
Three studies cited by Search Engine Roundtable and summarised by PPC Land point the same way: Google generated 96.5% of conversions across 100 sites analysed by Marie Haynes, ChatGPT sessions were 0.62% of Google organic volume, and SEO Works found 39 times more Google traffic than AI assistant traffic. Samples are small, but the practical conclusion is clear: for most SMEs, budget and attention should stay on channels that already bring customers.
What the three studies say
According to PPC Land, citing Search Engine Roundtable: Marie Haynes analysed 12 months of GA4 data from 100 sites and concluded ChatGPT sessions represent only 0.62% of the volume Google organic brings. ChatGPT converts at a higher rate, but Google generated 96.5% of conversions. SEO Works found that Google organic traffic was 39 times that of AI assistants. And Kevin Indig, with 12 sites, found no reliable relationship between AI visibility and brand or direct traffic.
The authors admit that the absence of a relationship found is absence of evidence, not proof there’s no effect.
The limitations worth acknowledging
Samples are small and generally sites from specific contexts. The data covers the year before October 2026, and this is a channel that changes fast. An assistant may influence decisions that later materialise through brand search or direct access, without appearing as a ChatGPT session. And a higher conversion rate on a tiny volume may be noise.
This doesn’t say AI doesn’t matter. It says that, today, and with this data, the measured traffic coming from it is a small fraction. To decide budget, you measure what brings customers, and the available data points to the classic channels.

ChatGPT vs Google traffic (100 sites)
0.62%
of Google organic volume comes from ChatGPT
Marie Haynes, 12 months of GA4 across 100 sites: Google generated 96.5% of conversions. SEO Works: Google 39 times AI assistants’ traffic.
The counterpoint: AI Overviews and visibility
In parallel, Semrush indicated Google’s AI Overviews appear in 88% of commercial keywords in a study of 400 keywords in the US (97.08% weighted by search volume). It measures only appearance, without clicks or traffic, and PPC Land notes Semrush sells the tools it recommends.
Putting the pieces together: AI answers appear a lot but send few clicks. There may be value in being cited, but it’s hard to measure, and there’s no proof it pays to spend a lot to achieve it. Prudence is the word.
What this means for your budget
For an SME in Portugal, the practical rule is simple: distribute the budget where cost per customer is known and within your limit. That means, as a rule, paid search, well-measured social ads, SEO, email and referrals. Reserve a small part (for example 5% to 10%) to test new channels, including AI, if you can measure.
The mistake would be moving budget from channels that work to channels that don’t yet have proof. An SME has no room to pay the price of being a pioneer unless the bet is small and learned from.

Budget follows what brings customers, not what's fashionable.
What to do in SEO and content
There’s no reason to stop making useful, clear and correct content: it serves customers, serves traditional search and may serve citation in AI answers. What makes no sense is paying for “AI optimisation” with promises of results, without proof.
If an agency sells you AI visibility, ask: how do you measure? What data do you have on traffic and conversions coming from AI? On which sites? With what sample? The October studies show the question is legitimate and the honest answer is often “it’s still hard to measure”.
How to measure traffic from AI in your account
In Analytics, create a segment or source report that identifies known referrals from AI assistants, and track sessions, conversions and value. Compare with other sources. If the volume is tiny, record it and move on; if it grows, dig deeper.
Also keep a record of customers’ remarks (“I saw it on ChatGPT”, “the assistant recommended you”), even informal. Direct contact reveals what traffic analysis doesn’t see. With a few months of records, you’ll have a concrete idea of AI’s weight in your business, without depending on third-party studies.
Questions for your manager or agency
What percentage of our traffic and conversions comes from AI assistants? How do we measure? What part of the budget goes to AI tests and with what criterion to continue? What do we do if results are weak? And how do we protect the budget of channels that work?
Good answers are concrete and humble. Bad answers are enthusiastic and vague. In October 2026, with the data available, the professional attitude is to test with discipline, not bet on faith.
Practical conclusion for whoever decides budgets
Keep the essentials on what brings proven customers, reserve a small slice to test AI, measure with a source report and ask agencies for proof, not promises. When the data changes, and it will, adjust. Deciding with today’s numbers doesn’t prevent changing your mind with tomorrow’s.
Frequently asked questions
Is ChatGPT of no use then?
That isn’t what the studies say. They say measured traffic is small. It may influence decisions in ways hard to measure. Test with discipline.
Do the studies apply to Portugal?
Samples are small and not specific to Portugal. Use them as an indication and confirm in your site’s data.
Should I change my SEO strategy?
Not because of this. Keep producing useful and correct content, and distrust promises of AI visibility results.