Do not move Local Service Ads budget on a model's read of your vertical. Count impressions, calls, time to first response, and booked jobs over one dated window first. Until those share a window, there is nothing to buy.
A rural west Tennessee plumber asked Gemini why his new-customer calls dried up. Gemini told him that a solo operator who used to rely on a well-optimized Google Business Profile and strong reviews to get free calls is "now effectively forced to pay for the exact same leads that used to be free." His own read matched the model's: "Our GBP is solid, our reviews are solid, our website is solid and continuing to get better but without LSAs the phone isn't ringing much from new customers."
Another operator looking at the same kind of listing said the Map Pack call button is still displaying for him.
Neither observation measures impressions, calls, answer rate, or booked jobs. Before you move budget into Local Service Ads on the strength of a paragraph a model wrote about your vertical, count the calls, and whether they get answered, before you buy the diagnosis.
A model can tell you what is happening across a vertical. Only your own instrumented loop (visibility, calls, response, bookings, revenue) can tell you what is happening in your business.
Two operators, two observations, no measurement between them
The plumber reports a decline in new-customer calls without LSAs, alongside a GBP, review profile, and website he rates as solid. That is a self-assessment of his assets and an impression of his phone.
The other operator reports one thing: the Map Pack call button is still displaying. That is a statement about an interface element on a screen. It does not prove call volume held. It does not prove demand held.
What it does prove is narrower and more useful: whatever is happening to the Map Pack call button is not happening uniformly to everyone. Any claim that Google removed a universal path to free calls has at least one operator standing in front of it as a counterexample.
Between the two of them, nobody has produced a record of impressions, calls received, answer rate, and jobs booked over the same dated window. That absence, not the disagreement, is the reason neither man can act on his own conclusion with confidence.
The market is moving. Nobody measured this plumber's slice of it.
Metricus reported that consumer use of AI tools to find local service providers rose from 6% to 45% in twelve months, and that one in three homeowners under 45 has used an AI assistant to find a home service provider in the past 90 days. The same reporting puts AI recommendations at about 1.2% of local business locations.
Gartner forecast in February 2024 that traditional search engine volume would drop 25% by 2026 due to AI chatbots and virtual agents.
That 1.2% figure is the one worth sitting with. If nearly half of consumers are asking a model and the model names roughly one in a hundred locations, the question stops being "am I ranking" and becomes "am I one of the named." Those are not the same question, and they are not measured the same way.
Whitespark's 2026 local ranking factors put Google Business Profile signals at roughly 25% of local ranking influence, up 15% year over year, with weight shifting to review recency rather than volume, plus profile completeness, posting frequency, behavioral engagement, and brand authority. "Solid reviews" and "recent reviews" are different assets. An operator grading his own profile from memory cannot tell you which one he has.
None of this establishes what happened to one plumber's phone in west Tennessee. Sector-level movement is a reason to measure. It is not a measurement.
Paid is not the safe fallback it used to be
If the conclusion from the Gemini paragraph is "just buy LSAs," that is also a decision that needs numbers, and the numbers moved.
Plumbing & Mechanical reported average LSA cost climbed from $50.46 in 2023 to just over $60.50 in 2024. Google data cited in the same January 2025 article showed a 19% overall increase in cost-per-conversion in Google Ads (phone calls, form fills, chats, or web bookings), with HVAC up 16% and electrical leads up 23%. Average home services touchpoints rose from 4.9 to 5.5 before a form fill or purchase, so each of those more expensive conversions also sits further down a longer path.
What Google changed on LSAs
Two structural changes make the platform less forgiving than it was.
In October 2025 Google replaced the Google Guaranteed and Google Screened badges with a "Google Verified" badge that often no longer appears visibly in search results, which cuts the trust signal homeowners were scanning for. Starting in 2025 Google discontinued credits for "job type not serviced" and "geo not serviced" leads. You now pay for leads you cannot serve if your category and service-area settings are imprecise.
That last one is the unglamorous item. An operator who switches on LSAs to fix a call drought, with loose service-area boundaries and a category list nobody has audited, is buying a leak with no credit path.
Before you blame discovery, time the answer
Here is the failure mode that looks exactly like a lead drought.
CallJolt's published response-time table (not a census of every shop) drops fast after the first few minutes:
| Time to first response | Conversion vs baseline |
|---|---|
| Under 1 minute | Baseline |
| 1 to 5 minutes | About 95% |
| 5 to 10 minutes | About 40% |
| 10 to 30 minutes | About 15% |
| Past 30 minutes | Under 5% |
Past thirty minutes, most of those callers have booked elsewhere. That cliff is easy to miss if you do not time first response.
Coverage makes it worse. 76% of the week (128 of 168 hours) falls outside standard business hours, and call-tracking data shows 35% to 45% of home service calls arrive during those off-hours. Most of the week is off-hours, and a large share of calls land in that block.
The arithmetic is plain: spend $5,000 a month on marketing, miss 62% of the calls it generates, and $3,100 a month ($37,200 a year) buys leads that never get a conversation. An operator with that profile who reads the Gemini answer may buy more discovery before he has established whether the leak is discovery, intake, or both, at a reported average LSA cost of $60.50 in 2024.
The discovery question and the intake question have to be answered in the same pass: use a search checkup to establish the visibility side, then set it against your call handling and booked revenue.
Four measurements, then the budget
Four measurements, in order, over a dated window. Do not change LSA spend until you have them.
- Visibility. Impressions and Map Pack presence for your service terms, plus whether AI assistants name you when asked for a provider in your area. Given that AI recommends about 1.2% of local business locations, "named or not named" is a binary worth recording, not inferring.
- Calls in. Every call that arrives, timestamped, with the off-hours block counted separately. Because call-tracking data puts 35% to 45% of home-service calls outside business hours, count your own off-hours calls before evaluating demand from what the office heard.
- Time to first response. Measured against the five-minute line, not estimated. CallJolt's table drops from about 95% at one to five minutes to about 40% at five to ten minutes.
- Booked jobs and completed revenue, tied back to those calls. This is the only step that separates a marketing problem from an intake problem from a sales problem.
Then fix whichever step the trail shows is leaking. Plumbing & Mechanical reports one home services company with a 45% cancellation rate for drain services that cut it to 35% with an AI dispatching tool and no increase in marketing spend. A ten-point reduction in cancellations, from a process change instead of more ad spend.
The plumber in Tennessee may well be right. His calls may be down and LSAs may be the answer for his market. Right now he is acting on a model's characterization of a vertical and his own impression of his phone, and another operator could still see the Map Pack call button, an observation the broad Gemini diagnosis did not account for.
An AI answer is a hypothesis about your market. Your own numbers are the test. Those four sit in the shop if you log them. Until they share one dated window, there is nothing to buy.
Frequently asked.
Why did a plumber's new-customer calls dry up even with a solid GBP, reviews, and website?
The Gemini paragraph does not establish that. The plumber reported a drop alongside assets he rates as solid. Another operator looking at the same kind of listing still saw the Map Pack call button. Neither man produced impressions, calls, answer rate, and booked jobs over the same dated window, so neither conclusion can be acted on with confidence.
Should I start Local Service Ads as soon as organic calls slow down?
Not from a model's characterization of your vertical. Average LSA cost climbed from $50.46 in 2023 to just over $60.50 in 2024, and Google discontinued credits for leads you cannot serve if category and service-area settings are loose. Count the calls, and whether they get answered, before you buy the diagnosis.
What four measurements should I track before I change LSA budget?
Visibility, calls in, time to first response, and booked jobs with completed revenue, in that order, over one dated window. Do not change LSA spend until you have them. Those four sit in the shop if you log them.
How fast does conversion drop if home-service calls sit unanswered?
CallJolt's published table drops from about 95% at one to five minutes to about 40% at five to ten minutes. Past thirty minutes it is under 5%. That cliff is easy to miss if you do not time first response.
Is the Map Pack call button gone for every business?
No. Another operator said the Map Pack call button is still displaying for him. That does not prove call volume held. It does prove the button is not gone uniformly, so a claim that Google removed a universal path to free calls has at least one counterexample.
Have Local Service Ads costs and policies changed recently?
Yes. Cost-per-lead rose, the Google Guaranteed and Google Screened badges were replaced with a Google Verified badge that often no longer appears in results, and credits for job-type-not-serviced and geo-not-serviced leads ended in 2025. An operator who switches on LSAs with unaudited categories is buying a leak with no credit path.