ChatGPT sends most of the AI traffic our clients see. So why track Gemini, Perplexity and the rest? For the same reason SEOs have always kept an eye on Bing: the audience is spread wider than the traffic suggests, and each platform tells you something different about what people want.
Tom LightIt’s a question we hear in almost every AI search conversation with our clients, from insurance and travel brands to automotive, property and charity organisations. If ChatGPT is where the traffic comes from, why not just track ChatGPT?
It’s a fair question. Across our clients, ChatGPT typically accounts for somewhere between 75% and 90% of the traffic that arrives from AI tools. On those numbers alone, it would be easy to treat every other large language model (LLM) as a rounding error, and to build your whole generative engine optimisation (GEO) plan around one platform.
But traffic is only part of the picture. People increasingly use more than one AI tool, the tools themselves find and choose sources in very different ways, and the same brand can be highly visible in one and almost invisible in another. By the end of this article, you should know why we track several AI models, how the big three differ, and how to use that data without spreading your effort too thin.
Let’s start with what’s true today. ChatGPT is by far the biggest source of AI referral traffic, both for our clients and across the web.
Statcounter’s global data shows ChatGPT drove just over 78% of AI chatbot referrals to websites in March 2026. That’s still a commanding lead, but it’s the first time its share has dropped below 80%, down from just over 84% a year earlier.
Over the same period, Gemini’s share almost quadrupled and it overtook Perplexity to become the second-largest source of AI referrals. Claude grew nearly tenfold from a very small base. The AI referral market is still dominated by one platform, but it’s fragmenting.
Referral share only counts the clicks. Usage tells a broader story.
Similarweb’s 2026 Generative AI Landscape found ChatGPT’s share of generative AI web visits fell from around 76% in June 2025 to around 52% in May 2026, while Gemini grew strongly. It estimates around 494 million people use ChatGPT, 310 million use Gemini and 86 million use Claude, but only around 29 million use all three.
Other research points the same way. a16z’s consumer AI report found around one in five weekly ChatGPT web users also used Gemini in the same week. Add Microsoft Copilot sitting inside Windows and Office, Grok inside X, and AI built into Google Search itself, and it’s clear your audience isn’t loyal to a single tool. They use whichever one is in front of them.
That matters because the person asking ChatGPT about your category today may ask Gemini tomorrow, and the answer they get could be very different.
All three tools follow a similar loop: read the question, search, pick sources, write an answer. But the way each one does it is different, and that changes which brands and pages get cited.
ChatGPTChatGPT decides for itself whether a question needs a web search. When it does search, it draws on Bing and OpenAI’s own index, built by its OAI-SearchBot crawler. It tends to cite fewer sources than the others, but leans more heavily on each one it uses, and favours encyclopaedic and established media sources.
Gemini is grounded in Google’s own search index. It often breaks a question into several smaller searches, runs them in parallel and combines the results into one answer. Its behaviour sits close to Google’s other AI features, such as AI Overviews and AI Mode, although the sources still differ.
Perplexity is built as an answer engine. It searches the web for almost every question, uses its own index alongside live search, and always shows its sources. Of the three, its citations are the closest match to traditional search rankings, and it leans heavily on community sites such as Reddit.
These differences show up clearly in the number of sources each one cites.
Because each tool searches differently, they rarely agree on which sources to cite.
A study of 161,286 prompts found ChatGPT, Gemini, Perplexity and Google AI Overviews shared only around 17% of their cited sources for the same prompt. Just 3.8% of sources were cited by all four.
BrightEdge found the overlap in cited sources between any two AI search surfaces ranged from 16% to 59%. Even Gemini and Google’s AI Mode shared only around a quarter of their sources.
Answers also change from one run to the next. Research on citation variability shows the same question asked twice on the same platform can produce a different set of sources.
We see the same thing in our own client data. We track a fixed set of prompts for each client across several AI platforms every week, using the same questions each time so the results can be compared. For one client, visibility looked very different depending on where you looked.
They were strong in Google’s AI Overviews and Gemini, but barely visible in ChatGPT, which was also where most of their AI traffic came from. Tracking ChatGPT alone would have shown a problem with no context. Tracking Google’s AI alone would have hidden the problem entirely.
