Three studies landed in the same week, all pointing in the same direction. AI search - ChatGPT, Perplexity, Google AI Overviews, Gemini - is sending a tiny fraction of the traffic that conventional Google organic search delivers. The coverage was predictable: headlines declaring that AEO and GEO programmes are not worth the investment. Some marketing directors will read those headlines and draw the wrong conclusion.
This is not a case for ignoring the data. The traffic gap is real and the studies deserve a serious read. But traffic volume is one metric in a much larger picture, and drawing budget decisions from it alone is a mistake. Here is how to think about this properly.
What the Studies Actually Measured
The studies all converge on click-through traffic referred from AI platforms compared to traffic from Google Search. On that measure, there is no contest. Google Search still drives orders of magnitude more visits to most websites than ChatGPT, Perplexity, or AI Overviews combined. If your success metric is raw referral session volume, AI search currently loses badly.
But traffic volume studies have a structural problem when applied to AI search: they measure what is trackable in analytics platforms. AI referrals are notoriously undercounted. Sessions arriving from ChatGPT and Perplexity frequently appear as direct traffic or are misattributed entirely, particularly when users copy a URL from a chat interface rather than clicking a hyperlink. GA4 does not reliably separate these sessions without custom channel grouping and referral source rules. The studies likely undercount AI-referred traffic to a meaningful degree.
There is also the question of what AI search is actually doing. A large proportion of AI search interactions do not produce a click at all - the user gets an answer within the interface. That is a genuine concern for click-dependent business models. But it also means a brand mentioned in those answers is influencing a decision without generating a session. Traffic studies cannot see that influence.
Volume vs Intent: The Metric That Actually Matters for Lead Gen
The question is not whether AI search sends more or less traffic than Google. The question is what that traffic does when it arrives. Anecdotally and across client programmes, AI-referred sessions tend to arrive later in the buying cycle. A user who has spent ten minutes in a ChatGPT conversation researching software options before clicking through to your site is a qualitatively different visitor to someone who clicked a broad-match keyword on the first search they ran.
If you are running a lead generation programme and you have CRM attribution set up properly - source captured at form submission, carried through the pipeline, and reported against closed revenue - then a small volume of high-intent AI-referred leads may deliver a cost-per-acquisition that justifies the programme entirely. The studies say nothing about this because they do not look at it. They measure sessions, not outcomes.
This is where most AI search debates go wrong. Practitioners making the case for AEO investment need to stop relying on traffic projections and start measuring what happens to AI-referred enquiries in the CRM. If those leads convert at a higher rate and close at a higher value, the traffic volume comparison with Google becomes much less relevant to the budget decision.
The Attribution Gap Is Still the Core Problem
The reason these studies carry so much weight is that most businesses cannot counter them with their own data. If your GA4 setup does not explicitly tag ChatGPT, Perplexity, and Gemini as referral sources in a custom channel group, you have no visibility into AI-referred sessions at all. If your form submissions do not capture UTM parameters or referrer data and pass them into your CRM, you cannot connect an AI-referred visit to a won deal six weeks later.
Gemini referrals via Google properties now pass UTM parameters in some configurations, which is a step forward. But the wider problem remains: the measurement infrastructure most businesses run was built for last-click attribution from Google Search, and it is poorly suited to capturing AI referrals. Until that is fixed, businesses have no ability to rebut studies like these with their own evidence - so the studies fill the vacuum.
The practical fix is not complicated, but it requires deliberate setup. The key steps are: custom channel groupings in GA4 to capture known AI referrers; hidden form fields to capture the referring source; CRM fields that store and preserve that source through the pipeline; and regular reporting that surfaces AI-source leads and tracks their progression to revenue. Without this, decisions will default to whatever the latest study says.
What This Means for How You Structure an AEO Programme
The studies should prompt a rethink of how AEO programmes are justified internally - not whether they should exist. A programme built on traffic volume projections is going to struggle against data showing AI platforms send a small fraction of Google's referrals. A programme built on brand visibility in AI responses, conversion quality from AI-referred sessions, and CRM-tracked revenue attribution is a much harder target to hit with a single week's worth of aggregate traffic studies.
There is also a timing argument worth making clearly. AI search is not mature. ChatGPT reached mass adoption faster than almost any consumer product in history. The proportion of searches conducted via AI interfaces is growing, even if the absolute referral numbers are small today. Getting the infrastructure right - the entity signals, the structured content, the citation-building through digital PR, the measurement setup - takes time. Waiting until AI search is sending comparable traffic to Google before investing means starting years behind the brands that moved earlier.
The studies are a useful corrective to the more breathless claims made about AI search traffic. They are not a reason to stop building for it. They are a reason to build the measurement layer that lets you make the case on your own data rather than relying on projections. If you cannot do that yet, that is where to start.
The Practical Checklist Before You Make a Budget Decision
Before concluding that AI search investment is not worth it, check whether you can actually see it. Are ChatGPT, Perplexity, and Gemini appearing as identifiable sources in your analytics? Are you capturing referrer data at the point of form submission? Are AI-referred leads tracked through your CRM to close? If any of these are missing, the absence of evidence is not evidence of absence - it is a gap in your measurement setup.
The studies that landed this week are measuring aggregate website traffic. They are not measuring your business, your category, or your buyer's research behaviour. Run the analysis on your own data first. If AI search is genuinely sending low-quality, low-volume traffic that does not convert, that is worth knowing. But make that call with your numbers, not someone else's.