AI names local businesses more often for the jobs their website names
We did not ask AI for "a plumber". We asked who to call for one job at a time: drain cleaning, a tankless water heater installation, a sewer line repair. We put 120 questions like those to AI across 10 local markets, read 480 answers, then checked whose website named that job. 1,164 comparisons.
- On average, AI named a business 35 percentage points more often for the jobs its website names. That is across the 63 businesses we could compare against themselves. The middle one gained 26 points.
- Jobs named in a page's address and jobs named only in a homepage link were both named by AI more often than jobs named in neither. With 32 link-only cases we cannot say which of the two did better.
- The typical business website names one in 12 of the jobs we asked about in its town, in a page address or a homepage link. We did not read page text. 16 of the 91 name half or more.
- We cannot say it is cause and effect. A shop might name the work because it is their main trade, and get named for being the specialist. We also did not check which jobs each shop actually does.
Go read your own homepage. Does a link on it, or the address of one of your pages, name the jobs people call you for? If not, add them there.
- Burst pipe repair
- Emergency drain cleaning
- Tankless water heaters
- Plumbing services
- About us
- Contact
Ask an AI assistant who to call for drain cleaning, then who to call for a tankless water heater installation, and it can name different businesses. Each job is a different question, and each one is a separate roll of the dice for every business in town.
We wanted to know what goes with being named. So we put 120 of those questions to AI across 10 local markets, four times each, and recorded every business it named. Then we checked the page addresses and homepage links of 91 businesses' websites in those markets to see which of the jobs each one names. That gives 1,164 pairs of "was this business named for this job" and "does its website name this job".
The pattern is strong and the honest reading of it is narrower than we would like. Both are below.
The same shop, compared against itself
The obvious way to test this is to line up every business and see whether the ones whose websites name more jobs get named more. That test is easy and close to worthless. A shop whose site names more jobs may simply be a bigger shop, and bigger shops may get named more anyway. You would be measuring size and calling it wording.
So we did it the other way. 63 of the businesses name some of the jobs we asked about in their town and not others. For each one we compared it against itself: how often AI named it for the jobs its website names, against how often it named the same business for the jobs its website does not. Same business, same size, same reviews. What differs is the job, and whether the website names it.
Naming ran 35 percentage points higher on the jobs a business's website names. 47 businesses did better on the jobs they name, 13 did worse, 3 came out even. That split is not a coin landing funny.
- Each mark: one business, compared only against itself.
- Left to right: how much more often AI named it for the jobs its website names, in percentage points. The line down the middle is ZERO, meaning no difference at all. A dot at +40 means that business was named 40 points more often for the jobs it names.
- Up and down: nothing. Dots are stacked upwards only so overlapping ones stay countable.
- What to look for: which side of the middle line the dots pile up on. Most sit to the right. Because each dot compares a business with itself, that lean cannot be explained by some businesses simply being bigger or better known.
All 63 businesses that can be compared against themselves
| business | jobs its site names | jobs it does not | AI named it: jobs it names | AI named it: jobs it does not | difference |
|---|---|---|---|---|---|
| Supreme Seamless Rain Gutter Systems | 1 | 11 | 100% | 0% | +100 points |
| Edgewater Rain Gutter Co | 1 | 11 | 100% | 0% | +100 points |
| Melendez Tree Service | 1 | 11 | 100% | 0% | +100 points |
| MVP Plumbing & Drain Services Inc | 1 | 11 | 100% | 0% | +100 points |
| Protection4Less | 1 | 11 | 100% | 0% | +100 points |
| Rooter Shooter Jetting | 1 | 11 | 100% | 0% | +100 points |
| ArborWorld | 1 | 11 | 100% | 0% | +100 points |
| California Arbor Care | 1 | 11 | 100% | 0% | +100 points |
| HB Security Cameras | 1 | 11 | 100% | 0% | +100 points |
| RCP Plumbing Inc | 1 | 11 | 100% | 0% | +100 points |
| Vista Smog & Test Only Center | 1 | 11 | 100% | 0% | +100 points |
| BrightView | 1 | 11 | 100% | 9% | +91 points |
| Fast Water Heater | 1 | 11 | 100% | 9% | +91 points |
| Way To Go Builders | 1 | 11 | 100% | 9% | +91 points |
| Ridgeline Outdoor Living | 2 | 10 | 100% | 10% | +90 points |
| Greenside Landscaping Inc | 1 | 11 | 100% | 18% | +82 points |
| Ecoworkz | 2 | 10 | 100% | 30% | +70 points |
| New Quest Electric | 2 | 10 | 100% | 30% | +70 points |
| Big O Tires | 3 | 9 | 100% | 33% | +67 points |
| Chloe Nail Lounge | 2 | 10 | 100% | 40% | +60 points |
| luv nail | 2 | 10 | 100% | 50% | +50 points |
| Valvoline Instant Oil Change | 2 | 10 | 50% | 0% | +50 points |
| Paramount Yards | 2 | 10 | 50% | 10% | +40 points |
| Cruz Electric | 3 | 9 | 100% | 67% | +33 points |
| Above All Landscape Company | 3 | 9 | 33% | 0% | +33 points |
| Sprinklers and Turf | 3 | 9 | 33% | 0% | +33 points |
| Vivid Landscape | 3 | 9 | 33% | 0% | +33 points |
| Land X Landscape Construction | 2 | 10 | 50% | 20% | +30 points |
| Ace Pelizon Plumbing | 7 | 5 | 29% | 0% | +29 points |
| Drain Doctor Plumbing & Rooter | 7 | 5 | 29% | 0% | +29 points |
| SAHARA NAILS AND LASHES | 1 | 11 | 100% | 73% | +27 points |
| Autonovation Mobile Auto Repair | 5 | 7 | 40% | 14% | +26 points |
| Kilowire Electric | 4 | 8 | 25% | 0% | +25 points |
| American Tire Depot | 4 | 8 | 25% | 0% | +25 points |
| Garvey Roofing | 1 | 11 | 100% | 82% | +18 points |
| Marvelous Nails And Spa | 1 | 11 | 100% | 82% | +18 points |
| So-Cal 1 Plumbing & Rooter | 6 | 6 | 17% | 0% | +17 points |
| SpeeDee Drains and Plumbing | 6 | 6 | 17% | 0% | +17 points |
| Craig Johnson Automotive | 6 | 6 | 17% | 0% | +17 points |
| Johnny's Auto Service | 6 | 6 | 67% | 50% | +17 points |
| J.I.G. Plumbing | 7 | 5 | 14% | 0% | +14 points |
| Rooter Hero Plumbing & Air | 7 | 5 | 14% | 0% | +14 points |
| Service Hero | 8 | 4 | 13% | 0% | +13 points |
| Nolasco Plumbing, Inc | 18 | 6 | 28% | 17% | +11 points |
| Done Rite Roofing LLC | 3 | 9 | 33% | 22% | +11 points |
| Trout Electric | 5 | 19 | 20% | 11% | +9 points |
| Raul's Auto Repair | 7 | 5 | 43% | 40% | +3 points |
| Solace Plumbing Heating and Air | 6 | 6 | 33% | 33% | 0 points |
| High Speed Plumbing | 6 | 6 | 33% | 33% | 0 points |
| The Electric Connection | 18 | 18 | 39% | 39% | 0 points |
| PONCE LANDSCAPING | 1 | 11 | 0% | 9% | -9 points |
| System Pavers | 1 | 11 | 0% | 9% | -9 points |
| Hoyo Rooter | 1 | 11 | 0% | 9% | -9 points |
| Rain Dance Roofing | 1 | 11 | 0% | 9% | -9 points |
| Brennan Electrical | 2 | 10 | 0% | 10% | -10 points |
| Leo's Auto Care | 4 | 8 | 50% | 63% | -12 points |
| SoCal Plumbing & Rooter | 6 | 6 | 0% | 17% | -17 points |
| Rancho Landscape and Tree Care | 1 | 11 | 0% | 18% | -18 points |
| TopCal Roofing | 3 | 9 | 33% | 56% | -22 points |
| San Gabriel Valley Roofing | 1 | 11 | 0% | 27% | -27 points |
| Corona Landscapers | 2 | 10 | 0% | 30% | -30 points |
| Power Pro Plumbing Heating & Air | 9 | 3 | 0% | 33% | -33 points |
| DEL MAR LANDSCAPING | 1 | 11 | 0% | 45% | -45 points |
AI named these businesses more often for the jobs their websites name.
Both ways of naming a job beat naming it in neither place
A job counted as named in one of two ways. Every word of it is in the address of one of the site's own pages, like /sewer-line-repair. Or it is only in the words of a link on the homepage. So we split the named cases along that line.
Jobs named in a page's address and jobs named only in a homepage link were both named by AI more often than jobs named in neither. With 32 link-only cases we cannot say which of the two did better.
This is not a page against no page. A link whose words name the job can open a page about that job, with an address that does not spell it out, and some of these do. So the split cannot tell you whether a page for each job is worth building.
- Each mark: one of the three cases we sorted each business-and-job pair into.
- Left to right: share of those cases where AI named the business for that job.
- Up and down: the three cases.
- What to look for: the size of the gap down to the bottom bar. That gap is the finding. The middle bar rests on few cases, so do not read a ranking into the top two.
The three cases
| how the job showed up | cases | named at least once | rate |
|---|---|---|---|
| in a page's address | 186 | 64 | 34% |
| only in the words of a homepage link | 32 | 18 | 56% |
| in neither | 946 | 189 | 20% |
Why would the words matter? Somebody asking who does sewer line repair in West Covina uses the name of the job, and the plainest reading of the pattern is that the assistant favours the sites that say it back. We did not test that. It is our reading.
Name the job plainly, where it can be found. A page's address and a homepage link both went with being named more often than naming it in neither.
The typical business website names one in 12 of the jobs we asked about
Set aside cause and effect for a second, because this next number does not depend on it. For each trade and town we asked about a dozen different jobs, and each one is a separate chance to be named or not named. The list of jobs is one our own tool generated. It is not a measure of what that town asks for most.
The typical business in this study names one in 12 of those jobs in a page address or a homepage link, the two places we looked. 28 of the 91 businesses name none of them in either place. A roofer's links say roofing. The questions are about roof leak repair, skylight repair, gutter installation and tile roof repair, one at a time. We did not read the text of the pages, so a job described only in a paragraph does not count here.
- Each mark: one business, drawn as a dot on the strip below the curve.
- Left to right: share of the jobs we asked about in its town that this business's website names. 0 means none, 1 would mean all of them.
- Up and down: the share of businesses that land in each slice.
- What to look for: where the mass sits. A pile at the left means most businesses share the same gap.
Coverage, counted
| cut | count |
|---|---|
| businesses that name no job at all | 28 |
| businesses that name half or more of the jobs | 16 |
| typical (median) share of jobs named | 8% |
| businesses measured | 91 |
| business-and-job pairs checked | 1,164 |
Whatever is going on underneath, most of these websites name few of the jobs we asked about in the two places we looked. 16 of the 91 name half or more.
What we would actually do with this
Here is the honest version of the advice, with the uncertainty left in.
Worth doing
- Write down every job you get paid for. Not your three headline services. The whole list, by the plain name of each job. This costs nothing.
- Check whether your homepage links and page addresses say those words. Open your site and look. The typical site in this study named one in 12 of the jobs we asked about.
- Put them where they can be found. A link on your homepage whose words name the job, or a page whose address does. Those are the two places we checked, and a job named in either was named by AI more often than a job named in neither.
