Pipelineos

Here's a maintenance tip I'd run today if I owned a service business: open your Google Business Profile and look at your service area.

I see owners list 25 cities. Sacramento. Roseville. Rocklin. Lincoln. Loomis. Auburn. Folsom. Granite Bay. Citrus Heights. Orangevale. Fair Oaks. Carmichael. Elk Grove. Galt. Wilton. And on it goes. The thinking is reasonable. More cities, more leads.

What actually happens is the opposite. Google looks at that list and tries to figure out where you really operate. If your reviews mention three cities, your photos are tagged in two, and your website talks about all 25 with the same paragraph and the city name swapped in, Google does not know what to trust. So it trusts none of it.

Here is what I would do instead. Pick the three cities where you do the most revenue. Not the most jobs. The most revenue. Build a real page for each one. Mention the neighborhoods you actually work in. The HOAs you have permits with. The soil conditions, the building codes, the permit office, the drive time from your shop. Add photos from jobs in that city. Ask the customers in those cities for reviews that mention the city by name.

Then drop the other 22 from your service area. I know that feels like leaving money on the table. It isn't. You were not ranking in those cities anyway. You were diluting the three that could actually pay your mortgage.

Depth beats breadth in local. It always has. The March core update just made it more expensive to ignore.

Speed-to-lead is the unsexy metric that beats almost every other marketing investment you can make this quarter.

Here's the scene. A homeowner fills out your contact form at 2:47pm on a Tuesday. She's standing in her kitchen looking at a slow drain. She just filled out three forms. Yours, and two of your competitors.

Whoever calls her back first usually wins. Not whoever has the best website. Not whoever ranks highest. Whoever picks up the phone.

The research on this has been consistent for over a decade. Lead response within five minutes versus thirty minutes changes contact rates by a factor of around 100. Within an hour versus a day, it falls off a cliff.

Most local service businesses I look at respond in two to twelve hours. Some don't respond at all because the form notification goes to an email nobody checks, or it gets caught in spam, or the owner sees it after dinner and figures he'll call in the morning. By morning she's booked.

So before you spend another dollar on ads, do this. Submit a test lead through your own form right now. Time how long it takes for someone in your business to call that number back. If it's more than fifteen minutes during business hours, you have a leak that no amount of traffic will fix.

Fix the response time first. Then buy more clicks.

Tracking is the part of marketing nobody brags about and almost nobody does well.

It's also where the biggest revenue gains hide.

Here's what I mean. An owner spends $4,000 a month on Google Ads. The agency reports 87 conversions. Looks fine. He keeps paying.

Then we open the account. "Conversions" includes every form view, every click on the phone number on desktop, every 10-second page visit. Not actual calls. Not actual booked jobs. The 87 number is mostly noise.

We rebuild it. One conversion action for phone calls over 60 seconds from the ad. One for form submissions that hit the CRM. One for booked appointments tied back to the click. Suddenly the real number is 23. Not 87.

That sounds like bad news. It isn't. Now he knows which campaigns produce the 23 and which produce the noise. He cuts two ad groups, shifts the budget, and the next month's real conversions go from 23 to 34. Same spend.

The work was boring. Sit with the account for three hours, map out what actually counts as a lead, wire it up, watch it for a week, fix what's wrong.

No clever strategy. No new channel. Just measuring the right thing.

If you're spending real money on ads or SEO and can't tell me, in plain language, what your last 10 paying customers cost you to acquire, that's the project. Before the next campaign. Before the next agency call. Before anything else.

Here's a myth I want to put to rest for local owners thinking about AI search.

The myth: you need to write content "for the AI." Stuff your pages with prompt-style headers. Add a hidden section that talks directly to ChatGPT. Buy a tool that scores your "AI readiness."

The reality: AI systems are reading the same web everyone else reads. Your service pages. Your reviews. Your Google Business Profile. The chamber directory you joined in 2019. The local news piece that mentioned you. The HomeAdvisor profile you forgot exists.

When somebody asks ChatGPT for a good electrician in Roseville, the model isn't visiting your site looking for secret AI signals. It's pulling from the corpus of evidence about your business that already exists across the web, then summarizing it.

So the practical work doesn't change much. It just gets weighted differently.

Write service pages that actually describe what you do, in which towns, at what price range, with what guarantees. Ask customers to mention the service and the city in reviews. Keep your name, address, and phone identical across every listing you can find. Get mentioned by local sources that already have authority: the chamber, the BBB, a trade association, a partner business, a local news outlet.

That's the work. It's the same work that made you findable on Google five years ago. The audience reading the evidence just got bigger.

If an agency is selling you something more exotic than that, ask them to explain in plain English what they're actually changing on the open web. If they can't, you have your answer.

Here is a quiet AEO truth that doesn't get talked about enough: your reviews are training data.

Not in the technical sense. In the practical sense. When an AI system tries to recommend a plumber in Roseville for a slab leak, it reads what people have written about plumbers in Roseville. If your 80 reviews all say "great service, highly recommend," you've given the system nothing to work with. You're a wall of beige.

The reviews that earn citations sound different. They name the problem. They name the city. They name what got fixed. "Called them Sunday night when my water heater flooded the garage in Rocklin. Tech was out by 9am Monday, replaced the unit, hauled the old one away." That review is evidence. It tells an AI system, and a human reader, exactly what kind of job you handle and where.

You don't write your customers' reviews for them. But you can ask better. After a job goes well, the ask isn't "leave us a review." It's "if you have a minute, it really helps when people mention what we worked on and the part of town you're in. Future customers searching for that exact thing find us because of it."

Same ask. Specific request. Different reviews. Different evidence trail.

The owner who builds that habit over a year ends up with a profile that AI systems can actually quote from. The owner who doesn't ends up wondering why competitors with fewer reviews keep getting mentioned instead.

Here is the AEO metric I keep coming back to when owners ask what to track: citation sources.

Not how often you appear in AI answers. Where the AI pulled the information from when it mentioned you.

I had a contractor ask me last month why ChatGPT described his service area wrong. I asked him to run the prompt and look at the sources cited under the answer. Three of the five were old directory listings with his previous address. One was a Yelp page he hadn't touched in four years. The fifth was his actual website, which had the correct info.

The AI wasn't lying. It was averaging.

If you want to know what AI systems will say about you, look at what they're already reading about you. That means your GBP, your top three or four directory listings, your Yelp, your BBB page, your chamber profile, and any local news mention. Open each one. Check the address, the phone, the service list, the hours, the service areas. Fix the ones that contradict your website.

This is boring work. It's also the work that changes what AI says about you in the next 60 days. Nothing else you do matters as much if the underlying evidence is wrong.

Prompt engineering is not the job. Evidence hygiene is.

Branded search lift is the AEO metric I'd watch before any other.

Here's why. When AI platforms recommend you, most customers don't click from the AI window. They open a new tab and type your business name into Google. That second search is the receipt.

So pull Google Search Console. Filter queries to your business name and common misspellings. Look at impressions over the last 90 days versus the prior 90.

If the line is climbing while your ad spend and review velocity are flat, something is mentioning you in places you can't see. That something is usually an AI answer, a Reddit thread, a local blog, or a directory you forgot existed.

If the line is flat, the evidence trail about your business isn't strong enough yet for any system, human or AI, to repeat your name.

It's a boring chart. It tells you more than any "AI visibility score" an agency will sell you.

Here is a quiet truth about AI search that doesn't make for a flashy headline.

72% of people who use AI search also still use Google. That's from Datos survey data cited by SearchLab.

So when an owner asks me whether they should "switch focus" to ChatGPT or stay on Google, I tell them the question is broken. Their customer is doing both. Often in the same sitting. They ask ChatGPT for a shortlist, then Google each name to check reviews, photos, and the website before they call.

That changes what you should actually be doing this week.

Your Google Business Profile has to back up whatever ChatGPT says about you. If the AI calls you a drain specialist in Roseville and your GBP photos are from 2022, your reviews don't mention drains, and your service page is one paragraph of generic plumbing copy, the customer closes the tab.

The work is making sure every place a customer might verify you tells the same story. Same services. Same cities. Same specialties. Same proof.

AI search didn't replace the buying process. It added a step at the front of it. The businesses that win the AI mention and then lose the verification check are the ones who think this is one channel instead of two working together.

Audit both sides this week. Read your GBP and your top service page like a customer who just heard your name from ChatGPT and is deciding whether to trust it.

Here is a platform comparison I wish more owners understood before they spend a dollar trying to "show up in AI."

ChatGPT, Google AI Overviews, and Perplexity are not one channel. They behave differently, source differently, and reward different things.

ChatGPT pulls heavily from Wikipedia, established publishers, and structured business data. If your business has consistent name, address, and phone across the major directories, plus a clean website with real service detail, you have a shot. If your only footprint is a page and a GBP, you are mostly invisible to it.

Google AI Overviews pull from Google's own index. So the work you already do for organic search and your Google Business Profile matters here. Coalition Technologies reported that by March 2025, AI Overviews appeared on about one in eight U.S. desktop searches, and in their 300,000-keyword sample, the top organic result lost roughly a third of its clicks when an Overview showed up. That is the channel where your existing SEO either pays off or quietly bleeds.

Perplexity leans hard on Reddit, forums, and recent web content. Niche, conversational, opinion-rich sources. A plumber with real reviews on Google and a few honest mentions in local Facebook groups or Reddit threads has more raw material here than one with a polished site and nothing else.

What I tell owners: pick the platform where your customers actually are, look at what it pulls from, and feed that. Do not chase all three at once with the same generic content. The evidence trail is different for each one.

Here is the part of AI search nobody wants to talk about: most of the work is invisible to you until a customer mentions it.

A homeowner asks ChatGPT for a roofer in Rocklin. Your name comes up. She doesn't tell you that. She just calls. You log it as a phone lead. Maybe she found you on Google. Maybe a neighbor mentioned you. Maybe ChatGPT did. You'll never know unless you ask.

This is why "how do I track AI search" is the wrong first question. The right first question is: are you asking every new lead how they found you, and writing the answer down?

Not a dropdown on the form. A real question on the call. "Quick one before we schedule, how did you hear about us?" Then a field in your CRM that doesn't get skipped.

Do that for 90 days and you'll learn more about your actual marketing mix than any dashboard will tell you. You'll hear ChatGPT. You'll hear Google. You'll hear "my sister-in-law." You'll hear the Nextdoor post you forgot about.

The boring tracking work always pays. Always. The owners who know where their leads come from spend smarter than the ones who guess.