small events. Pitch yourself into niche roundups and “best new” lists, and enter industry awards, even small ones.
On your own site, be specific. Real numbers, credentials, who you’ve worked with, case studies.
Have you found you’ve had more success in ecom vs b2b/lead gen or vice versa with AI search?
Asked by Jamie Bray, Emma Greenwood
We’ve had success in both, but they work differently and show up differently in the data.
Ecommerce results often appear in revenue faster, because the path from AI answer to purchase is shorter. ZUGU saw a 243% increase in AI platform traffic, and The Ordinary is now appearing in AI Overviews and ChatGPT for its hero product.
In lead gen and B2B, the influence is more hidden. People research in AI, then search your name or come back direct, so it shows up in branded search, “how did you hear about us” answers and sales conversations. Elite Renewables went from 84 to 337 monthly leads, many of which had touchpoints with AI.
ChatGPT ads and working across teams
I know it’s early days but is there any value in ChatGPT ads? Or is it simply more preferable to work on getting AI models to recommend your brand in its answers?
Asked by Megan Smith
I’d prioritise getting recommended in the answers. That’s where the lasting value is.
ChatGPT ads launched in the UK in June 2026 and are still a pilot. OpenAI says ads don’t affect the answer itself, so the two are separate channels. The ad sits beneath the answer, while the recommendation is in it.
If you have spare budget and your category fits, a small, well-tracked test is worth doing so you learn early. I just wouldn’t move money out of organic AI visibility to fund it.
I found out that you can now have an ad of your brand on ChatGPT, would you saw that is a good investment?
Asked by Beatriz Marques
It’s early days. ChatGPT ads launched in the UK in June 2026 in a pilot phase. They only show to users on the Free and Go plans, appear beneath the answer, and are supposedly targeted on the context of the conversation though I have seen ads show on irrlevant conversations.
If your category fits and you have budget to test, I’d try it with a ring-fenced amount. Don’t expect it to be cheap, because some sectors are already seeing high CPMs.
The important thing to know is that OpenAI says ads don’t influence the answer itself. Paying for an ad doesn’t get you recommended in the response. Organic visibility is still where the lasting value is. This is all moving fast, so I’d check the latest before committing budget.
Have you seen any businesses organic ChatGPT referrals grow significantly post spending on ChatGPT ads and then decrease significantly when they stop?
Asked by Camilla Lloyd-Davies
I haven’t seen this but I also don’t have enough data to give real insight. I’d be sceptical of anyone who has a strong answer yet. ChatGPT ads only launched in the UK in June 2026, and they’re still in a pilot phase.
OpenAI’s position is that ads don’t influence the answers themselves, so there shouldn’t be a direct organic effect.
Should AI search and SEO teams be part of the same team? What if they do not talk to each other?
Asked by Miren Ortigosa Arraiago
Yes. Optimising for ChatGPT, Gemini, Claude or whichever AI platform is a set of platforms you want to optimise to show up in. Google is the original platform we’ve been optimising to show up on. Now there are more. The SEO team have the raw skills needed to do this work for other platforms. The technical foundations and content work overlap a lot, and when two teams work separately they can end up producing competing content or chasing conflicting metrics.
AI visibility also needs PR, brand and product marketing, so it’s bigger than SEO alone. At Exposure Ninja we’ve built AI Search into how our SEO and Digital PR teams already work, rather than setting it up as a separate silo.