AI Search Took the Click. Multi-Location Brands Should Stop Grading Search by Traffic.

If your search report still opens with website sessions, it's grading the wrong test. For multi-location brands, I'd drop traffic as the headline search metric this quarter and replace it with actions counted store by store, like calls, direction requests, and booked appointments. I'd also shift effort away from publishing more general articles and toward making every location's listing and page impossible for an AI answer to get wrong.

I know how that sounds from someone whose job includes demand generation, writing on a blog. Traffic is the number most of us were trained to watch. It's also the number AI search is wearing down without much fanfare, and brands that keep leading with it will end up cutting the wrong programs.

The 8% problem

The clearest data I've seen comes from Pew Research Center, which studied the actual browsing of 900 U.S. adults. When a Google results page carried an AI summary, people clicked a traditional result in 8% of visits. Without one, they clicked in 15%. They clicked a link inside the summary itself in just 1% of visits.

That isn't a fringe experience. Pew's broader report on the same panel found 58% of respondents ran at least one search in a single month that produced an AI summary.

Google is now building its ad business around the answer itself. In May, Google said it is testing two ad formats that sit inside AI Mode answers, Conversational Discovery ads and Highlighted Answers, and plans to bring a chat agent called Business Agent for Leads into Search ads in the coming months. These are tests, and none of them is something you can buy in every market today. Still, when the platform starts selling placement inside the answer, the answer has become the storefront.

Back in February 2024, Gartner predicted traditional search engine volume would drop 25% by 2026. I can't tell you whether that exact figure came true, and I wouldn't plan a budget around it. I'd plan around the part nobody disputes, which is that more of the decision now happens before anyone reaches your site.

The customer still needs a place to go, though. Someone asking an AI tool for a roofer, a dentist, or a dinner spot still has to call, drive to, or book with one specific location. The click moved. The customer didn't.

One store's wrong hours never show up in a brand report

A single-site business can live with a fuzzy search metric. A brand with 80 locations can't, because a blended traffic number hides which stores the answers send people to and which ones they skip. Location-level marketing needs location-level measurement, and AI search raises the stakes on that.

When an AI answer recommends a business, it leans on structured facts like the listing, the hours, the services, the reviews, and the location page. If a franchisee changed holiday hours on the door but not on the listing, the answer is wrong for that store and right for the other 79. Your sessions report will never show you that. A call count by location will, eventually, as a dip nobody can explain.

That's why our SEO and GEO work starts with location data rather than content volume. Content still matters. For most of the brands we talk to, it just isn't the first bottleneck.

Where 1,470 of 1,844 leads came from

The best evidence I can offer is a portfolio we manage for Horsepower Brands, a multi-brand home services franchisor. Across 21 franchise locations, the program produced 1,844 verified leads, and 1,470 of those 1,844 came from GBP PowerRank call leads. For that portfolio, the listing was the lead engine.

Consistency is the other part of the story. The listing work runs through GBP PowerRank, and in that program 14 of 15 GBP locations improved rankings through 25 to 40 automated actions per month, covering posts, Q&As, videos, and review responses. It's unglamorous maintenance, done for every location every month, and it's the kind of work an AI answer depends on.

The full numbers are in the Horsepower Brands case study, including which figures we measured and which we estimated.

Try the 10-prompt location check this week

You don't need to buy anything to run this, and it takes about two hours.

1. Pick five locations. Use your two strongest by revenue, your two weakest, and your newest.
2. For each one, write 10 questions a real customer would ask. Mix short searches like "emergency plumber near me" with full sentences like "who can fix a leaking water heater in [neighborhood] this Saturday." Keep the long ones in. Pew found 53% of searches with 10 or more words produced an AI summary, compared with 8% of one- or two-word searches.
3. Run every question in Google, noting whether an AI Overview or AI Mode answer appears, and in one AI assistant your customers actually use. Name the city or neighborhood in the prompt so the results reflect that trade area.
4. Score each answer three ways. Does that location appear at all, are the facts it states correct (hours, phone, services, address), and which competitor gets named when you don't?
5. Pull the last 90 days of calls, direction requests, and website clicks from each location's Google Business Profile, and put them next to the scores.

Then use this table to decide what each location needs.

What you found	What it means	What to do next
The location appears in fewer than 3 of 10 answers, or any answer states a wrong fact	AI answers can't find or trust that store's data	Fix the listing and location page before you spend another dollar on content or paid search in that market
It appears in 3 to 6 answers and the facts are correct	Visible, but not the first choice	Compare reviews and listed services against the competitor that keeps getting named
It appears in 7 or more answers, but calls and direction requests are flat	Visibility isn't the problem	Look at reviews, hours, and the offer

When sessions should stay at the top of your report

Traffic still deserves top billing when the transaction happens on your site. A restaurant brand that lives on online ordering, a retailer with real ecommerce, or a healthcare group whose online booking flow is the conversion should keep sessions near the top of the report, because a lost click there is a lost order.

Content deserves a defense too. For considered purchases, a well-built guide is often what an AI answer cites, and being the cited source has value I can't fully count. If your listings are already clean everywhere, more content may be exactly the right next dollar. My argument is about order of operations. Fix the facts the answer pulls from first, then add to them.

The blind spots in all of this

Nobody can see every AI answer every customer gets. Answers shift by day, by device, and by how the question is worded, which is why the check above is a sample. A call from a listing isn't a sale, either. Some are existing customers, some are wrong numbers, and some never book. In the Horsepower Brands work, pipeline value is estimated using industry benchmark averages, and we label it that way.

Where a client opts into closed-loop attribution, we can follow a lead through to revenue. Where they don't, we say where the trail stops instead of filling the gap with a guess. A dashboard full of location-level actions is a better scoreboard than sessions, but it still isn't a strategy. Someone has to decide what to do about the store that never shows up.

Start with one store tonight

Pick your weakest location and ask an AI assistant three questions a customer in that town would ask. Write down whether you appear and whether the hours are right. That one page of notes will tell you more about your search program than last month's traffic report.

Sources

• Pew Research Center, "Google users are less likely to click on links when an AI summary appears in the results," July 22, 2025.
• Pew Research Center, "What Web Browsing Data Tells Us About How AI Appears Online," May 23, 2025.
• Google, "A new generation of ads for the AI era of Search," May 20, 2026.
• Gartner, "Gartner Predicts Search Engine Volume Will Drop 25% by 2026, Due to AI Chatbots and Other Virtual Agents," February 19, 2024.