Doheny Co.

Be the name AI gives.

The Doheny Score.

One number: how often AI names you when a customer asks who to call. Move it, and the next call is yours instead of the shop down the road.

What happens when your customer asks.

An assistant does not know anything about your town. It goes and reads. Every answer says what it read, so we can follow it backwards. Here is a real one, in auto repair.

1 · Your customer types this

Who should I call for auto repair in Boise, ID?

A buying question, with the town in it. One of the 5 we put to every assistant in this trade.

2 · Four assistants go looking, and they do not agree

ChatGPT61% of your score
Gemini20% of your score
Claude14% of your score
Perplexity5% of your score

Weighted by how likely your customer is standing in front of each one. Ask only one and you get a different answer, not a rougher one: across 160 calls the four agreed on who to name between 4% and 14% of the time.

3 · This is what they actually read

Shop websites55%
Carfax8%
Reddit7%
Yelp4%
BBB3%
Everything else23%

Share of all 157 sources cited across this trade’s answers. The single biggest one is not a directory. It is some shop’s own website.

4 · So the work splits in two

What we can write

The shop’s own pages, and the free listing the assistants lean on hardest.

What we can only earn

Reddit threads and review platforms. Real, slower, and never faked.

That is the whole of it. We read what the answer read, then we go and change it.

And it is a different picture in every trade. Same shape of question, different sources entirely, which is why the first thing we do is measure yours. The eight trades we have measured.

One year.

US consumers who used AI to find a local business

20256%
202645%

7x in one year.

Google’s share of local business discovery

202583%
202671%

Down 12 points.

Where people look for a local business now

1Google
2Facebook
3AI assistants
4Yelp
5Tripadvisor

Used in the past year to get a business recommendation

ChatGPT31%
Google AI Mode23%

Almost half your market now asks a machine first, and they arrive believing it.

It does not land like an ad. It lands like a friend told them where to go.

BrightLocal, Local Consumer Review Survey 2026. A representative panel of 1,002 US adults, published March 2026.

Not who you think.

US adults who use an AI chatbot, by age

18 to 2966%54% search
30 to 4961%53% search
50 to 6442%37% search
65 and up23%20% search
Uses an AI chatbotUses one to look something up

Pew Research Center, 5,119 US adults, surveyed 17 to 23 February 2026. The second figure is a subset of the first, which is why it is drawn inside it.

64%

of 30 to 44 year olds use AI to pick a local business. The highest of any age band, and the people who own the house and the car.

24%

of people 60 and over have used AI to find a local business. A year ago it was 6% for everybody.

BrightLocal, Local Consumer Review Survey 2026.

It leans younger. The drop-off comes far later than people plan for.

How the number is worked out.

Four inputs. Nothing else is in it.

1

The questions your customers actually type

Built from real search demand in your town and your trade

Who should I call to fix my furnace in Salt Lake City?I have a nail in my tire, repair or replace?Best tire shop in Fayetteville?Who is the best injector near me?

Not “is my business online”. Buying questions, with the town in them, because that is how a customer asks. The exact set for your trade is the part we keep to ourselves.

2

Asked of four assistants, weighted by reach

How likely your customer is standing in front of each one

61%20%
ChatGPT 61%Gemini 20%Claude 14%Perplexity 5%

Reach, not preference. Gemini hands out more names per answer than any of the four and earns nothing extra for it, because a recommendation nobody sees is worth nothing. If an assistant names no local business in your trade at all, it earns no weight either, and what it would have held moves to the ones that answered, so a quiet engine never drags your number down.

Where these weights come from, and what they leave out. They are US web-visit share, reported by Similarweb in August 2026, normalised across the four we ask. Visit share is a proxy for how many people use a tool, not for how many ask it who to hire locally, and it is the weaker of the two proxies we could use. Google’s AI Mode is not one of the four, and by our own chart above it is the second most used surface for a business recommendation. It has no stable way to ask it the same question twice and get a comparable answer, so measuring it weekly would produce a number that moves for reasons that are not you. We would rather leave it out and say so than fold in a reading we cannot stand behind. Same for Copilot and Grok. When that changes, the weights change and this page changes with them.

4–14%

How often the four agree on who to name. We asked the same 40 questions across all four, in eight trades, 160 calls.

So one assistant is not a rough version of the answer. It is a different answer.

3

Every answer scored on a ladder

What one answer is worth. Named is not all or nothing.

Named first1.00
Named second0.85
Named further down0.70 to 0.50
Read your site, named someone else0.25
Not named at all0.00

Average them within each assistant, weight those four averages by reach, multiply by 100. A score is a percentage, so it cannot go above 100. The published arithmetic is here.

4

Three things that earn nothing

0.00Your Google rankingYou can be first on Google and named by no assistant at all. We have measured exactly that. If Google moved this number, the number would hide the problem you hired us for.
0.00An answer that names nobodyAssistants sometimes give general advice and no names. That is not a race you lost, so it leaves the denominator entirely. In some trades that is up to six answers in ten.
0.00Anything we could not checkIf your Google listing carries no website, we cannot read your site, and the report says so. A number we did not measure is never printed as a nought.

