AI referral traffic vs organic search: What the data shows

What does the evidence actually show?

In her October 2026 article, “GEO is a distraction”, Marie Haynes describes an analysis of 12 months of GA4 data across 100 websites, covering more than 145 million sessions.

In that sample:

• ChatGPT sessions amounted to 0.62% of the volume delivered by Google organic search.
• Including the other LLM sources increased that figure to 0.73%.
• Across the 55 sites tracking conversions, Google organic delivered 27 times more conversions than all LLMs combined.

The SEO Works found a similar imbalance in its AI Referral Traffic Benchmark, although its figures and methodology differ.

Its analysis covered 95 websites with 12 complete months of data, from September 2025 to August 2026. Those websites received:

• 2,825,594 organic search sessions.
• 72,095 AI referral sessions.
• Approximately 39 times more sessions from organic search than from AI assistants.

AI referral sessions therefore equalled approximately 2.6% of organic search sessions in that dataset.

These are separate samples, not a universal benchmark for every industry. Haynes compares LLM referrals with Google organic specifically; The SEO Works compares them with organic search. Neither study establishes that every website receives the same proportion of traffic from AI.

Nevertheless, all three point towards the same practical conclusion: measurable assistant referrals remain much smaller than organic search in the websites studied.

A small fraction of organic traffic is not the same as a share of total traffic

This distinction matters when headlines start quoting percentages.

The SEO Works' 2.6% figure does not mean AI generated 2.6% of all website traffic. Its denominator is organic search sessions, not total sessions. Haynes' 0.73% figure is likewise a comparison with Google organic volume, not all traffic.

Total website traffic can include paid campaigns, email, social media, direct visits and other referrals. Adding those channels changes the denominator.

For illustration, imagine a website receives 10,000 organic sessions and 260 AI referral sessions, alongside 9,740 sessions from other channels. AI referrals equal 2.6% of organic volume, but only 1.3% of its 20,000 total sessions. These are hypothetical figures, not results from any of the three datasets.

When reviewing a claim about AI traffic, always ask: a percentage of what?

AI referral traffic does not capture every AI-influenced visit

A visitor clicking a link in ChatGPT or Perplexity is different from someone clicking a website link within a Google AI Overview.

The SEO Works identifies assistant referrals using referring sources such as ChatGPT, Claude, Gemini and Perplexity. Its benchmark explicitly excludes Google AI Overviews, because those visits arrive as ordinary organic sessions in GA4.

Google's own guidance also explains that traffic from AI Overviews and AI Mode is included within overall search traffic in Search Console's Web performance reporting.

There are further measurement gaps. A person might encounter a brand in an AI answer, then search for it later. An app or browser might strip referral information. Neither journey necessarily appears as an identifiable AI referral.

That means a small referral total is not proof that AI has no influence. Equally, a brand mention or citation is not proof that it generated a customer. Visibility, visits and commercial outcomes need separate measures.

What about claims that AI traffic converts better?

Higher conversion rates can make a small channel worth paying attention to. They do not automatically make it the most valuable channel.

Among the 55 websites tracking conversions in Haynes' sample, ChatGPT visitors converted at 13.8%, compared with 3.02% for Google organic visitors. Yet Google organic still delivered 27 times more conversions than all LLM sources combined.

The SEO Works found a more modest advantage. In its eligible 63-site conversion sample, AI referrals converted at 7.18%, compared with 6.08% for organic search: approximately 1.18 times the organic rate.

That advantage was not consistent:

• AI referrals converted worse than organic search on 23 of the 63 websites.
• Five websites recorded no AI conversions despite receiving AI referral sessions.
• Across the conversion sample, AI referrals delivered 3.2 conversions for every 100 delivered by organic search.

These conversion figures represent GA4 key events, not necessarily unique customers or completed sales. Businesses can also define key events differently, which limits comparisons between studies.

The sensible question is not simply “Does AI traffic convert better?” It is “How much qualified business does it generate, and what does acquiring that business cost?”

Growth headlines need context, too

Fast percentage growth can look impressive when the starting point is small.

