# How to Attribute Conversions to AI-Referred Traffic

- Canonical: https://www.geoly.ai/blog/ai-traffic-attribution
- Published: 2026-03-07
- Updated: 2026-07-07
- Author: [GEOly AI](https://www.geoly.ai/blog/author/geoly-ai)
- Tags: [How To](https://www.geoly.ai/blog/tag/how-to), [GA4](https://www.geoly.ai/blog/tag/ga4), [GEO](https://www.geoly.ai/blog/tag/geo), [AI Search](https://www.geoly.ai/blog/tag/ai-search), [AI Visibility](https://www.geoly.ai/blog/tag/ai-visibility)

> **Summary:** Turn dark AI traffic into provable ROI: catch AI referrers in GA4, UTM-tag the links you control, model zero-click influence, and attribute conversions and revenue to GEO.

The hardest part of proving GEO works is attribution. A click from Google search is neatly labeled in your analytics; a visit driven by ChatGPT often arrives as "Direct" or a generic "Referral," so the revenue it produces gets credited to the wrong channel — or to nothing at all. Marketers call this "dark AI traffic," and it's why GEO ROI looks invisible even when AI answers are quietly sending you buyers.

The good news: as AI search matures, it leaves more distinct footprints. This guide shows how to surface that data and attribute conversions to your GEO work — configuring GA4 to catch AI referrers, tagging the links you control, modeling the zero-click influence you can't tag, and closing the loop with off-site visibility so you can report real ROI instead of a shrug.

## Key takeaways

- AI touches your funnel two ways: referral clicks (a citation link in Perplexity or Gemini) and zero-click influence (a shopper reads a good ChatGPT answer, then types your domain directly, landing in GA4 as "Direct").
- Referral clicks are recoverable with a GA4 AI channel; zero-click influence is inferred, not measured — model it, don't fake a clean number.
- Add UTM parameters to every link you control, including entries in `llms.txt`, because some AI clients strip referrer data and dump the visit into Direct.
- Attribute on behavior and conversions, not just sessions. AI-referred visitors are pre-qualified, so compare their conversion rate and revenue against organic search, not their raw volume.
- GEOly connects the off-site half — where you appear and get cited in AI answers — to on-site sessions via its GA4 integration, so citations and conversions sit in one view.

## Step 1: Configure GA4 to catch AI referrers

GA4 captures referral data, but AI sources are scattered and easy to miss, so build a dedicated segment or channel. The domains that actually send clicks in 2026 include `chatgpt.com` and `openai.com`, `perplexity.ai`, `gemini.google.com` (and legacy `bard.google.com`), `claude.ai` and `anthropic.com`, and `copilot.microsoft.com`. Microsoft is the messy one: Copilot and Bing Chat traffic often shows as `bing.com` mixed with regular search, or falls into Direct.

The fast route is a custom channel group. In GA4 Admin, create one, add an "AI Search" channel above Referral and Organic (order matters — GA4 assigns each session to the first matching channel), and match session source with a regex covering those domains. For the full channel-group build, see [how to track AI traffic in GA4](/blog/track-ai-traffic-ga4).

Gotcha: this captures referral clicks only. It misses the zero-click influence entirely, so treat the number as a floor, not the whole picture.

## Step 2: Tag the links you control with UTMs

You can't tag the links ChatGPT renders, but you can tag every link you place in AI-friendly locations — and that's your defense against stripped referrers. Anywhere you can append parameters, add a consistent `utm_source` and `utm_medium` (for example `utm_source=chatgpt&utm_medium=ai`) so those sessions land deterministically in your AI reporting instead of relying on referrer detection.

This matters most for `llms.txt`. Since that file is a set of links you author for models, tag the destination URLs so any click that follows an `llms.txt` entry is attributable. Do the same for links in docs, community answers, and anywhere you seed content that AI may cite.

Gotcha: keep UTM values lowercase and consistent. GA4 treats `ChatGPT` and `chatgpt` as two different sources and will split your data across both.

## Step 3: Model the zero-click influence you can't tag

The largest share of AI's impact is invisible to referrer tracking. A shopper asks ChatGPT about your category, gets a favorable answer that mentions you, closes the tab, and later types your domain directly. GA4 records "Direct," but AI visibility drove that visit. You can't measure it cleanly, so model it — and be honest that it's an estimate.

