AI referral traffic vs Google organic: what three October 2026 studies show, and what they cannot measure

Is AI search worth investing in? Why the question blew up this week

In the space of four days, three independent analyses of real analytics data landed on the same uncomfortable conclusion for anyone selling or buying "generative engine optimisation" (GEO): AI assistants still send very little traffic compared with Google. On 22 October 2026, Barry Schwartz rounded them up on Search Engine Roundtable under the headline 3 New Studies Suggest AI Visibility & Search Not Worth Investing In, and Rand Fishkin’s whiteboard video on the same studies drew more than 300 comments on LinkedIn.

• Marie Haynes, [GEO is a distraction](https://www.mariehaynes.com/geo-is-a-distraction/) (21 October 2026). Twelve months of GA4 data from 100 randomly sampled websites, more than 145 million sessions.
• SEO Works, [AI referral traffic benchmark](https://www.seoworks.co.uk/downloads/ai-referral-traffic-benchmark/). GA4 data from 95 websites, September 2025 to August 2026, with conversions compared on 63 of them.
• Kevin Indig, [Does AI drive more direct traffic?](https://www.growth-memo.com/p/does-ai-drive-more-direct-traffic) (19 October 2026). A test of whether AI exposure shows up later as Direct visits or branded searches, across 12 sites and four brands.

The timing matters. The same month, ChatGPT’s referral data briefly broke: links lost their tracking tag for several days after the GPT-6 rollout, and many site owners saw their "ChatGPT traffic" halve overnight. So the debate is really two questions at once: how much does AI search send, and can we even measure it?

How much traffic does ChatGPT send compared with Google?

Haynes found that all AI assistants combined amounted to 0.73% of Google organic volume, and that ChatGPT referral growth stalled around the GPT-5 release in August 2025. SEO Works, using a wider definition of AI referrers (ChatGPT, OpenAI, Claude, Copilot, Gemini, Perplexity and Grok), found AI sent 2.6% of organic’s traffic over the year. Neither study counts Google’s own AI Overviews or AI Mode as "AI traffic": those visits arrive as Google organic.

The trend line is the part most coverage skipped. AI’s share is not flat: SEO Works reports organic’s lead narrowing from 44x to 33x after May 2026, with August 2026 at 26x. It also found AI referral growth tracked organic growth (a correlation of +0.557), which suggests the sites winning in AI assistants are largely the ones already winning in search.

Does AI traffic convert better? The studies disagree

The most quoted claim for AI search is that its visitors convert far better. The two studies with conversion data point in the same direction but by very different amounts.

SEO Works also broke conversions down by business type. AI traffic converted 1.51x better than organic for B2C sites, 1.46x for mixed sites and 1.21x for B2B services, but only 0.62x for B2B product sites, where six of ten sites saw AI convert worse. Against other channels, AI converted at 0.67x paid search and 2.48x Direct. Its conclusion is the most useful sentence in the debate: AI traffic "converts better than organic, but by far less than existing figures in circulation suggest."

What SEOs are saying about the GEO debate

Rand Fishkin, co-founder and CEO of SparkToro, said in his LinkedIn video post that he came away "less confident than ever" that more visibility in ChatGPT, Claude, Perplexity or Copilot would have much business impact for most companies, while noting Google’s AI Overviews may matter more. Not everyone in the thread agreed. Lily Ray warned that many GEO tactics are exploits the AI platforms will close, and that some risk demotions in Google. The strongest objection came from the local search side:

Shaw added that ChatGPT is still the top self-reported source in Whitespark’s software signup survey. That is the crux: referral traffic measures clicks, while much of AI’s influence happens without a click, or with a click that loses its label on the way.

Why does ChatGPT traffic show as Direct in GA4?