Your trade decides how much of this we can move.

Assistants build a local answer out of whatever they trust in that trade. Sometimes that is mostly your own website. Sometimes it is a directory we cannot touch. It changes what the work is, and it changes whether we should take your money at all.

How much of the answer is assembled from the business’s own website

Med spa & cosmetic61%

The rest comes from Yelp, Trustanalytica, Laseraway.

Salons & barbershops56%

The rest comes from Yelp, Modernluxury, Reddit.

Auto repair, tire & towing55%

The rest comes from Carfax, Reddit, Yelp, Bbb.

Home services: HVAC, plumbing, roofing51%

The rest comes from Reddit, Serviceagent, Angi, Bbb.

Dental & orthodontics51%

The rest comes from Reddit, Zocdoc, Yelp, Reviews.birdeye.

Accounting, bookkeeping & agencies40%

The rest comes from Expertise, Clutch.co, Clearlyrated.

Restaurants, cafes & bakeries14%

The rest comes from Uncorkedasheville, Exploreasheville, Reddit, Tripadvisor.

Residential real estate5%

The rest comes from Zillow, Fastexpert, Homelight, Homes.

Measured, not estimated: 160 first-party calls in August 2026. 8 trades, 5 buying questions each, put to all four assistants. That is 20 answers per trade, and each trade ran in one city, so trade and city are not separated here: read these as the shape of the difference between trades, not as precise figures. We are widening it. Source mix is also not the only thing that decides whether we take a trade on, which is why salons sit high on this chart and are not who we call.

Where we do our best work

Auto, med spa, home services. The answer is built out of things we can write.

Where we say so before you pay

Real estate answers come out of Zillow and its rivals, and those rank on closings we cannot manufacture. We tell you that first, not at day 60.

Seven steps. You act in two.

1

Your free report

We ask the four assistants what your customers ask.

Free, no card
2

You start, we start

$495 today, then monthly. Cancel any time. Every dollar back at 60 days if we miss.

Day 1
3

We copy your site over

Same design, same words, same domain. You approve the switch.

You, once
4

We do the work

Your pages, your listings, your reviews, the lists AI quotes.

Us, weekly
5

You tap approve or skip

Nothing publishes without your OK. Everything is reversible.

You, ~10 min a week
6

Your weekly note

What we did, and what moved.

Us
7

Day 60, we measure again

Both numbers side by side. We check it, you do not chase us.

Us

We start today. $495 now, then $495 a month, everything included, cancel any time, no contract. In 60 days you'll be showing up in AI answers more than you are today, or you get every dollar back.

Day 60.

We measure twice, and we come to you

Day 1Your score the day you sign up.
Day 60Higher than day 1, or every dollar back.

In 60 days you'll be showing up in AI answers more than you are today. If not, every dollar back. One refund per customer, and we show you the measurement either way.

Week one

You see the work.

Weeks to a couple of months

You get named consistently.

Anyone promising you the front of the answer in a week is not being straight with you.

How you know it worked.

Your score says whether AI is naming you. It does not say whether that put money in the till. Those are two different numbers and you should have both.

The one we measure

Your score, day 1 against day 60

Same questions, same four assistants, same formula. We run it and we bring it to you.

The one you count

Customers who say an AI sent them

One line on the ticket: how did you hear about us? It costs nothing and it is the number your CFO will actually want.

Add them up at 60 days and you have a measurement that moved and a count of people who told you why they walked in. If the first moved and the second did not, we would rather you knew that than not.

More than one location?

Assistants answer at city level, not street level

Shop 1Shop 2Shop 3Shop 4

“Who should I call in this city?”
One answer.

Four shops in one city compete to be the same single answer, so paying for all four buys it once.

Buy one location per market, not one per address. Twenty rooftops is nowhere near twenty subscriptions.

Easy to put on your payroll

Someone to approve the pages, chase the web firm, and know each town.

The part that does not come with the hire

Measuring four assistants every week in every market, and knowing what each one read and why it changed. And they only ever see your own data.

Worth saying plainly: at some size, bringing this in house is a real option, and we would rather talk about it than pretend otherwise.

Will this hurt our Google?

What Google actually penalises

Hundreds of thin pages nobody read before publishing.

Google’s published guidance judges whether a page is useful, not whether a machine helped write it.

What we do instead

A few good pages about services you really offer, every word approved by you.

It is additive. The same signals Google’s local ranking already rewards, and every word goes past your own people before it publishes.

What we will not tell you is that this lifts your Google ranking. We have not measured that, so we do not sell it.

What we will not do.

We will not promise you the top spot

We will not filter or fake your reviews

We will not tie you to a contract

We will not hand you homework

Google and the assistants decide who they name, not us. The 60-day promise above is the only one we make. The mechanics, with the evidence behind each one.

See your own number first.

Free, no card, about your business rather than about us. $495 a month when you are ready.

Questions? support@thedohenyco.com, and a person replies, usually the same day.