The SEO Works reports that AI referrals increased relative to organic search in its sample from May 2026. By August, they equalled 3.86% of organic sessions, up from 2.25% in September 2025.

Haynes describes a different pattern in her sample: ChatGPT referrals stopped growing exponentially around August 2025 and subsequently declined.

Those findings should not be combined into a single industry-wide trend. They cover different websites, source groupings and potentially different reporting windows. Haynes' article does not specify the exact start and end dates of its 12-month sample.

The useful takeaway is to monitor your own clients rather than assume that a headline trajectory applies to them.

What our own client analytics show

We deliver SEO behind the scenes for agencies, so we also looked at our own client analytics. Seven client GA4 properties gave us a complete year of data, covering 1 September 2025 to 31 August 2026.

Across those seven websites:

• Organic search delivered 712,958 sessions; identifiable AI referrals delivered 11,086.
• AI referral sessions equalled 1.6% of organic search volume, roughly one AI session for every 64 organic sessions.
• Site by site, AI's share of organic volume ranged from 0.6% to 4.5%, with a median of 2.0%.
• ChatGPT accounted for about 83% of the AI referral sessions.
• Those AI sessions produced about 9 key events per 100 AI sessions in GA4.

Two qualifications matter. First, like the published studies, these figures compare AI referrals with organic search only. The share of all sessions, once paid, email, direct and other channels are included, is lower still. Second, GA4 key events are not necessarily completed sales or unique customers, and each site defines them differently. Treat that conversion figure as an engagement indicator, not a revenue claim.

Our seven sites sit alongside the two published studies rather than replacing them. Different samples, same direction: measurable AI referrals remain a small fraction of the traffic organic search delivers.

Where should agencies put their effort?

Start with the opportunities already supported by customer demand and measurable outcomes.

1. Keep the SEO foundations strong

Google says its established SEO best practices remain relevant to AI Overviews and AI Mode. There are no additional technical requirements for appearing as a supporting link beyond the existing eligibility requirements.

That makes crawlable pages, clear business information, useful content and a dependable website practical priorities. An SEO audit can help identify obstacles before money is diverted into a new visibility initiative.

2. Improve content that helps people make decisions

Answer real customer questions. Explain your services clearly. Support claims with evidence and make the next step easy to find.

This is a stronger foundation than producing generic pages solely to chase AI citations. Relevant blog content should earn its place by helping the audience, not by filling a publishing quota.

3. Report traffic and outcomes separately

Track identifiable AI referral sources alongside organic search. Use a consistent period and record:

• Sessions and their share of total website traffic.
• AI referral volume relative to organic search.
• Meaningful key events, qualified enquiries and completed sales where available.
• Revenue, lead quality and acquisition cost where your reporting supports them.

The client figures in this article were produced this way. Check that key events are configured correctly before drawing conclusions. A newsletter signup and a signed client contract should not be treated as interchangeable wins.

4. Make AI visibility an accountable experiment

There can still be a case for AI SEO activity, particularly where research shows that prospects use AI assistants to compare suppliers.

Define the audience, target questions, deliverables, budget and review point. Assess citations and mentions separately from referral traffic and sales. Do not promise that a placement guarantees inclusion in an AI answer.

Where organic search already delivers substantial business, AI work should complement an ongoing SEO strategy, not replace it on the strength of industry excitement alone.

The bottom line

AI referral traffic is real. It can be valuable, and its importance will vary by website, audience and sector.

But the evidence across these three samples does not support treating AI assistants as a replacement for organic search. Measurable referrals were a small fraction of organic traffic, and higher conversion rates did not overcome the gap in volume.

For most businesses, the prudent starting point is to measure their own traffic mix, protect what already works and test AI opportunities proportionately.

Invest in customer outcomes, not just the newest visibility metric.

Sources

• Marie Haynes: GEO is a distraction, published 21 October 2026.
• The SEO Works: AI Referral Traffic Benchmark, covering September 2025 to August 2026.
• Google Search Central: AI features and your website.

Sources checked on 24 October 2026. Third-party findings come from the published research listed above; the seven-client figures come from our own client analytics, seven GA4 properties covering September 2025 to August 2026.