AI is sending traffic to your site. But is it traffic you can actually use?
You’ve seen it in your analytics: visits coming in from chatgpt.com, perplexity.ai, gemini.google.com, or claude.ai. Good news, on the surface. Your brand got mentioned in an AI answer, someone clicked through, and traffic followed.
Then comes the question most marketers get stuck on: is this a genuinely interested visitor, or a bot scraping your pages to train its language model? You can see the traffic. You just can’t see what it’s worth. A black box you’re steering on, and reporting on, without being able to look inside.
And that difference matters more than it looks.
Why this suddenly matters
AI search engines aren’t a niche channel anymore. ChatGPT, Perplexity, Gemini and Claude are used daily to answer questions that used to go to Google, including questions about vendors, tools, and B2B solutions. For marketers, that means a new part of the funnel they need to be able to explain: are we showing up, and is it paying off?
The problem is that most web analytics tools weren’t built for this. They see a session, a referrer, a user-agent, and categorise it as best they can. The distinction between “a language model indexing your content” and “a person who just got an answer and clicked through” usually isn’t made at all.
Cloudflare Radar showed just how large that problem has become: in June 2026, bots accounted for 57.5% of all HTML traffic on the web, humans for the rest. For the first time in the internet’s history, traffic is no longer predominantly human.
Two kinds of “AI traffic”, and they are not the same
Broadly, two streams arrive under the banner of “AI traffic”:
Crawlers and training bots. These are the bots run by OpenAI, Anthropic, Perplexity and others that visit your site to index content or train models. They generate pageviews, but zero buying intent. They don’t convert, they don’t read on, and they never come back as a lead.
Real, AI-referred visitors. These are people who asked ChatGPT, Perplexity, Gemini or Claude a question, got an answer that mentioned your brand or content, and chose to click through to your site themselves. This is the traffic your content, SEO and GEO investments were actually meant to earn.
The gap in scale between those two streams is large. Cloudflare Radar recorded that, in the week of 25 May to 1 June 2026, ClaudeBot crawled a site more than 11,000 times for every single visitor it actually referred. For GPTBot that ratio was roughly 1,276 crawls per referred visit, and for PerplexityBot around 111 to 1. For comparison, traditional Google search sits at roughly 5 to 1.
In other words: the overwhelming majority of what gets labelled “AI traffic” is traffic a marketer can’t use for anything at all.
Wikipedia is the clearest public example of what that actually costs. The Wikimedia Foundation reported that bandwidth for downloading multimedia content grew by 50% between January 2024 and early 2025, driven largely by AI crawlers pulling in bulk to train models and answer questions, rather than by more human readers. Automated requests already accounted for at least 65% of its most resource-intensive traffic, even though bots made up only around 35% of pageviews.
By March 2026, the Foundation said it was blocking or throttling roughly a quarter of all crawler requests that ignore its access policies, on some days amounting to billions of requests. If a site with Wikipedia’s scale and engineering resources is straining under this, the average B2B website has no way of telling the difference on its own.
Why this distinction affects your reporting, and your budget
Without that distinction, one of two things happens, and both are costly.
You count bot traffic as a result. Your content and GEO work looks like it’s paying off on paper, while in reality you’re mostly feeding training data to a language model. You report numbers to your manager or board that have no relationship to pipeline whatsoever.
Or you write off all AI traffic as “noise” and ignore it entirely, including the visitors who genuinely matter. That’s a shame, because this group tends to run warm: they had a specific question, got an answer that surfaced your brand, and chose to read further on their own.
Picture a marketer who’s spent months investing in content and GEO tactics to get surfaced in AI answers more often. Without a way to separate bot from visitor, that person either reports a success that isn’t real, or misses the proof that the investment is actually working. Neither version is one you want to steer on, let alone defend upstairs.
This is exactly the problem Leadinfo has helped B2B marketers solve for “ordinary” website traffic for years: 98% of your visitors leave without converting, and without the right layer of insight, you don’t know who the remaining 2% actually are. AI traffic has the same problem, just harder to see through.
The fix: AI Traffic Insights
That’s why we’ve added AI Traffic Insights to the Insights Inbox in Leadinfo.
Using our own database, Leadinfo recognises bots from all the major AI search engines and language models, and separates them from real, relevant visitors who click through to your site from an AI answer. Not an educated guess based on a single user-agent string, but recognition based on continuously updated bot data.
The result: in the Insights Inbox, you see exactly how much of your “AI traffic” is actually people, which AI source it came from (ChatGPT, Perplexity, Gemini, Claude), and which company is behind it. No noise, just signal.
There’s a second layer of value here, too. Seeing a consistent rise in real visitors from one specific AI source is a practical signal that your brand is being surfaced in those answers, in other words, that you’re visible in places you don’t yet have a dashboard for. No separate citation tool required: it follows directly from the distinction you can already make today.
We already recognise a broad and growing number of AI bots, and because AI traffic is expanding fast, we’re extending that recognition continuously. We’ll share concrete benchmarks as soon as we can back them up, precisely because we’d rather give you facts than an impressive-sounding number without the data behind it.
What this means for you as a marketer
Every euro and every hour you spend on content, SEO, GEO or ads should turn into traffic you can actually act on. With AI Traffic Insights, you’re no longer steering on a black box, but on facts: who you’re attracting, from which AI and AI Ad source like ChatGPT ads, and whether that’s a bot or a company that might actually buy.
That doesn’t just sharpen what you do differently tomorrow. It also makes you more credible to everyone you report to, because your numbers hold up, down to the source.
Want to see for yourself how much of your AI traffic is bot, and how much is a genuine visitor? Start your free trial: live in 5 minutes.