Every study above relies on GA4 recognising where a visit came from. That depends on the AI assistant passing a referrer or a tag such as utm_source=chatgpt.com, and October showed how fragile that is. Search Engine Roundtable reported on 14 October that ChatGPT links had started opening without a UTM tag or referrer around the GPT-6 release, so GA4 filed those visits as Direct. Cedric Huesler of Adobe found, across about 500 pages, ChatGPT visits falling by about 31,000 a day while Direct rose by about 26,000 a day on the same pages. On 15 October he posted that it was a bug and that tagging had returned.

LLM Pulse measured a drop of about 46% in daily ChatGPT sessions over 9–11 October across the sites it monitors, with total traffic down only about 2%, and Perplexity and Claude referrals unaffected. It also saw "gemini"-tagged visits almost disappear between 28 September and 2 October while visits from gemini.google.com held steady, which points to a tagging change rather than lost visitors. The lesson for anyone reading AI traffic charts this autumn: a sudden drop in an AI channel is more often a labelling change than a change in demand.

Kevin Indig’s study tested the opposite worry, that AI is quietly sending lots of untagged visitors. In his sample, most of the recent growth in Direct was low-quality: total Direct rose 140.9% between January and September 2026 while engaged Direct sessions barely moved, and one site logged 409,000 Direct sessions with only 198 engaged. Before you credit AI with a Direct spike, filter for engaged sessions.

How to measure AI visibility when clicks are not the whole story

No single tool shows AI visibility end to end. These are the sources available today and what each one can and cannot tell you.

Bing’s new Topics Portfolio, launched in beta on 13 October 2026, is the most practical addition. It groups the grounding queries where your site is cited into topics, shows each topic’s approximate size and how many sites compete in it, and lists "Opportunity Queries" where other sites are cited and yours is not. Microsoft is clear that the metrics are approximate "directional signals for planning", and small sites may see few topics or none.

What should businesses do about AI search now?

The studies do not say AI search is irrelevant. They say that, measured by clicks, it is small next to Google, growing slowly, and very uneven by sector. That supports a clear order of priorities.

1. Keep Google first. Google organic, which now includes AI Overviews and AI Mode, still drives the overwhelming majority of conversions in every dataset published this month. Strong SEO and technical SEO remain the foundation of visibility in AI answers too, since AI referral growth tracked organic growth in the SEO Works data.
2. Set up conversion tracking properly. 45 of Haynes’s 100 sites tracked no key events at all. You cannot judge any channel, AI or otherwise, without them.
3. Build an AI assistants channel group in GA4. Match referrers such as chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com and claude.ai, and keep utm_source=chatgpt.com visits in the same group. Annotate 9–15 October 2026 as a tracking gap for ChatGPT.
4. Check engaged Direct, not total Direct. If your top AI landing pages show a Direct rise, compare engaged sessions before attributing it to AI.
5. Open Search Console’s generative AI report and Bing’s AI Performance report monthly. Impressions and citations are the closest thing to a ranking report for AI answers. Use Bing’s Opportunity Queries to find pages worth improving.
6. Add one question to every lead form: "How did you hear about us?" with AI assistants as an option. This is the only way most businesses will see AI influence that never produces a tagged click.
7. Avoid shortcut tactics. Spend on clear, specific, well-sourced pages that serve people and machines equally, not on tricks aimed at one assistant. Our answer engine optimisation work starts from the same rule.
8. Weigh by sector. If you sell B2B products, the SEO Works data suggests AI visitors may convert worse than organic; if you are B2C or a service business, the upside is larger. Let your own quarter of data decide.

How to report AI search to a board or client

• Show AI as a share of organic, not in isolation. A 50% month-on-month rise in ChatGPT sessions can still be a fraction of one percent of search traffic.
• Pair every traffic figure with a visibility figure. Search Console AI impressions and Bing citations for the same period.
• Report conversions per channel with session counts beside them, so a high rate on a tiny base is not mistaken for a trend.
• Flag known tracking breaks such as 9–15 October 2026 for ChatGPT and late September 2026 for Gemini tags.

If you want this built for you, our reporting service puts organic, AI assistant and Search Console AI data in one view. For the Google side of the same story, read our analysis of the Google AI Overviews links update.