The Traffic AI Search Won't Admit It Took

Reddit's chief executive put it plainly: Google's AI Mode "does not drive traffic." That is not a complaint from a small publisher watching its referral logs go quiet. It is the head of one of the largest content platforms on the internet, in a position to see exactly what Google sends back when it pulls Reddit threads into an AI answer, saying the number is close enough to nothing that he felt the need to say so in public. Google has not offered a rebuttal with its own figures. It has instead opened a Connected Apps Preview for AI Mode and Gemini, extending the surface area where its answers sit, while the basic question of what that surface sends back to the web remains unanswered.

This is the position brands are now expected to plan around: optimise for a channel whose own operator will not confirm, in any verifiable form, what it delivers.

The platforms disagree with each other, in public

It would be one thing if brands simply lacked access to Google's internal numbers. That is normal; platforms have always kept some of their machinery private. What is unusual here is that the people running adjacent platforms are contradicting the implied story in public statements, not in private briefings. Reddit's CEO didn't hedge. Shopify's CEO, separately, dismissed peer ratings on AI use as something that "didn't matter," another instance of a platform leader waving away a metric that had been presented elsewhere as meaningful.

When the chief executives who sit closest to the data are the ones undercutting the headline metrics, brands are left triangulating between contradictory signals from people who have no obligation to agree with each other, and no shared methodology that would let an outside observer judge who is closer to right.

Big brands can absorb the uncertainty; smaller sites can't

Reporting on AI search results has already found that big brands dominate the answers AI Mode and similar tools produce, while smaller sites are left with narrow openings rather than broad visibility. That asymmetry gets worse, not better, under measurement uncertainty. A large brand with its own data science team, a media mix model, and enough baseline traffic to smooth out noise can afford to treat "AI search traffic" as an unverified line item and keep spending regardless. A smaller site cannot run that experiment. It needs to know, with some precision, whether a given piece of content earns anything back from being cited in an AI answer. Without that, the rational response for anyone outside the big-brand tier is to stop trusting the channel at all, which only deepens the concentration the reporting has already flagged.

Meanwhile the pipeline feeding these answers keeps widening. Discord is preparing to put public chats onto pages that search engines can show. Google has adopted a new AI training control, and Microsoft's Bing has added new controls tied to AI training from AI search grounding. Each of these moves changes what goes into the retrieval systems that produce an AI answer. None of them comes with a parallel disclosure of what comes out the other side in terms of traffic, clicks, or revenue attributable to a named source.

Real money is measured; this traffic isn't

Compare this to a part of the digital economy where the numbers are solid enough to anchor an acquisition. DoorDash's purchase of Instacart is being described as a challenge to the $2 billion ad business Uber has built. That figure exists because advertising on delivery platforms produces auditable transactions: impressions, clicks, orders, revenue share, all trackable through systems built explicitly to be checked by advertisers, finance teams, and eventually regulators. Nobody is asking Uber's leadership whether its ad business "drives" anything. The ledger answers that question.

AI search traffic has no equivalent ledger. There is no standard unit that a brand can put in a spreadsheet next to its paid search numbers and defend in a budget meeting. The industry commentary that frames this as a gap to be "closed" by 2027 treats it as a maturity problem, as though better tooling will eventually catch up. But a gap implies two points that are moving toward each other. Here, one side (the brands asking for numbers) wants a shared measurement standard, and the other side (the platforms generating the answers) has shown no sign of wanting to supply one, and in at least one case, has had its own adjacent partner publicly say the number is effectively zero.

What brands are actually reporting upward

When a brand tells its leadership that AI search is "driving traffic," it is usually reporting one of a few proxies: a rise in referral-like sessions with ambiguous sourcing, a mention in an AI-generated answer that was manually spotted, or simply the absence of a decline that gets attributed to AI search by default because nothing else explains it. None of these are the same thing as a platform-confirmed, auditable number. They are inferences dressed up as metrics, and they get reported up the chain with the same confidence as a paid search conversion rate, which has an actual advertiser-facing dashboard behind it.

That confidence is the problem. A brand that cannot distinguish between "AI search sent us this many visits" and "we guessed AI search sent us some visits because the chart moved" is not measuring a channel. It is writing a story about a channel and calling the story a metric.

The honest position

The honest position for any organisation spending time or budget against AI search visibility right now is to say plainly that the number it reports does not meet the bar it would demand from any other channel. That is not an argument for ignoring AI search, since big brands dominating these results clearly matters commercially. It is an argument against pretending the traffic figure attached to that effort has been verified by anyone with the authority to verify it.

Until Google, or Bing, or any of the platforms shaping AI training and AI search grounding, publish a methodology that a third party can check the way an advertiser can check a delivery app's ad numbers, the figure brands put in front of their own leadership is not a measurement. It is a placeholder standing in for one, and the platforms that could turn it into something real have so far chosen not to.