Tracking Traffic from AI Tools: Technical Considerations

Summary

Siteimprove now provides built-in AI Referral Traffic classification and a pre-configured AI Traffic Overview dashboard, which identify visits from known AI-powered chat tools and assistants using referrer-domain matching. This gives organizations a direct, in-platform way to see AI-driven traffic, rather than relying only on indirect methods. That said, referrer-based detection has inherent limits: AI tools that don’t send a referrer header, links that are copied and pasted, or tools not yet on Siteimprove’s reference list will still show up as direct or unclassified traffic.

Overview

With the rapid growth of AI-driven assistants and search tools (such as ChatGPT, Copilot, Perplexity, Claude, and Gemini), many organizations want to understand how these tools interact with their websites and whether traffic from them can be measured.

In January 2026, Siteimprove introduced a new traffic source classification, AI Referral Traffic, and in April 2026 followed up with a pre-configured AI Traffic Overview system dashboard. Together, these give meaningfully better, though still partial, visibility into AI-driven traffic. This article outlines how AI tools interact with websites, how Siteimprove's AI Traffic feature works technically, and what challenges and limitations remain.

How AI tools interact with websites

Crawlers / Bots

Many AI vendors operate crawlers (e.g., GPTBot, PerplexityBot, AnthropicBot) that scan and index web content. These appear in server logs with a recognizable User-Agent string. This activity represents content ingestion, not end-user visits.

User interactions

When AI tools provide links to users, people may either:

Click the link directly, which should generate a normal pageview in analytics.

Copy and paste the link into their browser, which typically shows up as direct traffic with no referrer information.

How Siteimprove's AI Traffic feature works

AI Referral Traffic classification

Siteimprove classifies a visit as AI Referral Traffic when its referrer domain matches a known AI chat or assistant domain — using the same domain-matching logic already applied to Search Engine and Social Media traffic. Siteimprove maintains a central, expandable reference list of AI organizations and their referring domains, including:

ChatGPT — chatgpt.com (OpenAI)

Gemini — gemini.google.com (Google / DeepMind)

Claude — claude.ai (Anthropic)

Microsoft Copilot — copilot.microsoft.com (Microsoft)

Perplexity AI — perplexity.ai (Perplexity)

Meta AI, Grok, DeepSeek, and other AI assistants, with new domains added over time

When a visit's referrer matches one of these domains, it is tagged as AI Referral Traffic rather than falling into generic external referrer or direct traffic buckets.

Where to find it

The Traffic Sources section in Marketing Analytics shows a breakdown of AI referrals, and you can build segment filters on this traffic source for use across any view that supports filters.

AI Traffic Overview dashboard

A pre-configured system dashboard (Dashboards → AI Traffic Overview) combines this traffic classification with content, SEO, and engagement data to help answer questions such as: Is AI driving traffic to your website? Which pages are appearing in AI-powered search results? Which AI platforms — ChatGPT, Claude, Perplexity, Gemini, and others — are sending visitors to your site? Some sections require an active Siteimprove Analytics install, an active crawl, and an SEO Enterprise subscription or trial for the full experience; EU datacenter accounts may need to build the dashboard manually rather than copying the standard template.

This feature does not introduce a new tracking mechanism or new data collection — it applies Siteimprove's existing analytics infrastructure (filters, dashboards, and policies) to referrer data that is already being collected, so no additional personal data is gathered.

Tracking challenges

No referrer information

Many AI tools do not send referrer headers when a link is clicked. Analytics platforms, therefore, cannot identify the visit as “coming from ChatGPT” or similar. Because AI Referral Traffic classification depends entirely on a referrer header being present and matching a known domain, visits from AI tools that omit the referrer still land in direct traffic, unclassified.

Server-side fetching

Some tools (e.g., Perplexity) fetch and cache site content on their servers. Analytics may register the crawler's request, not the user's subsequent interaction.

No standard identifiers

There is no universal parameter, such as utm_source=chatgpt, applied automatically. Only publishers in formal partnerships with AI vendors may see consistently tagged links, but this is rare. AI Referral Traffic classification helps here for the AI tools Siteimprove already tracks, but emerging or unlisted tools won't be classified until their referring domain is added to the reference list.

Copy/paste behavior

User behavior often results in “direct/unknown” traffic, indistinguishable from manually typed URLs — this is unaffected by AI Referral Traffic classification, since there is no referrer to match against.

What you can track

AI Referral Traffic (Siteimprove Analytics)

View the AI referral breakdown in the Traffic Sources section of Marketing Analytics, segment on it anywhere segment filters are supported, or use the pre-configured AI Traffic Overview dashboard to see which AI platforms are sending visitors and how they engage with your content.

AI crawlers in server logs

Check for known User-Agents (e.g., GPTBot) to measure indexing activity.

Direct traffic spikes

Unexplained increases in “direct traffic” may still correlate with your content being referenced in AI tools that don't send a referrer.

Custom UTM parameters

You can create unique links for testing (e.g., https://yourdomain.com/page?utm_source=chatgpt&utm_medium=referral&utm_campaign=ai_test). If these appear in your analytics, it suggests some AI-driven visits. Important: anyone (including other crawlers) can reuse such links, so this method is not foolproof.

Current limitations

AI Referral Traffic detection is based on referrer-domain matching against a known list of AI organizations. It does not capture AI-driven visits where no referrer is sent, or where the tool isn't yet on Siteimprove's reference list, and referrer behavior from AI tools may change over time as vendors update their products.

There is still no way to reliably measure how much or what type of content AI tools are accessing beyond crawler log activity — that can only be inferred indirectly.

Tracking is further blurred when multiple tools or crawlers reuse the same tagged URLs.

Some AI Traffic Overview dashboard sections require an SEO Enterprise subscription, and EU datacenter accounts may need to build the dashboard manually.

Conclusion

As of 2026, Siteimprove offers direct, built-in support for identifying AI Referral Traffic and a dedicated AI Traffic Overview dashboard, giving organizations meaningfully better visibility than the fully indirect methods previously required. That said, this tracking is still referrer-based: it cannot capture every AI-driven visit, and cases with no referrer, copied-and-pasted links, or emerging/unlisted AI tools will still appear as direct or unclassified traffic.

Organizations should treat AI Referral Traffic numbers as a reliable floor, not a ceiling, on their real AI-driven traffic — and continue combining it with crawler-log monitoring and UTM experiments for a fuller picture. As AI-driven discovery tools mature, Siteimprove's reference list and detection coverage are expected to expand; site owners interested in this topic should monitor Siteimprove's release notes and Help Center for updates on AI referral tracking.