AI traffic lands in the direct bucket
When a buyer clicks through from an AI assistant, most analytics tools see no clean source and file it as direct or none. The channel that drove the visit disappears.
Wave · Channel Affinity
Your buyer found you on ChatGPT. Does your attribution know it?
B2B buyers now start research inside ChatGPT, Perplexity, and Gemini, then arrive on your site. Most analytics stacks bucket that traffic as direct and lose the signal. Wave names each AI assistant as a first-class inbound channel and folds it into the channel prediction on every contact.
Quick answer
AI Referral Intelligence is a Wave capability that detects and classifies website traffic arriving from AI assistants, including ChatGPT, Perplexity, Claude, Gemini, and Copilot, as a first-class inbound channel rather than losing it in a direct or none bucket. Those AI referrals flow into the same Channel Affinity inbound prediction as UTM and referrer signals, so AI-referred contacts are scored on the same pipeline.
Key capabilities
- AI-assistant referral detection and classification
- ChatGPT, Perplexity, Claude, Gemini, and Copilot named distinctly
- First-class inbound channel in a B2B taxonomy
- Folds into per-contact channel predictions
- Per-tenant classification you control and extend
- Extensible taxonomy for new AI assistants
- Server-side, no tag manager changes
Last updated July 2026
The problem
The fastest-growing channel in B2B is invisible to your stack.
Buyer discovery has moved into AI assistants faster than any attribution model has kept up. Virtually no marketing team can answer how much of their pipeline comes from AI referrals, let alone which assistant sends their best-fit contacts.
Retrofits miss the signal
Most martech was designed before 2023 and is bolting on AI awareness after the fact. The signal class was never native, so it gets approximated and lost in the noise.
No contact-level view
Even teams that spot AI traffic in aggregate cannot tie it to a person. They cannot say this specific buyer researches through Perplexity, so they cannot act on it.
How Wave does it
Name every AI assistant, classify it, fold it into the prediction.
Wave reads each inbound arrival server-side, classifies the AI assistant that sent it, and routes that signal into the same channel model that scores every other arrival.
- 01
Capture the referrer server-side
Wave records each inbound channel touch, including the document referrer, with no change to your tag manager or marketing automation platform. The signal is read on Wave's side.
- 02
Classify against a B2B taxonomy
A canonical channel taxonomy names ChatGPT, Perplexity, Claude, Gemini, and Copilot as distinct first-class channels, alongside paid social, organic search, and partner programs.
- 03
Classify, your way
Wave recognizes each assistant against a per-tenant taxonomy you control and extend, so you can add new assistants as they emerge without waiting on us. Accurate, cost-efficient, and fully under your control.
- 04
Fold into the inbound prediction
AI-referred contacts flow into the same inbound channel prediction as UTM-based contacts, on one pipeline, with no separate reporting silo and no extra setup.
Where it fits
Part of Wave's person-level channel intelligence.
AI Referral Intelligence is the AI-assistant signal class inside Channel Affinity, one axis of Wave's person-level data layer. Channel Affinity predicts how a person arrives and where they respond, and AI referrals are now a native part of that arrival picture.
Your sources
Wave
HubSpot
Wave reads inbound touches from your existing systems through native connectors, classifies each one per tenant so your data never trains on anyone else's, and the resulting channel predictions write back to HubSpot today as contact properties (Marketo next). Every writeback is logged with previous value, new value, and the model version that produced it.
Works alongside Channel Affinity, Content Intelligence, The full Wave platform.
Why Wave is different
Built from day one to treat AI assistants as a channel.
The timing advantage is real. Incumbents are retrofitting AI awareness onto products designed before AI search existed. Wave treated AI-assistant referrals as a native inbound signal class from the start.
Most tools
Wave
AI-assistant traffic is bucketed as direct or none and lost.
Wave names ChatGPT, Perplexity, Claude, Gemini, and Copilot as first-class inbound channels.
AI awareness is bolted on after the fact, so the signal is approximate.
Wave classifies AI referrals natively and folds them into the same channel model as every other arrival.
AI traffic shows up only in an aggregate report, with no person attached.
Wave ties AI referrals to the contact, so you can act on how each buyer actually researches.
Your data trains a shared, cross-customer model.
Wave classifies and predicts per tenant. Your data never enters a shared model.
FAQ
Questions buyers ask about AI Referral Intelligence.
What is AI Referral Intelligence in Wave?
It is a Wave capability that detects and classifies website traffic arriving from AI assistants like ChatGPT, Perplexity, Claude, Gemini, and Copilot as a first-class inbound channel. Instead of losing that traffic in a direct or none bucket, Wave names the assistant and folds the signal into its channel prediction on every contact.
Which AI assistants does Wave track?
Wave classifies referrals from ChatGPT, Perplexity, Claude, Gemini, and Copilot as distinct channels in its canonical B2B channel taxonomy. The taxonomy is extensible per tenant, so you can add new AI assistants as they emerge without waiting on an engineering change.
How is this different from my analytics tool?
Most analytics stacks see no clean source on an AI-assistant click and file it as direct or none, so the channel disappears. Wave reads the inbound referrer server-side and classifies the specific assistant, then ties that signal to the contact rather than only to an aggregate report.
Do I need to change my tag manager or marketing automation platform?
No. Wave reads inbound signals on its own side through native connectors. There is no change required in your tag manager, your marketing automation platform, or your website code to start classifying AI-assistant referrals.
How does AI traffic affect my channel predictions?
AI-referred contacts flow into the same inbound channel prediction as UTM-based and referrer-based contacts. There is no separate pipeline and no separate silo, so an AI-referred buyer is scored alongside every other arrival in Channel Affinity's inbound axis.
Is the AI classification accurate?
Yes. Classification runs against a per-tenant taxonomy you control, which keeps it fast, accurate, and cost-efficient, and you decide how every channel is recognized. New assistants can be added as they emerge.
Does my data train a shared model?
No. Wave classifies and predicts one model per tenant. Your contact and engagement data never enters a shared or cross-customer model, and every prediction is stamped with the model version that produced it.
How do I see my AI referral mix?
Book a 20-minute walkthrough. We will configure a tenant against your stack and show the AI-assistant referral mix Wave would surface and how it folds into your channel predictions on real inbound signal.
See it on your data
See which AI assistants send you your best contacts.
Book a 20-minute walkthrough. We will run Wave against your stack and show the AI-assistant referral mix and channel predictions Wave would produce on your real inbound signal.
Request a demo Download the one-pager (PDF)