Named by the assistant homeowners ask first.

Panel and EV work is researched for weeks, and that research increasingly happens inside an AI answer. Across twenty buying-intent prompts, the business was named once.

PROMPTS NAMING YOU

1/2013/20

AI OVERVIEWS

014

ESTIMATE REQUESTS

9/mo34/mo

SERVICE PAGE SESSIONS

1,2403,910

KPI 01FROM 1 OF 20

13/20

Tracked prompts naming the business

KPI 02FROM ZERO

AI Overview appearances for core services

KPI 03

3.8x

Estimate requests per month

[ 01 / 04 ] • THE SITUATION

Schema was plugin-generated and contradicted the site: two service lists, an old suite number, and a business name that appeared three ways across directories. Answer engines resolve that ambiguity by naming someone else.

The content had a second problem. Pages described services in company language rather than answering the questions a homeowner types, so even where the entity was clear there was nothing quotable to lift.

Named in 1 of 20 tracked buying prompts

Plugin schema contradicting on-page facts

Business name recorded three different ways

No llms.txt, no FAQ or Service schema

[ 02 / 04 ] • WHAT RAN, IN ORDER

MONTHS 1–2Entity cleanupEvery place the name, address, and service list appeared was mapped, conflicts listed, and a single canonical record established before any markup was written.

MONTHS 2–4Schema in the buildLocalBusiness, Service, FAQPage, and Person schema written into the templates rather than bolted on, plus an llms.txt stating plainly what the business does and where.

MONTHS 4–7Answer-shaped contentPanel capacity, permit requirements, and honest cost ranges written as direct answers, with monthly prompt runs across three engines to see what changed.

[ 03 / 04 ] • BEFORE AND AFTER

Prompts naming the business (of 20)1 → 13

BEFORE

AFTER

AI Overview appearances0 → 14

BEFORE

AFTER

Estimate requests9 / mo → 34 / mo

BEFORE

AFTER

Organic sessions on service pages1,240 / mo → 3,910 / mo

BEFORE

AFTER

Named in 13 of 20 tracked prompts, from 1 of 20 at baseline.

MONTHLY PROMPT RUNS ACROSS CHATGPT, PERPLEXITY, AND AI OVERVIEWS, MONTH 7 VS. BASELINE

[ 04 / 04 ] • WHY IT WORKED

Answer engines read structure before they read prose. Contradictory entity signals are the most common reason a model defaults to a better-defined competitor, and it is a fixable problem rather than a budget problem.

Publishing real cost ranges did the rest. Homeowners were already comparing numbers, and being absent from that comparison was removing the business from the shortlist rather than protecting its margin.

Named by the assistant homeowners ask first.

See what AI says about you.