AI attribution: crediting what an AI answer actually influenced
Counting AI referral sessions is the easy half. Attribution asks the harder question: did an AI mention or citation actually influence a conversion that happened somewhere else, later, on a different device?
In short
AI attribution is the practice of crediting conversions influenced by an AI system's mention, citation, or referral of your brand - including cases where the AI interaction and the eventual conversion happen in different sessions, with no referrer header connecting them. It borrows from multi-touch attribution but adds a harder problem: most AI influence never leaves a trackable trail at all.
Why this is harder than normal attribution
The dark-traffic problem: someone reads an AI answer, closes the tab, and later searches your brand name directly or types the URL from memory. No referrer header survives that path, and no analytics platform sees the AI interaction that started it.
What can actually be measured
The trackable slice: direct AI-referral sessions with an intact referrer or UTM that convert on their own. Beyond that, branded-search lift correlated with visibility changes over time - useful, but correlation, not causation - and self-reported "how did you hear about us" data, which is honest but incomplete.
What has to stay labeled as inference
Any claim linking a rise in branded search to a specific AI mention is a hypothesis unless tested with a controlled before/after comparison - never present it as a measured fact just because the timing lines up. This is the same discipline that applies to every unverified claim in this space: label it, and say what would confirm or refute it.
Where this fits in a measurement stack
Attribution sits on top of AI traffic measurement (referral segmentation) and the AI visibility audit (mention tracking) - it's the layer that tries to connect the two into one number, with the confidence level stated honestly at every step rather than collapsed into a single misleadingly precise figure.
Common questions
Can I fully attribute revenue to an AI mention?
Rarely completely. The trackable slice is direct AI-referral sessions that convert; the rest - someone reads an AI answer, then searches your brand name later - has to be modeled as correlated, labeled as inference, and never presented as a measured fact.
Is this the same as multi-touch attribution?
Related but harder. Multi-touch attribution assumes a trackable touch at each step. AI-influenced conversions often have an untrackable step - the AI interaction itself - that conventional attribution models weren't built to handle.
What's the minimum viable version of this?
Segment true AI-referral traffic first, track its conversion rate on its own, and treat any broader brand-lift claim as a hypothesis to test, not a number to report as fact.
Do you build AI attribution models?
We build the instrumentation - referral segmentation, outcome joins - that makes attribution possible, and we label what is measured versus inferred at every step.
Put a real number on your AI traffic - labeled honestly.
Referral segmentation, crawler analytics, and outcome joins, with measured and inferred kept clearly separate.