- AI engines retrieve, they don’t guess: they pull from profiles, reviews, rankings, and pages they can verify, then name the companies the evidence supports.
- The signals are buildable: One consistent identity, review depth with specifics, crawlable answer-first pages, corroborating mentions.
- Nobody can force an output. Anyone guaranteeing an AI recommendation is guessing on your invoice.
- Measurement is X/12, with method: 12 real buying questions, three platforms, repeated runs, screenshots.
What happens when someone asks
When a customer types “electrician near me for a panel upgrade”, the assistant doesn’t consult a secret ranking. It searches, its own index, the live web, or both, pulls the sources it trusts about electrical companies in that area, and composes an answer from what those sources agree on. The engines differ in the mix (Google AI leans on its search and Maps data; Perplexity shows its citations; ChatGPT blends browsing with what it already knows), but the shape is the same: retrieval, then verification, then a recommendation it can defend.
That last part is the lever. An engine naming an electrician takes a small reputational risk. It prefers companies whose story checks out everywhere it looks: profile matches site, site matches reviews, reviews describe real work, and other sources corroborate all of it. Contradictions are how you get skipped, because the engine cannot safely recommend you, since you’re unsafe to recommend.
What an electrician can actually build
One identity, everywhere. Same name, address, phone, and service list on your site, Business Profile, and every listing. Every conflict is a reason to skip you.
Reviews with specifics. “Great job” proves little; “upgraded our panel for the EV charger, pulled the permits, passed inspection first time” is quotable evidence. Volume and recency matter; detail is what engines can cite.
Pages that answer buyer questions plainly. Panel upgrade costs, ev charger installation, aluminum wiring, whole-home surge protection, written so a machine can lift the answer and name the source. The SEO layer and the AI layer are the same work.
Crawler access. If your site blocks AI crawlers, you’ve opted out of the answer. Make robots.txt a decision, not an accident.
Corroboration. Chamber listings, supplier mentions, local press, directories, third-party confirmation that a real, established company stands behind the claims.
How to know if it’s working
The honest metric is Share of Answer: take 12 real buying questions for your market, ask them on Google AI Overviews, ChatGPT, and Perplexity, repeat each three times in clean sessions, and count the questions where your company is named in at least two of three runs. That’s your X/12, per platform, dated, with screenshots.
Run it monthly and the trend tells the truth. Single runs prove nothing, AI answers vary, which is why one triumphant screenshot is selling, not measuring. Beware percentage “AI scores” with no method, and guarantees of inclusion: nobody controls the outputs. The work is being easy to verify, cite, and recommend, here’s how we run that as a service.
The electrical starting list
In order of leverage: fix every name/address/phone conflict you can find; make the review ask systematic (every job, same-day, compliant, nudging customers toward specifics by asking about the project, never the sentiment); publish honest answers to your five most-asked questions (panel upgrade costs, EV charger installation, aluminum wiring, whole-home surge protection); check your robots.txt; then claim the two or three local listings that matter in your market.
Then baseline your X/12 before the work, so three months from now you’re reading evidence instead of vibes. The Free Visibility Check is exactly that baseline.
Want your X/12?
The free Visibility Check runs 12 real buying questions where your customers ask them, score, screenshots, first worthwhile fix. No sales call.
Get my free Visibility Check