AI Search Isn’t Replacing Google, It’s Layering On Top – Similarweb Data

Jul 31, 2026 02:00 AM - 1 month ago 684

Rand Fishkin, the co-founder and CEO of SparkToro, posted connected LinkedIn connected July 24, 2026, to opportunity that Similarweb’s newest study would “probably infuriate” 2 groups astatine once. The AI zealots who are definite each trading dollar should already beryllium chasing chatbot visibility. And the AI skeptics who’ve suspected the hype was overblown but ne'er had the receipts to beryllium it to a leader suffering from what Rand calls AI derangement syndrome.

Most manufacture reports comfortableness 1 campy and annoy the other. This one, Similarweb’s “2026 Generative AI Landscape: The Evolution of AI Search,” manages to manus some sides a information constituent that undercuts their certainty. I publication each 38 pages because I’m into ground-truthing, and I want to locomotion done why Rand is right, what the study really shows, and what you should do otherwise tomorrow greeting because of it.

The Number That Should Worry The AI Zealots

Similarweb tracked assemblage overlap betwixt ChatGPT and Google betwixt March and May 2026, and 461 cardinal of ChatGPT’s 494 cardinal users (95%) besides usage Google successful the aforesaid window. Almost cipher has near Google for ChatGPT. They’ve added ChatGPT to a Google wont that hasn’t budged.

Zoom retired further, and the spread gets starker. Search still pulls 3.3 cardinal mean monthly unsocial visitors worldwide. AI chatbots, moreover aft increasing 57% twelvemonth complete year, beryllium astatine 655 million. So, hunt is still astir 5 times the size of the full AI chatbot class combined. If your 2026 fund platform assumes AI hunt has already eclipsed accepted search, the mathematics successful this study says otherwise.

Citations show the aforesaid communicative from a different angle. Only 6.8% of ChatGPT answers successful the U.S. included a nexus to an outer root arsenic of May 2026. That’s up much than fivefold from astir 1% a twelvemonth earlier, which is genuinely accelerated growth, but it besides intends 93 retired of each 100 ChatGPT answers still nonstop cipher anywhere. Ethan Smith of Graphite makes a sharper constituent successful the study astir really users are folding those aforesaid prompting habits backmost into Google itself, pinch mean hunt magnitude climbing steadily since AI Mode launched. People aren’t abandoning hunt boxes. They’re conscionable typing longer sentences into them.

The Number That Should Worry The AI Skeptics

Now for the half of the study that punctures the different side’s confidence. Average monthly web visits crossed generative AI platforms deed 9.5 cardinal betwixt June 2025 and May 2026, up 70% twelvemonth complete year. App downloads worldwide climbed to 2.7 billion, up 134%. Half of each generative AI users are now 35 aliases older, compared to 61% nether 34 conscionable 2 years ago, which is the clearest awesome I’ve seen that this isn’t a Gen Z fad moving retired its inclination cycle. Michael Horrocks of Miro puts it plainly successful the report: “Growth concentrated successful younger demographics tin slice pinch trends; maturation spreading into older generations is often what durable, mainstream take looks like.”

Meta AI’s ain disclosed numbers backmost that up from a wholly different angle. Publicly reported monthly progressive users went from 384 cardinal successful September 2024 to 1.2 cardinal by March 2026, much than tripling successful 18 months, wholly by riding wrong Instagram, Facebook, WhatsApp, and Messenger alternatively than arsenic a standalone destination anyone had to activity out. And ChatGPT advertisement penetration successful the U.S. jumped from 14% of desktop chats successful May 2026 to 26% conscionable 1 period later. Whatever you deliberation astir the maturity of AI search, advertisers intelligibly don’t deliberation it’s a toy.

My publication is that some camps are pattern-matching disconnected the portion of the information that confirms what they already believed, and some are ignoring the half that complicates it. AI hunt hasn’t replaced anything. It has stacked a new, fast-growing, unevenly distributed furniture connected apical of a hunt ecosystem that was already there, and the practitioners who will triumph the adjacent 2 years are the ones measuring the stack alternatively of arguing astir which furniture matters more.

Why The Disconnect Between Citations And Clicks Matters More Than Either Camp Realizes

The astir useful floor plan successful the full deck, and the 1 I deliberation gets underappreciated successful the LinkedIn debate, comes from Aleyda Solis of Orainti. She points retired that 65% of the URLs ChatGPT cites beryllium 2 aliases 3 folders deep, the pages doing the existent evidentiary activity down an AI answer. But 58.8% of the referral postulation that AI sends backmost to sites lands connected the homepage, not the cited page astatine all. Cited pages and clicked pages are almost wholly different populations of URLs.

That azygous information constituent should reorganize really agencies study AI capacity to clients. If you’re only search whether your heavy merchandise aliases blog contented gets cited, you’re missing the truth that the humans who really click done are landing location other wholly and request their ain conversion path. Rand makes a related constituent successful the study itself, comparing this to really 20th-century advertisers proved advertisement and power walk worked by measuring assistance successful shop visits alternatively than counting who glanced astatine the sign. The system has changed. The subject of measuring downstream behaviour alternatively of aboveground impressions hasn’t.

3 Things To Change In Your Strategy This Week

1. Split your reporting into 2 abstracted metrics. Track citation complaint and citation files extent arsenic 1 cardinal capacity parameter that measures whether AI trusts your contented capable to usage it arsenic evidence. Track referral landing pages and downstream conversion arsenic a wholly abstracted KPI that measures what really happens erstwhile a quality clicks through. Conflating the 2 successful a azygous dashboard is really brands miss some problems astatine once.

2. Stop treating “AI visibility” arsenic a azygous category. Similarweb’s marque visibility scale shows really category-specific this already is. CeraVe leads beauty astatine an scale of 100 while NYX Cosmetics sits astatine 19 successful the aforesaid category. Kevin Indig of Growth Memo argues successful the study that stock of sound is the metric that matters present because it’s a relative comparison successful a stochastic system, not an absolute score. Pull your ain category’s leaderboard earlier you presume you’re winning aliases losing.

3. Match your contented to the platform’s existent audience, not the platform’s wide size. The affinity information shows ChatGPT skews toward mundane consumers researching restaurants, health, and fashion; Claude users are 25 times much apt than the mean searcher to sojourn assemblage sites and skew heavy toward professionals and students, and Gemini users over-index connected graphics, security, and hardware content. A azygous portion of “AI-optimized” contented aimed astatine each 3 is aimed astatine nary of them.

The Bottom Line

I don’t deliberation the AI zealots are incorrect that thing structural is shifting. Nine and a half cardinal monthly visits and a fivefold jump successful citation rates successful nether a twelvemonth are not a rounding error. But I besides don’t deliberation the skeptics are incorrect that astir of the industry’s AI panic is moving good up of the existent postulation numbers, fixed that Google still commands 5 times the assemblage of each AI chatbot mixed and 95 retired of each 100 ChatGPT users ne'er near Google successful the first place. Rand’s station nailed the discomfort because the information refuses to fto either broadside support its communicative simple. The brands that will really use from this study aren’t the ones picking a side. They’re the ones pulling the citation and referral numbers for their ain class this 4th and building a strategy astir what the information says alternatively than what the statement connected LinkedIn says.

More Resources: