Is GEO worth it? For most businesses, yes, but not for the reason the sales decks give. Three recent studies found AI assistants send a tiny share of website traffic compared with Google, and that part is true. The case for generative engine optimization rests on other things: visitor quality, being named when buyers ask an assistant, and fixes that also help ordinary search.
We’re an AI-search agency, so you should read us with that in mind. This post walks through what each skeptical study measured, where it stops being useful, and where we think GEO spend is justified and where it isn’t. It sits inside our AI search work and builds on two earlier pieces, how to measure AI search ROI when the click disappears and what AI referral traffic does once it lands.
What do the three studies actually say about AI search traffic?
They say AI assistants send a very small slice of traffic compared with Google, and that the conversion advantage people talk about does not make up for the gap in volume. Search Engine Roundtable’s Barry Schwartz rounded the three up on October 22, 2026 under a headline suggesting AI visibility may not be worth investing in.
The first study is Marie Haynes’ “GEO is a distraction,” published October 21, 2026. She looked at 12 months of GA4 data across 100 websites of varying sizes, more than 145 million sessions in all. ChatGPT sessions came to 0.62% of the volume Google organic drove, or 0.73% with all other LLMs added in.
The second is a benchmark from the UK agency The SEO Works covering 95 GA4 properties from September 2025 to August 2026. Across that period those sites got 2,825,594 organic sessions and 72,095 AI referral sessions, which the agency summed up as organic delivering 39 times more traffic. The page shows a last-modified date of October 14, 2026 and no separate publish date.
The third is Kevin Indig’s Growth Memo post from October 19, 2026. It asks a different question: is AI quietly showing up as direct traffic? Across 12 sites, he couldn’t detect it.
Do AI visitors convert better than Google visitors?
Sometimes, and by very different margins depending on who ran the numbers. In Haynes’ data ChatGPT visitors converted at 13.8% against 3.02% for Google organic. In The SEO Works data, AI referrals converted at 7.18% against 6.08% for organic.
Same direction, wildly different size. The SEO Works study also found AI traffic converted worse than organic on 23 of the 63 sites that had conversion data, and for the 10 B2B product businesses in its sample the pooled ratio was 0.62x, meaning AI visitors converted worse there. So “AI traffic converts better” is a tendency, not a rule you can build a forecast on.
What can these studies not tell you?
Quite a lot, and the authors say so themselves. Haynes calls hers a “small dataset,” and 45 of the 100 sites she sampled were not tracking any key events at all. That leaves 55 sites for any conversion math, and the article doesn’t explain how the sites were chosen or what counted as a conversion.
The SEO Works page describes its measured AI contribution as a floor. It identifies AI visits by matching session sources like chatgpt, claude, perplexity and gemini, and it leaves out Google AI Overviews entirely because those clicks arrive as ordinary organic sessions. It is also published by an agency that sells SEO services, which doesn’t make the data wrong but is worth knowing.
The blind spot all three share: they count visits that reach your site. They can’t count the buyer who got your name from an assistant, never clicked, and typed your brand into Google later. That person shows up as branded organic or direct, and none of these studies can separate them out.
Is AI traffic hiding inside your direct traffic?
Probably not in any amount you could measure. Indig tested whether AI exposure turns up as extra direct visits, and across 12 sites he found it wasn’t detectable. On average, direct sessions with identifiable AI traces were only 0.003% of total direct traffic.
His sample is small and skews toward B2B, and his own conclusion is narrow: he couldn’t detect hidden AI traffic without self-reported attribution. He also found that direct traffic itself looks inflated by bots. In one filtered sample, total direct sessions rose 140.9% from January to September while engaged direct sessions changed by about minus 1%. At one site, GA4 logged 409,000 direct sessions and only 198 counted as engaged.
The practical lesson is a good one even if you ignore GEO completely: check your direct traffic for bots before you credit it to anything.
Where does GEO spend still make sense?
It makes sense where the cost is low and the payoff doesn’t depend on AI clicks. That is our view, not a finding from any of these studies, and you should test it against your own numbers.
- Fixes that also help regular search. Clear page structure, direct answers near the top, accurate schema and clean crawlability help classic SEO whether or not an assistant ever quotes you. The SEO Works study even found AI growth and organic growth moving together across 61 websites (a correlation of +0.557), though correlation says nothing about which causes which.
- Making sure you’re named at all. If a buyer asks an assistant for options and you’re absent, no analytics report will ever show the loss. Our piece on brand mentions versus citations covers the difference.
- Pages that already get AI visits. If you can see AI visitors landing somewhere, tightening that page is cheap.
It makes less sense as a large, standalone budget line justified by promised traffic. The studies give you no basis for that promise.
What should you fix first if you only have a small budget?
Start with measurement, then the pages, then the mentions. In that order, because each step tells you whether the next is worth paying for.
- Set up the tracking. Build a GA4 channel grouping for the main assistants and add a “How did you hear about us?” field to your forms. Without it you’re guessing.
- Clean your direct traffic. Compare direct sessions with engaged direct sessions before you read anything into the totals.
- Fix the pages assistants already send people to. Lead with the answer, the price or fit, and proof.
- Check whether you’re named. Ask the assistants your buyers use the questions your buyers ask, and write down who comes back.
- Set a review date. Pick a quarter, re-run the numbers, and decide with your own data whether to spend more.
Small businesses and local shops should keep the ask modest. The studies give no sample you can lean on for a business your size, so run the cheap steps and let your own conversion data make the case.
The skeptics are right about volume: AI assistants send a small fraction of what Google does, and the conversion edge is inconsistent. They can’t see the buyers who were influenced and never clicked. Spend small and early on tracking, page fixes and being named, skip any budget justified by traffic forecasts, and let your own data decide the next step.
Not sure where AI search fits for your business?
We’ll check what assistants say about you today, what your analytics can and can’t see, and which fixes are worth the money.
Sources
- Marie Haynes, “GEO is a distraction,” October 21, 2026 (12 months of GA4 data, 100 websites, over 145 million sessions; her dataset is not published).
- The SEO Works, “How Much Traffic Do AI Assistants Actually Send, And Does It Convert Better?” AI referral traffic benchmark, 95 GA4 properties, September 2025 to August 2026; page last modified October 14, 2026, no separate publish date shown.
- Kevin Indig, “Does AI drive more direct traffic?” Growth Memo, October 19, 2026 (12 GA4 properties).
- Barry Schwartz, “3 New Studies Suggest AI Visibility & Search Not Worth Investing In,” Search Engine Roundtable, October 22, 2026.
- Note: the section questions above are written as likely follow-up questions; we did not pull keyword demand data for this post, so treat their search demand as unverified.