It's Not ChatGPT Taking Your Local Leads. It's Google.

Somebody told you AI ate search, and the story fit your quarter. Your rankings held, your map pack position held, and the phone still went quiet, so ChatGPT became the obvious suspect. It is the wrong suspect. ChatGPT did not take your local leads; Google rearranged its own results page and took them itself, and the first move is not a new AI strategy, it is thirty minutes spent finding out where your leads actually come from.

Your Scapegoat Sent One Client 2% of Google's Volume

Sterling Sky's "The State of Local SEO in 2026," published June 26, 2026, tracked a large multi-location client whose ChatGPT traffic grew from 0.1% of Google's volume to 2%. That growth rate sounds dramatic until you sit with the second half of Joy Hawkins's finding: even after that jump, ChatGPT was still sending only 22% of what Bing sent. That is Bing, the search engine nobody in your industry has held a strategy meeting about in years.

I have spent 17 years in search, 15 of them in SEO and 10 of them running Google Ads, and this is not the first time I have watched an industry blame a new arrival for a shift the incumbent made on purpose. The difference this time is that the numbers are public and recent, so you do not have to take my word for any of it. Both halves of that experience matter here, because the change you are feeling shows up on the organic side and the paid side of the same results page.

Translate that into your calendar. If ChatGPT is a two-percent channel and Google is the hundred, then a month spent rewriting your site for AI assistants while the Google side goes unaudited is a month spent moving the smallest lever in the building. That is the decision the number should change.

The buyer-intent split says the same thing from a different angle. First Page Sage's "Google vs ChatGPT Market Share: 2026 Report," published August 11, 2026 by Evan Bailyn, puts transactional queries at roughly 90% Google and 5% ChatGPT, and navigational queries at roughly 93% Google and 3% ChatGPT. Transactional and navigational are the two that matter to a local service business: someone ready to buy, and someone looking up a specific company by name. Those are your booked jobs.

Source: First Page Sage, "Google vs ChatGPT Market Share: 2026 Report," Evan Bailyn, August 11, 2026.

Read the top-ranking articles on this exact question and you will find plenty of alarm about AI and very little about where your leads went. Pair those two published datasets with what I see inside client accounts and the picture stops being mysterious: AI referral traffic is a line item you need a magnifying glass to find, while the Google surfaces those same clients have depended on for years quietly changed shape underneath them. Both things are true at once, and only one of them is costing you money this month.

Blaming ChatGPT Is Easier Than Auditing Google

Misattribution is not stupidity. It is what happens when the loud story arrives before the boring data does. Every conference talk, every LinkedIn post, and every cold email in your inbox this year has been about AI, so when revenue dips, the brain reaches for the explanation it has heard forty times rather than the one it has heard zero times.

There is a second reason, and it is mechanical. AI referrals do not show up as AI referrals in your analytics. They land in the Direct bucket, which means the owner looking at a report sees no ChatGPT line at all, concludes the tracking is broken, and fills the silence with the assumption that AI is taking everything. I wrote the walkthrough for that specific problem in how AI referrals get filed under Direct traffic in GA4, and it is worth an hour before you accept any conclusion about your channel mix.

The third reason is that AI visibility genuinely is hard, which makes it feel important. Search Engine Land reported in January 2026 that AI assistants recommend only 1% to 11% of the locations available in a given market, and that AI local visibility can be up to 30 times harder than ranking in Google. That is a real finding and I am not waving it off. Difficulty and revenue are separate questions though, and confusing them is how a business ends up with a six-month AI project attached to a two-percent channel.

Search Engine Journal ran a piece by Matt G. Southern on August 17, 2026 under the headline "AI Assistants Are Choosing Local Businesses For Your Customers." The headline is accurate and the trend is real. A trend being real does not make it the thing that emptied your calendar last quarter.

Google Kept Your Ranking, Took Your Screen, and Sold It Back

Picture an HVAC company in Sacramento. It has the same three vans, the same owner answering the phone, and the same position two in the map pack it has held since 2023. Summer arrives, the rank tracker stays green, and bookings come in noticeably lighter than the previous summer. Nothing about that company's ranking changed. What changed is what a phone screen shows a homeowner at 2pm on a 104-degree day.

Sterling Sky's 2026 report, using data from Places Scout, found that AI local packs featured 5,943 unique businesses compared with 18,330 in regular 3-packs, roughly 32% as many. Across 322 markets, 88% had fewer unique businesses appearing in AI local packs than in traditional ones. The AI version typically shows one or two businesses instead of three, and it carries no call button.

Source: Places Scout data via Sterling Sky, "The State of Local SEO in 2026," Joy Hawkins, June 26, 2026.

Sit with the call button line, because it is the one that touches your bank account directly. A homeowner with a dead air conditioner does not want to read your website. They want to press a green button. Sterling Sky's report also cites Jepto data covering 179 Google Business Profiles across 34 US law firms over two years, showing clicks-to-call from those profiles in a steady decline, and the decline is specific to mobile, which is exactly where the call button lives. No single percentage was published for that decline, so I am not going to hand you one. The direction is the point, and the direction is down on the device your customers use when the problem is urgent.

