A number stopped me last week, and it’s stayed with me since.
A mid-market cybersecurity firm went digging through its own GA4 and CRM. Over a year it picked up 167 sessions referred from AI tools: ChatGPT, Perplexity, the usual names. Those 167 visits turned into 23 qualified enquiries. Booked calls and demo requests, the real ones, not newsletter signups.
In the same window, Google organic sent that firm 3,840 sessions. Those produced 96 enquiries.
Put the two side by side and it gets a bit uncomfortable. AI was 4% of the firm’s traffic. It drove 19% of its qualified pipeline.
👉 4% of sessions, 19% of pipeline.
That’s the line you repeat in a Monday budget meeting, the one that makes a CMO move spend. And it’s the kind of number racing through marketing content right now, usually boiled down to one tidy stat: AI visitors convert at 14.2%, Google organic at 2.8%. Roughly five times better.
I’ve seen that exact pairing, 14.2% and 2.8%, down to the same two decimal places, in at least half a dozen blogs this month. That’s the point where my curiosity turned into suspicion.
Where that stat actually comes from — and where it doesn’t
The cybersecurity case study is real. It sits inside the 2026 AI Search Benchmark Report from Opollo, a small agency that does marketing for IT-services firms. The report tracked 312 companies that sell IT and technology services to other businesses, from $2M regional managed-service providers to $80M cybersecurity consultancies, across North America, Australia and the UK, from January 2025 to January 2026.
And credit where it’s due, the methodology is disclosed. They excluded firms with under 10 AI sessions a month, stripped out outliers above 35% conversion, reported both mean and median, and manually validated how the traffic was classified. Conversion was defined narrowly: qualified enquiries only. Mean AI conversion came out at 14.2%, Google organic at 2.8%. Roughly a 4–5x gap.
So the number isn’t fabricated. But look at what it actually is. One boutique agency, 312 IT firms, three English-speaking markets. That is the entire foundation for a stat now quoted as if it were an internet-wide law.
This is the bit that nags at me. That same 14.2%-vs-2.8%, same decimals, turns up on RankScience, an SEO agency for startups, attributed not to Opollo but to a separate “12 million website visits” analysis credited to something called “Superprompt.” A different study, a different claimed sample, the same two decimal points, and no shared methodology anywhere in sight.
I went looking for that Superprompt report. I couldn’t find a checkable version of it independent of the blog citing it.
The irony almost writes itself. The whole premise of answer engine optimisation (AEO) is that AI should cite its sources accurately. And the marketing industry, writing breathlessly about exactly that, is passing one number hand to hand, between reports that never spoke to each other, without checking where it came from.
The visitors really do convert better — the exact number is just fuzzy
Let me be fair here, because I think the direction is right. AI-referred visitors do convert better, and the mechanism makes sense to me.
When someone lands on your site from Google, they’re still comparing. When they arrive from ChatGPT or Perplexity, the model has already compared, filtered and recommended you. The click comes after the shortlist, not before it. So the visitor turns up warmer.
And the corroboration doesn’t all trace back to Opollo, which is the part that matters. Conductor — a marketing-software company whose enterprise platform helps brands with SEO and answer engine optimisation — citing Knotch data, found LLM-referred visitors convert at twice the rate in a third of the sessions of other traffic. Semrush ran its own study, over 500 high-value topics, published mid-2025, and put the premium at 4.4x. HubSpot cites the same number, though without saying where its own figure comes from, so treat that as one data point, not two. Seer Interactive, a data-driven marketing agency, measured ChatGPT traffic converting at 15.9% against Google organic’s 1.76%. Shopify has reportedly seen AI-referred sessions converting around 50% higher on product pages, with ~14% higher order values, though those are secondhand figures I couldn’t verify against Shopify’s own data.
Different studies, different samples, different definitions of “conversion”, all pointing the same way. That is what real corroboration looks like, and it convinces me far more than one number repeated word for word.
The volatility nobody puts in the headline
What the tidy stat leaves out is simpler: this signal is new, and it swings.
Adobe’s data on the very same channel flipped from 38% worse than other traffic in March 2025 to 42% better in March 2026. Call it an 80-point swing in twelve months. A channel that can reverse direction inside a year is still an emerging behaviour finding its shape, not a settled law.
And Adobe published that flattering 42% figure on the same day it launched its own LLM-optimisation product. Adobe’s own methodology note admits the figure isn’t independently audited. I’m not saying it’s wrong. I’m saying that when a vendor’s headline number and its new product launch land on the same morning, you read the number with one eyebrow raised.
The volume, meanwhile, is still tiny. Conductor’s study of 13,770 domains found AI referrals make up 1.08% of all website traffic. High intent, low volume, still immature, all true at the same time.
So the honest picture is a warm, promising, fast-growing sliver of traffic, and not something to bet the whole budget on this quarter.
