AI Visibility for Agencies

AI search attribution with no referrer / no click

Updated May 20268 min read
MentionRadar TeamWe run the MentionRadar query index across ChatGPT, Gemini & Grok.
The short answer

AI search frequently produces no click and no referrer: the user reads the assistant’s answer and never visits the site, and when they do, the visit often arrives with no identifying source. That breaks last-click attribution for AI entirely. The realistic model has two layers. First, measure visibility at the source — track which buyer queries the AIs cite the brand on, because being named in the answer is the event, whether or not a click follows. Second, corroborate downstream impact with the signals you can see: branded-search lift, direct/unattributed traffic trends, self-reported sourcing (“how did you hear about us?”), and assisted-conversion patterns. Stop trying to force AI into a click-based funnel; attribute it the way you attribute brand and PR.

Why AI search leaves no referrer

Three things conspire. First, AI answers are often zero-click — the assistant synthesizes the answer inline, so the user’s need is met without a visit. Second, when a user does follow a citation, the traffic can land without a recognizable referrer string, so it falls into “direct” in most analytics tools. Third, a lot of AI-influenced buying happens with a lag — the user sees a brand named in an answer today and searches for it directly next week. None of that fits the clean ranking-to-click-to-conversion chain SEO reporting relies on.

What this breaks (and what it doesn’t)

It breaks last-click attribution for AI search — you cannot point at a row in analytics and say “this conversion came from ChatGPT.” What it does not break is the underlying value: being the brand an assistant names on a high-intent question is influence on the buyer’s shortlist, the same way a glowing analyst mention or a top Reddit thread is. The error is treating an unmeasurable click as the only proof of value. Reframe the question from “did it get a click?” to “are we the brand the AI recommends?”

Layer 1: measure visibility at the source

The cleanest, most direct signal is upstream of any click: which queries do the AIs cite the brand on? That’s observable without touching the user’s analytics — you query the models and read the result. This is exactly what a reverse AI search index captures, and it’s why visibility, not clicks, is the primary metric for AI. The free Domain Check shows it for any domain across all three models.

Layer 2: corroborate downstream impact

You can’t prove the click, but you can build a credible body of corroborating evidence:

  • Branded-search lift. AI exposure often drives later branded searches. A rise in branded query volume that tracks visibility gains is suggestive (correlation, not proof — label it).
  • Direct / unattributed traffic trends. Watch the “direct” bucket; AI referrals often hide there. Annotate it against visibility milestones.
  • Self-reported sourcing. Add an “AI assistant / ChatGPT” option to “how did you hear about us?” on forms and in sales discovery. Imperfect, but the most direct downstream signal you’ll get.
  • Assisted conversions. Look for patterns where AI research plausibly preceded a known channel touch.

How to report it without over-claiming

Separate the two layers explicitly. Present visibility as measured fact (these are the queries you’re cited on, here’s the trend) and downstream impact as corroborating, partly modeled signal (here’s what we observe that’s consistent with that visibility). This honesty is what makes the report credible — and it’s the backbone of the CFO conversation in how to prove AI search ROI. Put the structure into your white-label report.

The mindset shift to sell clients

Help clients internalize that AI search is a brand and shortlist channel, not a direct-response channel. They didn’t demand a last-click number for a strong analyst report or a #1 spot in a respected “best of” roundup, and AI citations are the same kind of asset. When the “but is it real revenue?” objection comes, you’re ready — see the vanity-metric response. Start by measuring the source: run the free Domain Check to establish the visibility baseline every other signal hangs off.

Frequently asked questions

Why does AI search send no referrer?
Most AI answers are generated rather than linked. The user reads the response inside ChatGPT, Gemini, or Grok without clicking through, so there is no outbound visit for analytics to record as a referral.
What attribution model can I actually defend?
Switch from counting clicks to measuring presence: how often the brand is cited for the queries that matter, tracked over time, alongside corroborating demand signals like branded search, self-reported sourcing, and assisted conversions.
Does any AI traffic show up in analytics at all?
Some does — when a model links a source and the user clicks, you may see referrals from the assistant's domain. But it is partial and inconsistent, so it cannot be your primary proof.
How do I explain this to a client used to last-click?
Be candid that AI answers are largely zero-click, then reframe success as share of answer for buying queries plus branded-search lift. An honest presence metric earns more trust than a forced click number.