For the full breakdown of why three names became one and what earns the remaining slot, read why ranking in the map pack no longer guarantees Google shows you.

Shrinking the free real estate is only half of what happened. Sterling Sky's ranking reports show local pack ads on mobile appearing in about 1% of tracked queries in early 2025 and about 22% by December 2025. Local Services Ads went from roughly 11% to roughly 31% over the same stretch. That happened inside one year. On a phone, where the visible area above the fold is about the size of a postcard, twenty extra percentage points of paid units is not a cosmetic change, it is the difference between being seen first and being scrolled past.

Hawkins summed up the free half in one plain sentence: "AI local packs are only surfacing about 32% as many businesses as traditional 3-packs." Two independent things moved in the same direction inside the same company's product: fewer free slots, more paid ones. Neither of those changes was made by OpenAI.

The month-by-month response to a shrinking organic surface is its own subject and I have written it out in detail in what to do when your ranking holds but your phone stops ringing. What belongs here is the attribution: the squeeze is Google's, the timeline is 2025 into 2026, and the evidence is Google's own results pages.

Five Moves That Point at the Right Company

None of this requires a technical background. It requires an afternoon and a willingness to look at your own data before accepting anyone's narrative, including mine. Do them in order, because step one changes how seriously you take the other four.

Each of those steps has a concrete version. For step one, the GA4 path is Reports, then Acquisition, then Traffic acquisition, with the date range set to the last 12 months. For step two, your money queries are the three searches that produce booked jobs, which means your service plus your city, not your brand name, because the people typing your brand name were already on their way to you. For step three, rebuilding the call path means a tappable phone number in the first screen of your homepage and your Google Business Profile, running on a tracking number so those calls get counted as leads instead of vanishing into the office handset. For step four, the criterion is one comparison: your cost per booked job from ads against your margin on that job. If the margin covers the cost with room left, the paid space is worth buying. If it does not, you work around it and you know exactly why.

Step five is the one owners skip, and it is the cheapest. I have worked with 300-plus businesses across the USA, Canada, the UK, Singapore, Australia and New Zealand, in medical, ecommerce, local services, fintech, education and real estate. The pattern that separates the calm operators from the panicked ones is not budget or industry. It is whether someone in the room can name where the leads came from without guessing.

When you can name the source, the work becomes obvious. A London ADHD clinic I worked with got booked solid for three straight months, to the point where they hired more specialists and outsourced the overflow. That did not come from chasing a channel that was in the news. It came from being the obvious choice on the surface where the patients were already searching, and from making it effortless to act once they found it.

Measure Leads, Not Weather Reports

Position two in a pack that now shows one business is not position two. It is an entry on a list nobody sees. Rankings have quietly become a weather report: interesting, widely discussed, and disconnected from whether anyone walked through your door.

Distrust three metrics in particular. Average ranking position stays flat while the number of visible slots collapses underneath it. Impressions can rise while clicks fall, since an impression counts even when your listing sits under a wall of ads. And any AI visibility score sold as a headline number prices a small win as a large one.

My own site taught me the plainest version of this. I rebuilt it from a PageSpeed score of 59 to 97, and the thing that made the work worth doing was not the number itself, it was that a faster page changed what people did once they arrived. Metrics earn their place by predicting behavior. Everything else is decoration.

Frequently Asked Questions

Should I ignore ChatGPT completely then?

No, but size the effort to the channel. Sterling Sky's 2026 report found ChatGPT at about 2% of Google's volume on a large multi-location client, and First Page Sage put Google at roughly 90% of transactional searches in August 2026. Treat AI visibility as a long-term position you build steadily, not an emergency that justifies pausing the work on Google. If someone is selling you an urgent AI package, ask them what share of your leads it is meant to protect.

How do I find out where my leads actually came from?

Open your analytics and look at traffic by source over the last 12 months, then compare Google against every AI tool combined. Check the Direct bucket carefully, because AI referrals get filed there and make the AI share look like zero. Then match those sources against your actual booked jobs, not just website visits. Two numbers, one afternoon, and most of the guessing ends.

Do I have to start running ads to get local leads now?

Not automatically, but you should decide instead of drift. Sterling Sky's ranking reports show local pack ads on mobile going from about 1% of tracked queries in early 2025 to about 22% by December 2025, with Local Services Ads moving from about 11% to about 31%, so the free space above your listing genuinely shrank. Whether paid makes sense depends on your cost per booked job and your margin on that job, which are numbers you already have. Work those out first, then buy the space or work around it deliberately.

The uncomfortable arithmetic in all of this is that Google can shrink your free visibility and sell you the space it took, and nothing about that violates any rule. Your ranking report will keep saying you are fine while the screen tells your customer a different story. If your calls are down and your position is not, you are not looking at an AI problem, you are looking at a pricing decision made by the company that has always sent you most of your business.

If you want a second pair of eyes on your own split before you spend another quarter on the wrong fix, book a call with me and bring your analytics. Thirty minutes is enough to tell you which company actually took your leads.