ChatGPT referrals to B2B websites nearly quadrupled: how to report AI visits by account
Demandbase Labs reported on August 12, 2026 that monthly ChatGPT-referred visits to B2B websites in its dataset rose 303%, from about 645,000 in June 2025 to 2.6 million in June 2026. Perplexity referrals declined, and Gemini and Claude stayed flat. For B2B teams, the useful step is to report AI-referred visits by account and landing page, then compare them with CRM progression, while accepting that referral data misses visits where the AI platform sends no referrer.
Key takeaways
- Demandbase Labs saw monthly ChatGPT-referred visits grow 303% between June 2025 and June 2026 across its customer base.
- Other AI platforms did not grow the same way in that dataset: Perplexity fell and Gemini and Claude stayed flat.
- Referral reporting measures visits, not citations, zero-click influence or automated crawlers.
- The valuable view is accounts and pages: which target accounts arrive from AI platforms and which pages they reach.
What Demandbase reported
Demandbase Labs analyzed more than 11 billion website visits and 75,645 campaigns across 1,584 Demandbase instances between June 1, 2025 and July 31, 2026. It counted attributable visits where a user clicked a link from an AI assistant.
Monthly ChatGPT-referred visits rose from about 645,000 in June 2025 to 2.6 million in June 2026. Volume more than doubled in May 2026 compared with earlier months. In the same data, Perplexity referrals declined and Gemini and Claude stayed flat.
Why account-level reporting is the point
Total AI referral sessions are a vanity number for enterprise B2B. A sales leader wants to know which target accounts are researching the category through an AI assistant, which pages they reach, and whether those accounts move toward an opportunity.
An account-based platform such as Demandbase can resolve a share of web visits to accounts. Combine that with GA4 referral segments and your CRM to see AI-referred engagement at the account level, instead of a single traffic line.
What referral data cannot see
- Visits where the AI platform strips or withholds referrer information.
- Platforms your analytics does not recognize as AI sources yet.
- Pages without your analytics or Demandbase tag.
- Citations and answers that influence a buyer without a click.
- Automated crawlers, which are not human visitors.
Set up the report
- In GA4, create a segment for sessions from known AI sources and keep the source list in a shared document so it can be updated.
- In your account-based platform, filter web activity to the same sources and list the accounts and landing pages involved.
- Reconcile the two views once a month and note the gaps rather than forcing the totals to match.
- Join the account list to CRM stage so you can see which identified accounts progressed.
- Review the pages these visits reach. If they land on thin pages, strengthen those pages first.
Metrics to track
| Metric | Why it matters |
|---|---|
| Identified AI-referred accounts | Shows which target accounts use AI assistants in research. |
| Engaged-session rate | Separates real research from bounces. |
| High-intent page visits | Shows whether AI traffic reaches product, comparison and pricing pages. |
| Form conversion rate | Compares AI-referred visitors with other sources. |
| Opportunity creation | Connects the traffic to pipeline in the CRM. |
Frequently asked questions
Does 303% growth mean AI search drives most B2B traffic?
No. The figure is growth in ChatGPT-referred visits within the Demandbase dataset. The report does not say what share of total traffic this is.
Should we report AI referrals to the board?
Yes, as a small, honest line with its limits stated, tied to accounts and pipeline rather than raw sessions.
Why do some AI visits appear as direct traffic?
When a platform sends no referrer, analytics cannot attribute the visit, so it can fall into direct. That is why referral counts understate AI influence.
Related service: Generative engine optimization. See also pricing and client results.