Two signals help. First, watch Direct traffic for correlation: when your AI visibility rises (more mentions, more citations), does branded Direct and branded organic search lift in the same window? A consistent co-movement is evidence of zero-click influence. Second, run periodic surveys or "how did you hear about us?" fields at checkout, and count the AI mentions. Neither is perfect; together they bound the effect.

Gotcha: don't quietly fold estimated zero-click into your "AI referral" number and present one clean figure. Report measured referral and modeled influence separately, or the moment someone audits it, the whole attribution loses credibility.

## Step 4: Attribute on conversions and revenue, not sessions

Volume is half the story and the less useful half. AI-referred visitors are pre-qualified — they read a summarized answer before clicking, arrive with intent, and often head straight to a pricing or product page rather than top-of-funnel content. Attribution that stops at session counts misses this.

In GA4, filter to your AI Search channel and build an exploration comparing conversion rate, average engagement time, and revenue per session against organic search. Break it down by landing page to see which AI answers send buyers versus browsers. This is what turns "we get some AI traffic" into "AI answers about our category convert at X and drive Y in revenue" — the number that justifies GEO budget. For a full ROI framework, see [how to measure GEO ROI](/blog/how-to-measure-geo-roi-guide-2026).

Gotcha: with lower volume, small samples swing conversion rates. Give the channel enough time and sessions before drawing conclusions, and watch trend, not a single week.

## Step 5: Close the loop with off-site visibility

GA4 tells you what happened after the click; it can't tell you how often you appeared in AI answers, whether you were cited, or how your Share of Model compares to competitors. That off-site half is where the attribution story completes. Without it, you're measuring the visits AI sent and ignoring the answers where you were mentioned but didn't earn the click — or weren't mentioned at all.

GEOly measures that off-site side — your visibility, mentions, and citations across ChatGPT, Gemini, Perplexity, Copilot, and Grok — and connects it back to on-site sessions through its GA4 integration. Read together, the two halves answer the real question: not just "how much AI traffic did I get," but "which AI answers drove it, and where am I losing answers I should win." For the deep analytics build, including a Looker Studio dashboard on top of this channel, see [how to track GEO and analyze AI traffic in GA4 and Looker Studio](/blog/how-to-track-geo-analyze-ai-traffic-ga4-looker-studio). Start a free three-day trial at `app.geoly.ai`.

## Common mistakes

- Reporting all AI impact as one clean number by merging measured referral with estimated zero-click. Keep them separate and labeled.
- Blocking AI crawlers while trying to attribute AI traffic — no crawl, no citations, no attributable clicks.
- Leaving `llms.txt` and seeded links untagged, so their clicks vanish into Direct.
- Judging AI traffic by volume and dismissing it as small, while ignoring that its conversion rate may be your highest.

## How to verify

- Send yourself a test: follow a real link to your Perplexity or ChatGPT answer, complete a session, and confirm the next day it lands in your AI Search channel, not Referral.
- Check your source/medium report for the AI domains in your regex and confirm they're matching.
- Confirm UTM-tagged `llms.txt` links show up under your chosen `utm_source` in GA4.
- In GEOly, cross-check that rising citations line up with lifts in AI-referred and branded Direct sessions.

## FAQ

Why does AI traffic show up as Direct? Some AI clients strip the referrer, and in-app browser flows often carry no source, so GA4 defaults to Direct. UTM-tagging the links you control is the fix for the portion you can influence.

Can I ever fully attribute zero-click influence? No. It's inferred from correlation and self-reported surveys, not measured. Model it, bound it, and report it as an estimate alongside your hard referral numbers.

How is this different from just setting up a GA4 channel? The channel captures referral clicks. Attribution goes further — tagging your own links, modeling zero-click, and tying conversions and revenue (plus off-site citations) to GEO so you can prove ROI.

Which conversions should I attribute to AI? Start with the sessions in your AI Search channel and their conversions, then layer modeled zero-click influence separately. Judge on conversion rate and revenue per session, not raw traffic.

How does GEOly help prove GEO ROI? It measures where you appear and get cited in AI answers and links that to GA4 sessions, closing the loop between visibility and revenue. See [what GEOly is](/blog/what-is-geoly-ai) and more tutorials on the [GEOly AI author page](/blog/author/geoly-ai).
