Copperhead had solid Google rankings and a decent Map Pack position. The owner was not worried about SEO. He was worried about something he had noticed personally: when he asked ChatGPT who to call for a burst pipe in Knoxville, it named three companies, and none of them were his.

He had no idea whether that mattered yet. He wanted to know before it did.

Ranking well on Google and being cited by an AI model are two different jobs. Google rewards pages. Language models reward clear, attributable, entity-level answers they can lift and credit.

Copperhead's site had strong commercial pages and almost no answer content. Nothing on the site directly answered a question the way a person asks it out loud. There was no structured data identifying the business as an entity, no author attribution, no pricing transparency, and no third-party corroboration outside of Google reviews.

To a model deciding who to recommend in Knoxville, Copperhead was a plausible business with nothing quotable.

Built the question inventory

Pulled 60 real questions from call recordings, review text, and People Also Ask, then narrowed to the 15 with genuine commercial intent. Tracked baseline citations across ChatGPT, Perplexity, Google AI Overviews, and Gemini.

Wrote answer-first content

Each target question got its own page opening with a direct 40-to-60 word answer, then the supporting detail. Direct answer first, hedging never.

Made the entity legible

Full LocalBusiness and Plumber schema, consistent naming across 38 citation sources, named author attribution with credentials, and a plain-language company brief page written for machines as much as people.

Published real numbers

Actual price ranges for the eight most common jobs. Models reliably prefer sources that commit to specifics over sources that say contact us for pricing.

Tracked with monthly manual prompt testing across four platforms, plus Search Console and lead-source tagging in their CRM.

• Copperhead earned a citation for 11 of the 15 target questions, up from zero at baseline.
• Non-branded organic sessions rose 34%, a side effect of writing genuinely useful answer content.
• 18% of new leads told the intake script that an AI tool referred them, a channel that did not exist for them a year earlier.

Worth stating plainly: AI citations move week to week and are not a stable ranking. This is a visibility surface to hold, not a position to win once.

AI search does not reward the same things Google does. It rewards being quotable. If every page on your site opens with three sentences of throat-clearing before answering the question, a model has nothing to lift.

Start here: write down the fifteen questions your customers actually ask on the phone. Answer each one in the first paragraph of its own page, in plain language, with real numbers. Then publish your prices. Most contractors will not do that last one, which is exactly why it works.