Why the CMO surveys contradict each other so wildly
This is where it gets genuinely confusing for anyone trying to benchmark themselves.
Read one survey and you’ll feel hopelessly behind. Conductor’s CMO report says 73% of organisations already call their AEO programs “advanced or very advanced,” 94% plan to increase investment, and AEO already eats 12% of their digital budget.
Pick up another and the panic drains away. Acquia, a digital experience software company, found only 20% of marketers have actually started implementing AEO, which means about four in five haven’t. Minuttia, a content agency for B2B SaaS, found that even among people who say they “do AEO,” only around 29% have anything resembling a resourced, active program. The trade publication Digiday found 27% haven’t changed their strategy at all, and 16% aren’t even familiar with the terms. Modus and Semrush found 85% of CMOs call it a priority but rate their own confidence at 7.5 out of 10, with unclear KPIs as the top complaint.
“73% are advanced” and “80% haven’t started” sound like they can’t both be true. They can. The two surveys were simply looking at different rooms.
Conductor surveyed enterprise only, companies with 500 or more employees. Of course they’re advanced; they have the budgets and the teams. The broader-population surveys count everyone else. So the gap isn’t really a mystery. It comes down to who was in the sample.
Which matters enormously for who’s reading this.
What this means if you’re not running an enterprise budget
Most of the marketers I work with sit firmly in that broader population. A 40-person D2C brand in Bengaluru. A B2B founder in Pune doing the marketing between sales calls. They aren’t in Conductor’s room, and they never were.
So when that reader sees “94% of CMOs are increasing AEO investment” and feels behind, I want to be clear about what’s happening. That stat came from companies with 500 or more employees spending 12% of a large digital budget. It didn’t come from anyone who looks like your team.
You’re not losing an enterprise race. You were never entered in it.
And the barriers are exactly what you’d expect. Across the surveys, budget and expertise top every list. Acquia found 45% blocked by budget, 40% by a lack of internal expertise. Minuttia found around 40% of respondents, including teams already doing AEO work, have no dedicated budget line for it at all. Most of the people “doing” this are doing it on structure rather than money.
Which is the good news, really. Structure is cheap.
A playbook that fits a resource-constrained team
If I had a small budget and no AEO team, this is roughly the order I’d work in.
First, fix your measurement. You can’t manage what collapses into “direct.” One dataset found 70.6% of AI-originated traffic arrives without referrer headers, so it lands in analytics as direct traffic and quietly disappears. Segment your AI referrals deliberately in GA4 before you touch anything else. If you can’t see the channel, every argument you make about it is really just a guess.
Second, define conversion the way Opollo did. Qualified enquiries: booked calls, demos, real contact submissions. Not newsletter signups, not “engaged sessions.” The 5x story only holds when you measure quality. Measure signups instead and you’ll flatter yourself into a bad decision.
Third, audit what actually gets cited. Conductor ranked the content types that earn AI citations, in order: blog content, video, articles, news, product pages. Look at your own top-cited pages, then make more of what already works.
Fourth, treat citations as a three-to-six-month project rather than a two-week schema sprint. FAQ schema and structured content help, but topical authority, being genuinely, repeatedly the clearest answer in your category, is what earns the citation. That work is slow. It also lasts.
And the India reality, plainly: no source I trust gives an AI-referral conversion number specific to India. The headline stats are global, US-heavy and B2B-heavy. Indian ecommerce, meanwhile, is still overwhelmingly mobile and search-query-driven. Industry estimates put roughly 78% of ecommerce traffic on mobile and more than 60% of purchase journeys starting with a search query. Nobody has measured the AI-referral share for India specifically either, but if the global pattern holds, and Conductor puts AI referrals at around 1% of traffic industry-wide, India’s share is probably similar or smaller, given how much more search- and mobile-driven the market still is.
So the real ROI case today is quieter than “reallocate the media budget.” It’s this: get structurally ready and get cited, cheaply, now, so that when the sliver becomes a slice, you’re already in the answer.
The point
An industry built entirely on the promise of getting cited accurately by AI has, so far, mostly cited each other inaccurately, one narrow study’s decimals laundered through a dozen blogs that never checked.
I’m not telling you to ignore the stat. AI-referred visitors genuinely do convert better, and that part is strategically real. What I’m saying is this: verify before you rebuild your roadmap around a number you saw six times this week. Which, when you think about it, is exactly the discipline AEO is supposed to reward in the first place.
Do the cheap, structural work, and measure it honestly. The panic can wait…..
🔹 Have you actually segmented your AI-referred traffic yet, or is it still hiding inside “direct”? And has anyone reading this checked what those viral 14.2% and 2.8% numbers are really built on? I’d genuinely like to hear from the ones who’ve looked.