- Use the plain name of the job. We looked for each job by the name the question used, so that is the wording this result is about.
Worth being careful about
- Do not list work you do not want. If naming a job is partly a signal of what you specialise in, filling your homepage with everything may dilute what you are known for. Name what you actually want to be called about.
- Do not buy a hundred thin pages on the strength of this. We did not measure what is on a page or how much of it there is, so nothing here says a stack of near-empty pages per keyword pays.
The question is not "is my website good". It is "does my website say the thing somebody just asked for, somewhere they can see it".
List your jobs, then check whether your homepage links and page addresses name them. That is free and it takes twenty minutes. None of the 91 sites we measured named every job we asked about in the two places we looked.
How we measured this
We put one fixed question per job to a real AI assistant, four times each, and recorded every business it named. Each question named the trade, the job and the town, and asked which local businesses to consider. Asking once tells you what it said that second; asking repeatedly makes it a measurement. Then we read each business's own website, the addresses of the pages in its sitemap where it has one (up to 2,000 of them), and the links on the page we had scanned for it, and counted a job as named only when every word of it appears in a page's address or in the words of a link on that page, ignoring filler such as "services" and most plural endings. That page is the homepage for 83 of the 91 businesses and a location or landing page for the rest; this article says "homepage link" for all of them. A business counts as named by AI for a job when it appears in at least one of the answers to that question. No business paid to be here, none was told it was being measured, and nothing was re-run to get a nicer answer.
The 120 questions are 79 different jobs, some asked in more than one town.
The AI answers were collected between July 28, 2026 and August 2, 2026. The websites were read by July 30, 2026.
Scope
- This is a strong association, not proof of cause. A business may name a job prominently because that job is what it is known for.
- We did not check which jobs each business actually does. A business that does not offer a job will not name it and should not be named for it. Those cases are counted with the jobs named in neither place.
- The jobs are a list our own tool generated, about a dozen for each trade and town. They are not a measure of what each town asks for most.
- 91 businesses, 10 markets, mostly West Covina. 68 of the businesses and 816 of the 1,164 pairs are there. 3 of the 10 markets have three businesses or fewer in the study. Real sample, not a census.
- Every business here was named by AI for at least one job. They come from the businesses AI named in each market. A business AI never names is outside this study.
- The matching rule looks in two places and errs toward missing real mentions. It does not read page text, and now and then it matches words scattered across one link or one address. Real coverage is likely higher than we report. That most likely shrinks the difference we measured, and it makes the gap in the last chart look wider than it is.
- One assistant, one stretch of time. The dates above and at the foot of this page say when.
See what an AI assistant can work out about your business
The free scan reads your site the way an AI assistant does and tells you what it can and cannot work out about your business, including the services it found on your site. No sign-up, no card.
Corrected October 11, 2026. An earlier version of this page said we put 1,486 questions to AI across 10 markets. That figure counted every market we hold. The ten markets in this study were asked 120. It also said a homepage menu link did as well as a whole page, and that you do not need a page for every job. What we compared was a job named in a page's address against a job named only in the words of a homepage link, which tests neither. Its short version said "35% more often" where the measurement is 35 percentage points. Its foot said "Measured September 29, 2026", the day the numbers were computed, when the answers were collected in July and August. It said nobody had closed this gap, when 16 of the 91 name half or more of the jobs. And it described the jobs as what each town asks about. They are a list our own tool generated. Its title said AI recommends the businesses whose website names that exact job. The finding is "more often", and the match is not exact. And it said businesses had a page, or no page, for a job, where we checked only page addresses and the words of homepage links. The measurements themselves have not changed.
c07198cc, data fingerprint 980d8a28646d. These are the numbers as of that date, on that corpus. The web is being rebuilt around how AI reads it, so they will move. When they move enough to change the advice, we publish a new study rather than quietly editing this one. See every study we have published.