Quick Answer

Not as a traffic channel, at least not yet. New studies published this week found that AI assistants send a very small share of the visits Google does, and Google still drives almost all conversions. That does not mean AI search is useless. It means you should keep Google first, treat AI visibility as a brand channel, and measure both before moving budget.

What changed

Three pieces of research landed within days of each other, and they point the same way. On October 23, 2026, Marie Haynes published an analysis on Search Engine Journal covering 100 websites over 12 months and more than 145 million sessions. On October 22, Rand Fishkin of SparkToro pulled several studies together in a post asking whether AI visibility is a marketing boondoggle. The same day, SISTRIX published a study on which page features are linked to being cited by AI tools.

The headline number from Haynes is that ChatGPT sessions were 0.62% of the volume Google organic search delivered, and all large language models together reached 0.73%. Her own conclusion was blunt: the growth many people expected did not arrive, and ChatGPT referral growth plateaued in August 2025.

What the studies found

Study What it measured Finding What it means for you
Marie Haynes (SEJ, Oct 23) GA4 data, 100 sites, 12 months All LLMs sent 0.73% of Google organic volume. ChatGPT converted at 13.8% vs 3.02% for Google, yet Google drove 96.5% of conversions (964 vs 36) AI visitors are high intent but too few to move your numbers yet
SEOWorks UK (via SparkToro, Oct 22) 63 sites, Sep 2025 to Aug 2026 Organic search sent about 39 times more traffic than AI assistants. AI traffic converted worse than organic on 23 of 63 sites Do not assume AI visitors always convert better
Kevin Indig (via SparkToro, Oct 22) 12 sites, indirect effects on branded search and direct visits No reliable link found between AI visibility and those indirect gains The hidden benefit is unproven, so measure it before you pay for it
SISTRIX (Oct 22) 96,418 URLs, 17,380 domains, DACH region, Q2 2026 Pages with a table of contents (1.7), Article schema (1.6), five or more H2s (1.6), FAQPage schema (1.5) and a header-row table (1.5) were cited more. llms.txt showed no effect Clear page structure is the safe, cheap part of AI optimization

Sources: Search Engine Journal, Marie Haynes (Oct 23, 2026); SparkToro, Rand Fishkin (Oct 22, 2026); SISTRIX (Oct 22, 2026). SISTRIX results show correlation, not proven cause, and cover only the DACH region.

Why this matters

If you have been told you must buy an AI visibility package right now, these numbers are worth showing your provider. The data does not say AI search is dead. Fishkin himself believes AI likely influences some buyers in some sectors, and he suggests treating it like a brand channel, closer to a billboard than a traffic source. His warning is against restructuring budgets around AI visibility as if it were a traffic channel.

There is also a measurement problem. Haynes found that 45 of the 100 sites tracked no key events in GA4 at all. If you cannot see conversions, you cannot tell whether any channel is paying off. Read these studies as a reason to measure better, not as a reason to stop. Our guide on SEO vs AEO vs GEO explains where each fits.

What to do about it

  1. Keep Google organic as your first priority. It delivered 96.5% of conversions in the Haynes data. Protect that before you fund anything new.
  2. Set up conversion tracking. Define key events such as form fills, calls and bookings in GA4, then segment referrals from ChatGPT and other assistants to see what they really send you.
  3. Make your pages easy to read and cite. A clear structure, FAQ sections, tables and a visible update date help readers and are linked to more citations in the SISTRIX study. Skip llms.txt as a priority.
  4. Cap AI-specific spend until you have your own data. Test with a small budget for one quarter and compare against your baseline. Our SEO cost guide shows what a balanced budget looks like.
  5. Get a second opinion on your plan. Our SEO, AEO and GEO services start with measurement, so you can see what each channel is worth before you commit.

The bottom line

The honest read is mixed: AI search is real but small, Google still pays the bills, and good page structure helps both. Build on what works, measure what is new, and see how the pieces connect at GCL Marketing, or browse more articles by topic.

Frequently Asked Questions

Is AI search traffic worth chasing?

Not as a traffic channel, based on these studies. In Marie Haynes’s data from 100 sites, all large language models together sent 0.73% of the volume Google organic did. Treat AI visibility as a brand channel, keep Google as the priority, and measure both so your own numbers decide.

Why do ChatGPT visitors convert better but matter less?

In the Haynes data, ChatGPT visitors converted at 13.8% against 3.02% for Google organic. But there were so few of them that Google still drove 96.5% of conversions, about 27 times more than all large language models combined. High intent from a tiny volume is still a tiny result.

Should I stop doing GEO?

No, but keep it in proportion. The SISTRIX study of 96,418 URLs found that pages with clear structure, such as a table of contents, FAQ schema and tables, were cited more often. Those same habits help readers and Google too, so they are cheap to keep. Just do not rebuild your budget around them.

Does llms.txt help AI tools cite my site?

Not according to SISTRIX. In its study, cited URLs on domains with an llms.txt file made up 11.7% of the total, against 14.0% across all domains, so there was no measurable benefit. The study shows correlation only and covers the DACH region, so treat it as a signal, not a final verdict.

How do I know what AI sends my business?

Set up conversion tracking first. Haynes found that 45 of the 100 sites tracked no key events in GA4, so they could not see value either way. Define your key events, such as form fills and calls, then segment referrals from ChatGPT and other assistants to see real numbers for your own site.