If you only track one AI model, you’re only seeing one version of how AI describes your brand.
None of this is new to SEO. For years, Google has been the focus for almost every UK search strategy, and for good reason. But we’ve always kept an eye on Bing.
Bing’s share of UK search looks small overall, at around 6% across all devices. On desktop, though, it’s closer to 12%, thanks to Windows, Edge and workplace machines. For some audiences, particularly older users and people searching from work, Bing’s share is far higher than the average suggests.
SEOs have long worked on the principle that what works for Google works for Bing. Crawlable sites, clear content, strong authority and a trusted brand perform well on both. You don’t need a separate Bing strategy, but you do need to know how you’re performing there.
The same applies to AI search. The fundamentals are shared across every model: clear answers, real expertise, and a brand that’s mentioned and trusted across the web. You don’t need a different strategy for each AI tool. But monitoring several of them gives you far more data, and shows what each model thinks users want for each topic.
There’s one twist. Because ChatGPT draws on Bing’s index, Bing now matters more than its search share suggests. If your pages aren’t indexed in Bing, ChatGPT may struggle to find them, however well you rank on Google.
ChatGPT is where the traffic is today, so that’s where we focus. But tracking only ChatGPT is like doing SEO without ever looking at Bing. You miss part of your audience, and you miss what the other platforms are telling you about your topic.
Before deciding where to focus, check your own data. This is the first thing we do for every client, and it’s where our 75% to 90% ChatGPT figure comes from.
In GA4, open Reports, then Acquisition, then Traffic acquisition. Change the dimension to Session source, then add a filter using a regex that matches the main AI platforms:
.*chatgpt\.com.*|.*perplexity.*|.*gemini\.google\.com.*|.*copilot\.microsoft\.com.*|.*claude\.ai.*
This shows how many sessions, key events and how much revenue each AI tool sends you. Two caveats from experience: some AI traffic arrives with no referrer and is counted as direct, and Google’s AI Overviews and AI Mode traffic is counted as Google organic, so neither will appear here. Treat these figures as a floor, not the full picture.
Your own split matters more than any industry average. If Gemini or Perplexity sends a larger share for your brand than the global figures suggest, that’s a strong reason to track it closely.
Tracking more than one AI model isn’t about spreading effort evenly. It’s about getting better information. In practice, it helps in four ways.
It shows where to prioritise. Weight your reporting by where the traffic and audience actually are. For most brands today, that means ChatGPT first.
It warns you early when things shift. Gemini’s rise over the last year happened quickly. Brands tracking it saw the change coming.
It reveals content and PR gaps. Different models favour different sources. If you’re cited in Perplexity but not ChatGPT, that points to specific sources or content types you’re missing.
It separates signal from noise. AI answers vary from run to run. Seeing the same pattern across several models gives you much more confidence that it’s real.
For most clients, we track a core set of AI platforms: ChatGPT, Google AI Overviews and AI Mode, Gemini and Perplexity, adding Claude or Copilot where the audience justifies it. A few principles guide how we use the data:
Report by platform and by topic, not as one blended visibility score.
Weight priorities by traffic and audience, so ChatGPT gets the most attention while others are monitored.
Cover the technical basics for every model. Make sure AI crawlers such as OAI-SearchBot and PerplexityBot can access your site, and check your coverage in Bing Webmaster Tools as well as Google Search Console.
Focus on fundamentals that work everywhere: clear answers, original expertise and a strong brand presence across the web.
ChatGPT is the most important AI platform for most brands today, and it should get most of the attention. But it isn’t the only one your audience uses, and it isn’t the only one shaping how they see you. Tracking several models costs relatively little and gives you a far more complete picture, just as checking Bing always has.
If you’d like to see how your brand shows up across AI platforms, talk to us. Our GEO discovery shows how each AI engine currently sees your brand, where competitors appear instead and what to do next. To see where your brand sits on the wider journey, read AI search maturity: where brands are now, and how to move up. And if AI Overviews are already affecting your traffic, see our guide to why your organic traffic is dropping.
Our GEO discovery shows how ChatGPT, Gemini, Perplexity and Google’s AI features describe your brand, where competitors appear instead and what to do next.
