AI Search vs Google Traffic: Is AI Taking Your Clicks?

If your Google traffic is sliding and someone told you AI search is eating it, the honest answer is: partly true, and also the wrong thing to panic about. AI answers really are taking some clicks — but they're taking the clicks that were worth the least, and the ones you keep are worth more. The instinct to chase the lost sessions back is the mistake. Let me walk through what's actually happening, what you can see in your analytics, and what to measure instead.

The zero-click shift, honestly

Here's the mechanic. Someone used to type a question into Google, get ten blue links, click one of yours, read the answer, and leave. That was a session in your dashboard. Now they ask ChatGPT or read Google's AI Overview, get the answer in the response, and never click. The session that used to exist simply doesn't. That's the zero-click shift, and no amount of optimizing your page brings that particular click back — because the user's need was met before they ever reached a link.

But notice which clicks vanish. They're the informational ones — "how long does biltong last," "what's the difference between drip and soaker hose," "is a space heater safe for a nursery." The AI resolves those fully, so the click evaporates. That's the honest bad news, and pretending otherwise helps no one.

Here's the part worth sitting with, though: those were rarely your best clicks. The informational visitor read your answer, got what they came for, and left — no cart, no booking, no email captured. They inflated your session count and did little else. When the AI absorbs that traffic, the graph looks like a loss, but the business barely feels it. What you're watching leave is mostly the shallow end of the pool.

Why raw sessions can fall while your influence rises

Now the part that gets missed. Your session count and your actual influence are two different numbers, and in the AI era they've come unglued from each other.

When an AI answers "what's a good quiet space heater for a nursery" and names your product, that recommendation reached a buyer at the exact moment of decision — and it produced zero sessions. Your traffic graph shows nothing. Your influence went up. If you were only watching sessions, you'd conclude AI search hurt you, when it just recommended you to someone ready to buy.

This is why the traffic frame is quietly broken. The thing that matters most now — being the brand named inside the answer — is close to invisible in a tool built to count clicks. Raw sessions can fall in the same quarter your brand gets recommended more, because the recommendation increasingly happens where your analytics can't see it. For the deeper version of how those decisions get made, see how AI picks which brand to recommend and the full answer engine optimization playbook.

Reading AI referral traffic in your analytics

You can see part of it, and it's worth seeing — as long as you remember it's a floor, not the whole picture.

Some AI tools do send clicks, and those arrive as referral traffic. In GA4, open Traffic acquisition and look at the session source for the AI hostnames: chatgpt.com, perplexity.ai, gemini.google.com, and copilot-style referrers. Segment those out and watch the trend over months, not days. That's your visible, clickthrough AI traffic — the times an answer cited you and the user clicked to verify or go deeper.

Two honest caveats. First, Google's AI Overviews mostly don't announce themselves — those clicks still land tagged as ordinary Google organic, so you usually can't cleanly separate AI-driven Google visits from classic ones. Second, and more important: this referral number is always an undercount. It only captures the answers where someone clicked. Every answer that named you and satisfied the user without a click — the majority — leaves no referral trace at all. So treat rising AI referrals as a good sign, but never mistake that visible slice for the size of your real influence.

Informational clicks you lose, transactional clicks you keep

The single most useful distinction here is query intent, because it predicts which clicks you lose and which you keep.

Informational queries — questions with a factual answer — are the ones AI resolves in the response. Those clicks you largely lose, and no page edit reverses that. Transactional and high-intent queries — someone ready to buy, book, compare specific options, or sign up — still send real clicks, because at that point the person has to reach your actual page to act. The AI can recommend the restaurant; the diner still clicks through to see the menu and reserve. The AI can name the tool; the buyer still lands on your pricing page.

So the healthy pattern in your dashboard is: informational sessions drifting down, conversion-intent sessions holding steady or rising, and revenue flat or up. If that's what you see, your traffic didn't get worse — it got more qualified. You shed the skimmers and kept the buyers.

The practical move is to stop reading your traffic as one lump number and split it by intent. Look at which landing pages and queries are shedding sessions. If it's your blog posts, guides, and "what is" explainers, that's the informational bleed, and it's expected. If it's your product, pricing, booking, or comparison pages, that's a real problem worth chasing. Most of the time it's the former, and the alarm quietly deflates once you see where the drop actually lives. Understanding which questions fall on which side is what conversational search and AI intent is really about, and it's the cleaner lens than the old keyword split — more on that in AEO vs SEO.

Why chasing traffic is the wrong frame

Put it together and the strategy writes itself. The recommendation now happens inside the answer. That's the whole game. A buyer asks a machine what to get, the machine names two or three options, and the decision is largely made in that moment — often with no click for you to count.

If you optimize for traffic, you'll pour effort into clawing back informational clicks that never converted, and you'll be measuring the one thing that's structurally shrinking. If you optimize for being named in the answer, you're playing where the decision actually gets made. The right question isn't "how do I get my sessions back," it's "when a customer asks AI what to buy in my category, am I the one it recommends — and if not, why not?"

That reframe is also freeing, because it's measurable. Keep a fixed list of the real questions your customers ask, run them in ChatGPT, Perplexity, and Google's AI answers, and record how often you get named and whether the answer is accurate. That coverage rate — paired with your conversion-intent traffic and revenue — is the honest scoreboard. Watch it move instead of watching raw sessions. The full method is in track your AI visibility, and the work of actually earning those citations is in get cited by AI search.

None of this means Google traffic is dead or that you should stop caring about clicks. High-intent clicks still convert and still matter. It means the number on your traffic dashboard is no longer the number that runs your business. The brands that come out ahead are the ones who noticed early that the click became optional and the recommendation became everything. If you want that citation layer built across your whole site instead of audited one page at a time, that's the problem I built RunOctopus to solve.