The Greatest Customer Acquisition Channel Most Businesses Are Still Sleeping On

TL;DR:

Intent compression is when an AI system collapses research, comparison, and shortlisting into a single conversation before a buyer ever reaches a company's website. AI referral traffic is still tiny, just 0.14% of total web visits across 50,000+ sites analyzed by Semrush in 2025, but it grew 66% that year against 2.4% for traditional organic search, and Adobe's July 2026 data shows AI-referred retail traffic converting 60% higher than non-AI traffic for an 11th consecutive month. Conversion premiums for AI traffic range widely by methodology, from roughly 3x to 23x depending on the study, so no single multiplier should be treated as universal. The honest read: AI is not yet a proven cheaper acquisition channel, but it is already qualifying demand before your funnel sees it, and most companies have no one accountable for it.

I am Adnan, co-founder of Strategi, and this is the shift I think most businesses are underpricing right now: not that AI search exists, but that it has quietly become a customer acquisition channel with its own economics, and almost nobody has assigned anyone to own it.

What Is Intent Compression?

Intent compression is the process by which an AI system collapses research, comparison, qualification, and shortlisting into a single conversation, before a buyer ever visits a company's website. In the traditional buying journey, someone forms an intent, searches, opens several websites, compares options, and builds a shortlist over days or weeks. In the AI-mediated version, the buyer states their full requirement once, the assistant does the researching and filtering, and the buyer arrives at a recommended company already carrying a conclusion.

The difference is what the system knows before it starts. A search for "best accounting firms" hands Google a category and, at best, a city. The same buyer describing a 100-person company expanding into a new market, needing cross-border tax structuring, and explicitly ruling out large firms, hands an AI assistant company size, industry, timeline, requirement, exclusion, and purchase intent in one message. All of that context exists before any research happens, and all of it is available to the system doing the researching.

By the time that buyer reaches a recommended firm's website, they often already understand what the firm does, why it fits, which alternatives were considered and eliminated, and why this firm was worth a conversation. That is a structurally different visitor from someone who clicked a display ad out of curiosity, which is the most plausible explanation for why AI referral traffic keeps behaving strangely in the data below.

AI Referral Traffic Is Small and Growing Faster Than Everything Else

AI referral traffic remains a small fraction of total web traffic, and it is growing faster than any other channel measured. Semrush's analysis of more than 50,000 websites across 17 industries found AI traffic climbing 66% over 2025, from 462 million to 767 million monthly visits, while still making up only 0.14% of all measured traffic against 64.69% for direct and 16.04% for organic search. Paid search grew faster in percentage terms, up 76% over the same period, and organic search declined across 13 of the 17 industries Semrush analyzed, which matters: this growth is arriving alongside organic decline, not simply stacked on top of a stable base.

Globally, Similarweb's 2026 Generative AI Landscape report put average worldwide AI referral visits at 770.7 million a month between June 2025 and May 2026, more than double the year before. That figure counts only clicks out to external websites, not visits to the AI platforms themselves, which is a much larger and separately growing number.

Demandbase's own platform data tells a similar story from the B2B side. Drawn from more than 11 billion website visits across 1,584 customer accounts, the company reported monthly ChatGPT referral visits to the B2B sites it measures rising from roughly 645,000 in June 2025 to 2.6 million in June 2026, a 303% increase, with a sharp jump in May 2026 that more than doubled the prior run rate in a single month. That jump is worth pausing on. Demandbase never explains what caused it, and eight months earlier, a single undisclosed OpenAI ranking change had already cut ChatGPT's outbound referral traffic by roughly half in one month. A channel that can double or halve on an unannounced ranking change behaves less like a stable acquisition channel and more like a dependency on one company's product decisions. That volatility is exactly why we treat citation guarantees as indefensible at Strategi: the ground moves too fast for any promise about permanent placement to hold.

There is also a measurement problem underneath the growth numbers. Separate Similarweb research found that AI-recommended brands were roughly 2.5 times more likely to receive a website visit within seven days of being named, but only about 8.8% of those AI-influenced visits arrived as a direct, trackable referral. Most of the rest showed up in analytics as branded search instead. Referral-traffic figures like the ones above are real, but they are undercounting the total effect AI recommendations have on demand.

Does AI-Referred Traffic Actually Convert Better?

AI-referred traffic converts better than non-AI traffic in every major study measuring it, though the size of the premium varies enormously by methodology and should not be treated as one universal number. Adobe Digital Insights, analyzing more than 1 trillion visits to US retail sites, measured AI-referred traffic converting 60% higher than non-AI traffic in July 2026, the 11th consecutive month AI conversion has pulled ahead. The same July 2026 data showed AI-referred visitors spending 59% more time on site, adding items to cart at a 28% higher rate, and being 33% less likely to bounce, with a 14% higher overall engagement rate.

The trajectory behind that number is more interesting than the headline figure. A year earlier, AI-referred retail traffic was converting worse than average, not better. In March 2025, AI traffic converted 38% worse than non-AI traffic. By March 2026, the same measurement had flipped to 42% better, with 37% higher revenue per visit. That is roughly an 80-percentage-point swing in twelve months, and it did not happen gradually. It reads more like a threshold being crossed than a channel slowly maturing.

The most extreme individual example comes from Ahrefs, which published its own first-party numbers in June 2025. Over a 30-day window, AI search traffic accounted for just 0.5% of Ahrefs' total visits but drove 12.1% of all signups, a 23x higher conversion rate than traditional organic search, with AI-referred visitors browsing 50% more pages and 80% of that traffic landing on the homepage, free tools, and product pages rather than blog content. Ahrefs is explicit that this describes its own site, not an industry benchmark, and that AI search visitors click through so rarely that the ones who do are heavily self-selected for intent.

