GA4 referral reports can make AI-assisted discovery look negligible. New analyses summarized by Marie Haynes suggest that conclusion is incomplete: AI assistants currently send far fewer visits than Google Organic, but the visitors who do arrive can convert at a higher rate. More importantly, an AI answer, citation, or recommendation may influence a buyer before any measurable referral click occurs.
That creates a measurement problem for Generative Engine Optimization (GEO). Referral traffic remains useful evidence of direct visits from AI platforms, but it is not a complete measure of whether a brand was surfaced, trusted, compared, or selected during an AI-assisted research journey.
In Marie Haynes' analysis of GEO and referral data, 12 months of GA4 data across about 100 sites, representing roughly 145 million sessions, showed just how small direct AI referral traffic remains relative to Google Organic.
ChatGPT traffic amounted to 0.62% of Google Organic traffic in the dataset. Counting all AI assistants increased that figure to 0.73% of Google Organic traffic. Google Organic generated 964 conversions, compared with 36 from AI and LLM referrals combined. In absolute terms, Google Organic therefore drove about 27 times as many conversions as all AI assistants in the sample.
But the conversion-rate comparison points in a different direction. ChatGPT referrals converted at 13.8%, versus 3.02% for Google Organic. This does not mean AI is replacing search traffic, nor does it prove that an AI referral itself caused the conversion. It does show that the relatively small group arriving through ChatGPT was more likely to complete the measured conversion in this dataset.
| Measure in the Haynes dataset | Finding | What it indicates |
|---|---|---|
| ChatGPT traffic relative to Google Organic | 0.62% | Direct AI referrals were very small in volume. |
| All AI-assistant traffic relative to Google Organic | 0.73% | Adding assistants did not materially close the volume gap. |
| Conversion rate | ChatGPT 13.8%; Google Organic 3.02% | The observed AI-referral audience converted at a higher rate. |
| Recorded conversions | Google Organic 964; all AI/LLM referrals 36 | Google Organic still produced far more measured conversions. |
A parallel analysis cited in Search Engine Journal's reporting reached a similar broad conclusion. Across 108 million sessions from 57 sites, collected from January 1 to November 2, 2026, AI traffic represented around 0.66% of total sessions. Its AI visitors converted at roughly five times the rate of typical search users.
The two analyses are not market-wide estimates. They cover selected sites, use GA4 attribution, and involve different samples. They nevertheless challenge a simple interpretation of low referral volume: low clicks do not necessarily equal low commercial influence.
Traditional web analytics is designed around an identifiable visit. A user clicks a result, lands on a site, and the referrer or channel helps explain the session. AI-assisted journeys can be less visible.
A prospective customer may ask an assistant to explain a category, compare options, shortlist providers, or assess a product's suitability. The assistant can mention brands or cite sources during that process. The user may then search for the brand later, return directly, click a paid ad, or convert through another channel. In that sequence, the earlier AI interaction may have helped form intent, while the final session receives the credit.
This is the "agentic" influence Haynes highlights. It is not a claim that every AI mention produces a sale. It is a reminder that last-click or referral-based reporting cannot record every touchpoint that shapes a decision.
Attribution is also technically inconsistent. Depending on the assistant, browser, app, and integration, a visit may appear in GA4 as a referral, an AI-assistant source, another channel, or not in a way that cleanly identifies the AI origin. That makes platform-to-platform comparisons especially difficult.
Direct AI referrals should still be monitored. They can reveal which assistant sources send visitors and how those visitors behave. The mistake is treating that channel alone as a complete GEO scorecard.
A practical measurement framework can combine several signals:
This approach also changes optimization priorities. Pages that clearly answer user questions, expose important facts in accessible formats, and use accurate structured data can be easier for systems to interpret. Strong entity signals and consistent brand information can likewise support trustworthy representation across search and AI-generated answers. These are readiness measures, not guarantees of citations or rankings.
The findings offer a useful corrective to two extremes. AI referrals are not yet comparable with Google Organic in the cited data, so businesses should not disregard established organic search performance. At the same time, dismissing GEO because AI referral percentages are small risks overlooking where buyers first encounter and evaluate brands.
If your reporting treats AI traffic as a side channel, create a GEO Search Leads account to assess your website's GEO readiness and identify optimization opportunities around AI-friendly content, entities, and structured data. That gives your team a practical starting point for connecting AI visibility work with the topics and pages that matter most to your acquisition strategy, without assuming referral traffic tells the whole story. Does AI referral traffic currently rival Google Organic traffic?
No. In Marie Haynes' 12-month GA4 dataset, ChatGPT traffic equaled 0.62% of Google Organic traffic, and all AI assistants combined equaled 0.73%. Google Organic also generated far more recorded conversions.
Why can AI referrals have a higher conversion rate but fewer conversions?
Conversion rate measures the share of visitors who convert, while total conversions also depend on visitor volume. The cited dataset recorded a higher ChatGPT conversion rate, but Google Organic sent much more traffic and therefore delivered 964 conversions versus 36 from all AI and LLM referrals.
Can GA4 measure the full impact of AI-assisted discovery?
No. GA4 can record some direct referrals, but AI conversations, citations, and recommendations may influence a user before they visit through another channel or convert later. Source attribution can also vary by assistant and integration.
What should businesses measure for GEO in addition to clicks?
Track identifiable AI referrals alongside brand mentions and citations in relevant AI answers, coverage of important topics, conversion quality, and broader assisted paths such as branded search or direct visits. These signals should be interpreted together rather than as proof that AI caused a conversion.
The available GA4 analyses show that direct AI referrals remain small beside Google Organic, even when they convert efficiently. The more consequential lesson for GEO is measurement: referral reports capture direct visits, not every AI interaction that may shape buyer intent. Businesses should retain rigorous traffic and conversion reporting while adding visibility, citation, and content-readiness signals to understand AI-assisted discovery more fully.