How to Track AI Traffic to Your Website (ChatGPT, Perplexity, Gemini, and More)

AI platforms are sending real traffic to websites right now, and that traffic converts at rates that can make your usual channels look average by comparison. ChatGPT-referred visitors convert at roughly 15.9%, versus about 1.76% for Google organic search. The problem is that a big chunk of this high-value traffic is landing in your analytics as "direct," completely unidentified. This guide walks through how to track it properly, platform by platform, method by method.

Why standard analytics misses so much AI traffic

Before getting into setup, it's worth understanding exactly why the problem exists. Three separate things cause AI-sourced traffic to be undercounted:

Copy-paste behavior. When ChatGPT includes a URL in its response, a lot of users copy that URL and paste it into their browser rather than clicking through directly. A pasted URL carries no referrer information. It arrives at your site looking exactly like someone who typed your URL from memory. In your GA4, it shows as direct.

Mobile app traffic. ChatGPT's iOS and Android apps, and Gemini's mobile app, don't consistently pass referrer headers when users tap outbound links. This traffic also shows as direct. Given that mobile accounts for the majority of ChatGPT sessions, this is not a small problem.

Deliberate noreferrer links. Some AI platforms explicitly set noreferrer on outbound links, meaning no referrer data passes even when a user clicks directly. Behavior varies across platforms and changes as these products update their link handling.

"Industry estimates suggest 20 to 40 percent of AI referral traffic is misattributed as direct visits, due to mobile app behavior, noreferrer links, and inconsistent UTM implementation across AI platforms."

That 20 to 40 percent gap is significant. For any site doing meaningful volume, it represents a real blind spot in your understanding of what's driving your business. The rest of this article is about closing that gap.

Method 1: GA4 custom channel group (do this today)

This is the minimum viable setup and it's free. Go to Admin > Data Display > Channel Groups in your GA4 property, then create a new Channel Group.

Add a channel called "AI Traffic" and under Channel Conditions, set source matches regex to:

chatgpt\.com|openai\.com|perplexity\.ai|gemini\.google\.com|claude\.ai|copilot\.microsoft\.com|you\.com|grok\.x\.ai|meta\.ai|phind\.com

Once this channel group is active, you'll see AI Traffic as a separate channel in your Traffic Acquisition reports. You can segment conversions, engagement rate, and session duration by this channel to understand how AI-sourced visitors behave compared to other sources.

You should also check for UTM-tagged traffic separately. ChatGPT Search automatically appends utm_source=chatgpt.com to some outbound links. Filter your acquisition reports for this UTM source to catch ChatGPT Search sessions that the referrer regex might miss.

"ChatGPT referral traffic converts at 15.9 percent compared to just 1.76 percent for Google organic search. Visitors from AI recommendations arrive with higher intent because they've already received a qualified recommendation before they click."

That conversion rate difference is why getting this tracking right matters. If you're receiving any volume of AI-referred traffic and measuring it poorly, you're likely making suboptimal decisions about where to invest your marketing budget.

Method 2: Server log analysis and crawler monitoring

Here's where most marketers haven't looked yet, and where some of the most useful insights live. AI platforms don't just send users to your site. They also send crawlers to read your content before deciding whether to include it in answers. By monitoring which AI crawlers are visiting your pages and correlating that activity with subsequent referral traffic, you get a leading indicator of AI-sourced visits before they show up in your analytics.

The crawlers to look for in your server logs:

• GPTBot and OAI-SearchBot (OpenAI / ChatGPT)
• ChatGPT-User (used when ChatGPT accesses content in real-time during a conversation)
• PerplexityBot (Perplexity)
• Google-Extended (Google's crawler for Gemini and AI Overviews)
• ClaudeBot (Anthropic / Claude)
• Applebot-Extended (Apple AI features)

If a page on your site suddenly starts getting heavy GPTBot visits, there's a reasonable chance it will start appearing in ChatGPT answers shortly after, or already is. Spotting this pattern lets you correlate your content changes with AI crawler behavior and eventually with citation frequency.

One data point worth sitting with: researchers found that for every one visitor Claude sends to a website, its crawler visits approximately 500,000 times. For ChatGPT, the ratio is about 3,700 crawler visits per referred visitor. For Perplexity it's about 700 to one. This means you may have substantial AI crawler activity with relatively few direct referral clicks, and that the clicks that do come through are disproportionately valuable.