That self-selection point matters for the whole conversion-rate conversation. Independent estimates of the AI conversion premium range from roughly 3x to 23x depending on the study and the industry measured, with at least one large e-commerce study finding ChatGPT referrals underperforming traditional channels on conversion and revenue per session. The honest summary is not "AI traffic converts dramatically better." It is that AI traffic tends to be smaller, more qualified, and further along in the decision than traffic from a channel the buyer was interrupted into, and the size of that advantage depends heavily on what you are selling and to whom.

Buyers Have Already Moved, Even If Measurement Hasn't Caught Up

Buyer behavior has shifted well ahead of most companies' ability to measure it. Gartner surveyed 645 B2B buyers between August and September 2025 and found 45% had used generative AI during a recent purchase, mainly to gather information on vendors and products, with buyers reporting an average of seven information sources used overall. That last figure is the right corrective to any claim that AI has replaced the rest of the funnel. It has been added to it, not substituted for it. The same survey wave found 67% of B2B buyers prefer a sales-rep-free buying experience, a separate but reinforcing trend: buyers want to do more of the qualifying work themselves, with or without AI in the loop.

There is a genuine counterweight in Gartner's own data that is worth stating plainly rather than leaving out. The same buyers who are turning to GenAI still overwhelmingly want a human to check its work. 69% of B2B buyers prefer to validate AI-generated insights with a sales rep, and 51% say they are more likely to encounter misleading information from GenAI than from other sources. AI is qualifying demand earlier in the journey. It has not made the rest of the sales process optional.

In B2B software specifically, the shift is further along. G2 surveyed 1,076 software buyers and decision-makers in March 2026 and found 71% now rely on AI chatbots somewhere in their software research, up from roughly 60% seven months earlier, and 51% now start their research with an AI chatbot more often than with Google, up from 29% a year earlier. The influence is not passive. In the same survey, 69% of buyers said an AI chatbot's guidance led them to choose a different vendor than they originally planned, and a third bought from a vendor they had never heard of before the AI recommended it.

This pattern is not limited to B2B. McKinsey surveyed 1,927 US consumers in August 2025 and found half of them now intentionally seek out AI-powered search, with a majority of those users calling it the top digital source behind their buying decisions. McKinsey projects, as a forecast rather than a measured figure, that $750 billion in US revenue will funnel through AI-powered search by 2028.

Meta Creates Demand. Google Captures It. AI Qualifies It.

The clearest way to place this channel is against the two that already dominate most acquisition budgets. Meta is built to create demand: the buyer was not looking for anything, an ad interrupts a feed, and interest gets manufactured from nothing. Google is built to capture demand: the buyer already has a problem, goes looking for a solution, and a ranked page or paid ad meets them there.

AI search does neither of those jobs. It understands and qualifies demand that already exists. The buyer explains their situation once, and the system researches alternatives, eliminates poor fits, and returns a shortlist with reasoning attached. This does not make Meta or Google less important. It adds a layer above both of them, and that layer increasingly decides which companies even make it into the consideration set those other channels are competing for.

The uncomfortable version of that statement is this: a company can now lose a customer without that customer ever seeing its ad, its search ranking, or its website. If an assistant names three competitors in an answer and leaves one out, that company was never in the running to begin with. It never had a chance to compete on price, creative, or landing-page quality, because the elimination happened inside a conversation it had no visibility into.

What Actually Gets a Company Recommended

Getting recommended does not reduce to ranking well inside a chat interface, and treating it that way produces shallow work. An AI assistant recommends a company when it has enough corroborated evidence to conclude the company is genuinely the right answer to a specific question. That evidence draws on a wide surface: clear positioning and defined services, stated industry and audience specificity, technically parseable website content, third-party mentions and reviews, coverage in independent publications, comparison content, and any place a claim a company makes about itself gets independently confirmed.

The objective is not to manipulate a model into saying a name. It is to make sure that when a model looks for a reason to recommend a company, a legible one exists, sourced from somewhere the model already trusts.

The Prompts That Actually Matter

Visibility only matters when it attaches to commercially relevant conversations. A company mentioned across hundreds of informational prompts is worth less than one recommended consistently across a small number of genuine buying questions. The prompts worth optimizing for look like specific buying events with a shortlist attached: which commercial interior design firm should handle a large office build-out, which cybersecurity provider fits a mid-sized fintech company, which manufacturer can supply apparel at scale for a national dealer network, which broker should manage a significant real estate purchase. Each of those is a decision with money and a timeline behind it, not a curiosity question.

The chain worth building runs from a specific commercial prompt, to a recommendation, to a qualified inquiry, to a customer. It does not run from raw mention counts to impressions to a visibility score that cannot be tied to a dollar. One thing worth stating plainly: there is no robust cross-industry evidence yet that AI search is a cheaper acquisition channel than Google Ads, Meta Ads, or SEO. Building the evidence base that earns a recommendation still costs real money and real time. What is structurally different is that organic AI recommendations are not governed by a cost-per-click auction. Companies compete on evidence, positioning, and independently verifiable authority rather than by outbidding a competitor for the same click.

Nobody at most companies is accountable for any of this yet. There is usually a person for SEO, a person for paid search, a person for social, and no one whose job is making sure an AI system recommends the business when a buyer describes a problem it solves. That gap, more than any single statistic above, is what makes this window worth paying attention to now rather than once every competitor already has. What we work on at Strategi is turning that gap into an owned, measured part of how a business gets found.

Read more about Strategi and what we do.

^{[1][2][3][4][5][6][7][8][9][10]}