The Gap Between AI Search and Local Leads Is Smaller Than You Think

2026-09-15 — marketing audit SME India

I watched someone click on a plumber's ad. Then click right past everything and open ChatGPT instead. The ad click? Already paid for. The lead that walked somewhere else? That one stings.

This is the real problem with AI search.

It's not that AI search is invisible to local businesses. It's that we're looking at it wrong. According to data from CallRail, their customers are seeing somewhere between 1% and 2% of their phone calls coming directly from AI citations—calls that click through from a ChatGPT response, or a Gemini result, landing directly on your site. That's small. But here's the thing: that figure has roughly doubled since January. From nothing to small to something is the arc that matters right now.

The trouble is measuring it.

Two signals actually work. One is the direct click—when someone reads an AI response, clicks the citation link, and calls you from there. That looks exactly like any other referral in your logs. The other is just asking. A customer says, "ChatGPT told me to call you." Both have exploded by over 100% since the start of the year. Both still feel marginal. That's because they started from near-zero.

What doesn't work yet is the obvious thing.

Google has advertising tools. OpenAI doesn't. Not in any way that talks back to your dashboard. The intermediary platforms that let you track ad spend to actual revenue—the ones that made Google Ads and Facebook Ads manageable—just don't exist for ChatGPT ads. Actually, that's not quite right. They exist in some form. They're just not developed enough to trust, and from what we hear, OpenAI isn't particularly eager to help marketers understand the connection either. So most of us are stuck in a blind spot, watching citations arrive in GA4 without much idea where they'll land.

But there's something strange about the timing that matters more.

Half of web traffic hits a local business outside normal business hours. With AI search? That number jumps to nearly two-thirds. People don't start their day asking ChatGPT about plumbers at 8 a.m. They start in the afternoon and keep browsing until well past midnight. When they finally call, your business is closed. Your voicemail is full. Or they call your competitor because an AI might show three options instead of one. You paid for the click. You paid for the ad. You don't answer at 2 a.m., and now you're watching a lead walk out the door, no matter what your attribution tools say.

The real leverage isn't in tracking. It's in being available when the lead actually lands.

This is why call handling outside business hours isn't optional anymore. It's not inspirational. It's just math. More of this business happens when you're not there. Something like 65% of consumers called a business in the past month. They didn't schedule it. They called when they needed something. If you're running a service business—HVAC, dental, legal, whatever—and you're still thinking of the phone as an after-hours liability instead of a revenue channel, you're already losing.

One more technical thing that actually works: server-side phone number swaps. If you tell a server to show a different number to ChatGPT's crawler than it shows to Google's crawler, the AI picks up your tracking number correctly. It has to happen on the server side though. The crawlers that AI agents use don't run JavaScript. They don't execute scripts. They just grab what's there. If you're trying to do this in the browser, it won't work. Most people get this wrong. So do we, sometimes.

The bigger picture is this. You don't have to solve AI search attribution perfectly. You don't have to understand every click. You just have to not miss the calls when they arrive, no matter what time it is or which tool sent them.

Source: "CallRail's AI search engine attribution automatically tags whether an inbound call came from ChatGPT, Perplexity, Gemini, or Claude, not just "AI."" — Search Engine Journal