AI Search Layers Over Google, Not Replacing It – Similarweb

Rand Fishkin, co-founder and CEO of SparkToro, posted a bold prediction on LinkedIn on July 24, 2026: that Similarweb’s latest report would “probably infuriate” two opposing camps. On one side, the AI zealots who believe every marketing dollar belongs in chatbot visibility. On the other, the AI skeptics who dismissed the hype as inflated but lacked the hard data to prove it to a boss suffering from what Fishkin calls “AI derangement syndrome.”

Most industry reports soothe one faction and irritate the other. Similarweb’s “2026 Generative AI Landscape: The Evolution of AI Search” breaks that pattern, handing both sides a data point that shatters their certainty. I read all 38 pages because I value ground truth, and I want to break down why Fishkin is right, what the report actually reveals, and what you should change in your strategy starting tomorrow.

The Stat That Should Unnerve AI Zealots

Similarweb tracked audience overlap between ChatGPT and Google from March to May 2026. Of ChatGPT’s 494 million users, 461 million (95%) also used Google during the same period. Almost nobody has abandoned Google for ChatGPT. Instead, they’ve added ChatGPT to a Google habit that remains unchanged.

Zoom out further, and the disparity grows starker. Search still attracts 3.3 billion average monthly unique visitors worldwide. AI chatbots, despite growing 57% year over year, sit at just 655 million. That means search is roughly five times the size of the entire AI chatbot category combined. If your 2026 budget deck assumes AI search has already eclipsed traditional search, this report’s math contradicts you.

Citations tell a similar story from another angle. Only 6.8% of ChatGPT answers in the U. S. included a link to an external source as of May 2026. That’s up more than fivefold from roughly 1% a year earlier, which is genuinely fast growth. But it also means 93 out of every 100 ChatGPT answers still send nobody anywhere. Ethan Smith of Graphite makes a sharper point in the report: users are folding those same prompting habits back into Google itself, with average search length climbing steadily since AI Mode launched. People aren’t abandoning search boxes. They’re just typing longer sentences into them.

The Stat That Should Unnerve AI Skeptics

Now for the half of the report that punctures the other side’s confidence. Average monthly web visits across generative AI platforms hit 9.5 billion between June 2025 and May 2026, up 70% year over year. App downloads worldwide climbed to 2.7 billion, up 134%. Half of all generative AI users are now 35 or older, compared to 61% under 34 just two years ago. That’s the clearest signal I’ve seen that this isn’t a Gen Z fad running out of steam. Michael Horrocks of Miro puts it plainly: “Growth concentrated in younger demographics can fade with trends; growth spreading into older generations is often what durable, mainstream adoption looks like.”

Meta AI’s own disclosed numbers back that up from a completely different angle. Publicly reported monthly active users went from 384 million in September 2024 to 1.2 billion by March 2026, more than tripling in 18 months, entirely by riding inside Instagram, Facebook, WhatsApp, and Messenger rather than as a standalone destination anyone had to seek out. And ChatGPT ad penetration in the U. S. jumped from 14% of desktop chats in May 2026 to 26% just one month later. Whatever you think about the maturity of AI search, advertisers clearly don’t consider it a toy.

My read is that both camps are pattern-matching off the piece of data that confirms what they already believed, and both are ignoring the half that complicates it. AI search hasn’t replaced anything. It has stacked a new, fast-growing, unevenly distributed layer on top of a search ecosystem that was already there. The practitioners who will win the next two years are the ones measuring the stack instead of arguing about which layer matters more.

Why the Disconnect Between Citations and Clicks Matters More Than Either Camp Realizes

The most useful chart in the entire deck, and the one I think gets underappreciated in the LinkedIn debate, comes from Aleyda Solis of Orainti. She points out that 65% of the URLs ChatGPT cites sit two or three folders deep, the pages doing the actual evidentiary work behind an AI answer. But 58.8% of the referral traffic that AI sends back to sites lands on the homepage, not the cited page at all. Cited pages and clicked pages are almost entirely different populations of URLs.

That single data point should reorganize how agencies report AI performance to clients. If you’re only tracking whether your deep product or blog content gets cited, you’re missing the fact that the humans who actually click through are landing somewhere else entirely and need their own conversion path. Fishkin makes a related point in the report, comparing this to how 20th-century advertisers proved billboard and radio spend worked by measuring lift in store visits rather than counting who glanced at the sign. The mechanism has changed. The discipline of measuring downstream behavior instead of surface impressions hasn’t.

3 Things to Change in Your Strategy This Week

1. Split your reporting into two separate metrics. Track citation rate and citation folder depth as one key performance indicator that measures whether AI trusts your content enough to use it as evidence. Track referral landing pages and downstream conversion as a completely separate KPI that measures what actually happens once a human clicks through. Conflating the two in a single dashboard is how brands miss both problems at once.
2. Stop treating “AI visibility” as a single category. Similarweb’s brand visibility index shows how category-specific this already is. CeraVe leads beauty at an index of 100 while NYX Cosmetics sits at 19 in the same category. Kevin Indig of Growth Memo argues in the report that share of voice is the metric that matters because it’s a relative comparison in a stochastic system, not an absolute score. Pull your own category’s leaderboard before you assume you’re winning or losing.
3. Match your content to the platform’s actual audience, not the platform’s overall size. The affinity data shows ChatGPT skews toward everyday consumers researching restaurants, health, and fashion. Claude users are 25 times more likely than the average searcher to visit university sites and skew heavily toward professionals and students. Gemini users over-index on graphics, security, and hardware content. A single piece of “AI-optimized” content aimed at all three is aimed at none of them.The Bottom LineI don’t think the AI zealots are wrong that something structural is shifting. Nine and a half billion monthly visits and a fivefold jump in citation rates in under a year are not a rounding error. But I also don’t think the skeptics are wrong that most of the industry’s AI panic is running well ahead of the actual traffic numbers, given that Google still commands five times the audience of every AI chatbot combined and 95 out of every 100 ChatGPT users never left Google in the first place. Fishkin’s post nailed the discomfort because the data refuses to let either side keep its story simple. The brands that will actually benefit from this report aren’t the ones picking a side. They’re the ones pulling the citation and referral numbers for their own category this quarter and building a strategy around what the data says rather than what the argument on LinkedIn says.