AI Search Is Eating Your Traffic. Here’s How to Measure It.

Your organic traffic is flat or down. Your Google rankings haven’t moved. Your SEO agency says everything looks fine. And yet, somehow, fewer people are calling. At some point you have to consider that the problem isn’t your rankings. It’s that a growing slice of your potential customers never visited a search results page at all. They asked ChatGPT, got an answer, and moved on. You weren’t in the answer. Nobody told you.

measure AI search traffic customer impact

AI-generated search responses from ChatGPT, Perplexity, Google’s AI Overviews, and Gemini are absorbing queries that used to produce clicks. The traffic didn’t disappear. It just stopped being traffic. This post is about how to measure that gap, what it’s actually costing you, and what you can realistically do about it without hiring a team of people to stare at dashboards all day.

01Your Analytics Are Lying by Omission

Google Analytics shows you what happened after someone arrived on your site. It has nothing to say about everyone who got their question answered upstream and never arrived at all. That’s the core problem. Your session counts, your bounce rates, your conversion funnels: all of that is measuring a population that’s quietly shrinking while the tool shows you a stable line and you assume things are fine.

The first thing to check is Google Search Console. Specifically: impressions versus clicks, over time, broken down by query type. If impressions are holding steady or rising while clicks are falling, that’s not an SEO problem. That’s AI Overviews absorbing the answer and returning zero clicks to anyone. You showed up to the party; the bouncer just stopped letting people through.

Pull the data for the past 12 months. Filter to informational queries: anything that starts with “how,” “what,” “why,” “best,” “is it.” Compare click-through rates from 18 months ago to now. If the CTR on those queries dropped by more than 20–30% while rankings stayed put, you’re watching the AI redirect happen in real time. Reports from early AI Overview rollouts suggested the feature appeared across the majority of measured search categories within months. CTR on affected queries dropped measurably across the board. The clicks went somewhere else. That somewhere is a text box.

02The Referral Source Your Analytics Probably Isn’t Catching

When someone clicks a link inside a ChatGPT response or a Perplexity answer, that visit often arrives in your analytics as direct traffic: no referrer, no source, no campaign. It looks identical to someone who typed your URL directly into their browser. A warm lead. Except it isn’t. It’s a stranger who found you through an AI tool, and your system has no way to tell the difference.

Perplexity has started sending a referrer header in some cases (perplexity.ai), so filter your direct and referral traffic for that string and see if anything shows up. ChatGPT does not consistently pass referrer data, which is its own kind of gift to anyone trying to understand what’s happening. You may be getting more AI-referred traffic than you think, or less, and your current setup cannot tell you which.

The practical fix: create UTM-tagged versions of your most important pages and URLs, and start submitting them to content that AI tools tend to cite: industry directories, authoritative third-party articles, Wikipedia-adjacent sources. If an AI tool picks up your tagged URL and someone clicks through, the UTM parameter survives. You’ll see it. It’s not a complete solution, but it’s a measurement you can actually run.

03What “Being Cited by AI” Actually Means for Your Business

There’s a version of this story where AI search is good for your business: the AI recommends you by name, the user searches for you directly, and you get a high-intent visit from someone who’s already been told you’re the answer. That happens. It’s real. It is also roughly as likely as your accountant spontaneously complimenting your filing organization.

The more common scenario is that the AI synthesizes an answer from several sources, names none of them, and the user closes the tab satisfied. Your content contributed to the answer. You received nothing. Think of it as donating your expertise to a very confident summarizer that doesn’t do citations at dinner parties. It absorbed your best explanation of how brake calipers work, served it to 4,000 people last Tuesday, and filed no paperwork.

The way to check whether you’re in the “named” camp or the “donated” camp is manual and tedious, which is appropriate given everything else about this situation. Take your 10 highest-traffic informational pages: the ones where you’re genuinely trying to answer a question your customers have. Paste those questions into ChatGPT, Perplexity, and Google’s AI Overview. See if your brand appears. See if your content is paraphrased without credit. Do this monthly. It’s not automated, it doesn’t scale elegantly, and it is currently your best option.

04The Number That Tells You Whether to Care

Not every business is equally exposed to AI search disruption. The question is how much of your current customer acquisition depends on people discovering you through informational search queries. If someone Googles “what causes that clunking noise when I brake” and you’re an auto shop that answered that question well, you were getting that traffic. Now you might not be. If your business runs entirely on referrals and repeat customers, you can put this article down and go do something more useful with your afternoon.

The diagnostic: in Google Analytics, go to Acquisition > Traffic Acquisition and look at what percentage of your new users come from organic search. Then, in Search Console, look at what percentage of your organic clicks come from informational queries versus navigational ones (people searching your brand name) or transactional ones (people looking to buy something specific). If more than 40% of your organic traffic is informational, you’re exposed. The AI is eating the top of your funnel and you’ll feel it over the next 12–18 months if you haven’t already.

Here’s the uncomfortable math: if informational queries represent 40% of your organic traffic and AI Overviews cut CTR on those queries by even half, you’re looking at a 20% reduction in overall organic reach without a single ranking change. Your Search Console will show impressions holding. Your agency will say nothing is wrong. The phone will keep getting a little quieter each month until the pattern is too obvious to explain away. Most businesses will notice around month eight and describe it as a “slow period.”

05Three Things You Can Actually Do About It

Tactical options, in order of effort and impact:

  • Own your brand queries. Navigational traffic, people searching your name, is largely AI-resistant. The AI doesn’t intercept “[Your Business Name] pricing” or “[Your Business Name] reviews” because there’s no generic answer to give. Invest in getting mentioned by name in third-party content, directories, and review platforms. Make “find you specifically” the action, not “find someone who does this.”
  • Go transactional earlier. Informational content that exists to build awareness is the most exposed category. Transactional content, comparison pages, pricing pages, pages that help someone decide, gets less AI Overview treatment because the AI can’t confidently close the deal for you. Push your mid-funnel content to do more closing and less educating.
  • Build a list. Email, SMS, a loyalty program, a recurring touchpoint of any kind. AI search disrupts discovery. It does not disrupt a relationship you’ve already built. If someone is on your list, they don’t need to discover you again. The businesses that will feel this least are the ones that converted discovery into something durable before the discovery channel got hollowed out.

06Track It or Accept That You’re Guessing

Set up a monthly 20-minute review. Pull your Search Console CTR data for informational queries. Check your referral sources for any AI platform strings. Run your top 10 question-based pages through ChatGPT and Perplexity manually. Note what’s changing. That’s the whole system.

You’re not going to automate this perfectly. The data infrastructure for AI search attribution is still being built, and most of it doesn’t exist yet at a price point that makes sense for a small or mid-sized business. The tools that do exist tend to be enterprise-grade, built for teams with dedicated analysts and budgets that don’t flinch at four-figure monthly subscriptions. That is not most of the businesses actually dealing with this problem. What you can do is notice the trend before it becomes a problem you’re explaining to yourself in hindsight, which is a lower bar than it sounds.

The businesses that figure this out are not the ones waiting for a perfect dashboard. They’re the ones who accepted that the measurement is ugly and imperfect right now, started tracking what they could, and adjusted when the numbers moved. You now have enough to start. The gap between knowing this and doing something about it is, historically, about six months of not quite getting around to it. You’re going to finish reading this and open a new tab for something unrelated. That’s fine. Just make sure the Search Console tab is the one after it.