Method 3: What GA4 is doing with Perplexity Comet and ChatGPT Atlas specifically

This is worth a specific mention because the behavior is evolving. Perplexity's desktop browser (Comet) typically passes referrer information, so those visits are more likely to show up accurately in GA4. ChatGPT's standalone browser (Atlas), on the other hand, often masks its origin, and sessions from it tend to blend into direct traffic.

"Perplexity Comet typically passes referrer information and appears as a clear referral source, while ChatGPT Atlas often masks its origin, blending in with direct traffic."

The implication is that your GA4 AI traffic segment is likely closer to the full picture for Perplexity visits than for ChatGPT visits. For ChatGPT specifically, the direct traffic gap is probably larger than you'd expect from the platform's market share alone.

Method 4: Post-purchase surveys

Given how much AI-sourced traffic arrives as direct, behavioral analytics alone can't close the gap fully. Post-purchase surveys asking "how did you first hear about us?" with "AI assistant recommendation (ChatGPT, Claude, Perplexity, etc.)" as a specific option give you first-party attribution data that no analytics tool can provide.

This might sound low-tech, but for high-value conversions, it's often the most accurate signal you have. A customer who found your brand through a ChatGPT recommendation, then visited directly a day later, then converted has a journey that will never show up in your GA4 AI traffic channel. Only a survey can capture it.

Some ecommerce and SaaS businesses are reporting that when they add AI assistants as an explicit option in post-purchase attribution surveys, it shows up as a meaningful discovery channel far earlier than their behavioral analytics suggested it would.

Which AI platforms are sending the most traffic right now?

Current data (2025 to 2026) shows a clear hierarchy that's worth knowing for prioritization:

• ChatGPT dominates, accounting for 77 to 87 percent of all AI referral visits depending on the dataset and measurement approach. This is where most of the traffic volume sits.
• Perplexity sits at roughly 15 percent of AI referral visits and is notable for passing referrer data more consistently than other platforms. It's also growing fast.
• Google Gemini and AI Overviews account for most of the remainder. AI Overviews in particular appears in a rapidly growing proportion of Google searches, making it an increasingly significant traffic source.
• Claude, Copilot, Grok, and Meta AI each represent smaller shares currently but are growing. Claude is gaining traction with professional and research-oriented users specifically.

This hierarchy matters for prioritization. If you're starting from scratch on AI traffic optimization, ChatGPT and Perplexity are where you'll see the most immediate return on effort.

Reading your data once you have it set up

Once you have your tracking in place, here are the specific questions worth asking:

Is AI traffic growing as a share of total traffic? Even if the absolute numbers look small, the trend matters. A consistent upward trend in your AI channel segment, correlated with content or optimization changes you've made, tells you something real.

Which pages receive the most AI referral traffic? These are your most AI-friendly pages. Study what makes them different from pages that don't receive AI traffic and apply those lessons systematically.

What's the engagement and conversion rate of AI traffic versus other channels? AI-referred visitors almost always look different from organic search visitors on quality metrics. Higher pages per session, longer session duration, better conversion rates. Understanding this difference makes the business case for investing in AI visibility optimization.

Is there a lag between increased crawler activity and increased referral traffic? If GPTBot or PerplexityBot starts crawling certain pages heavily, look for referral traffic from that platform to increase over the following weeks. This correlation pattern helps you predict traffic changes before they show up.

For the complete picture, including how to connect AI traffic to actual revenue rather than just sessions, the detailed guide on attributing conversions from AI search covers that specifically. And for understanding the underlying visibility that produces this traffic in the first place, see the guide on how to get cited in ChatGPT and other AI search.

Common mistakes that produce misleading data

• Treating all direct traffic as direct. As discussed, a significant portion of AI-referred traffic arrives as direct. If your direct traffic has grown over the last twelve months without a clear explanation, AI is likely part of the story.
• Only tracking one AI platform. Different platforms serve different audiences and produce different traffic characteristics. Tracking only ChatGPT misses Perplexity's growing contribution.
• Ignoring crawler data. Referral traffic is a downstream consequence of crawler activity. The upstream side, which pages AI crawlers are visiting most frequently, gives earlier signals.
• Not tracking over time. A single snapshot of AI traffic tells you very little. The trend over three to six months, correlated with your content changes, is where the insight lives.
• Not having "AI assistant" as an explicit option in attribution surveys. If you're using post-purchase surveys but haven't added AI as a channel option, you're not capturing that data at all.

For the full picture on platform-by-platform visitor behavior, the companion guide on tracking visitors from ChatGPT, Perplexity, and Claude goes into more detail.