How to Get Recommended by AI Tools

Updated May 2026 • 9 min read

A homeowner in Lexington wakes up to a brown stain on the ceiling after a Saturday-night thunderstorm. By 6:47 a.m. she opens ChatGPT and types, "We have a leak in our roof in Lexington SC — needs an inspection today, single-story 1990s ranch with architectural shingles. Who would you call?" The AI answers in 12 seconds, names two roofing contractors with one-sentence descriptions, and the homeowner calls the first one named. By 9 a.m. that contractor is on a ladder; by Tuesday morning the contract is signed.

The other roofers in Lexington, Chapin, and Irmo who could have handled the job did not lose because their work is worse. They lost because the AI did not name them. This article is about how to be the contractor it names — and not by accident.

What "Recommended by AI" Actually Means

AI assistants do not have a ranked list the way Google does. They have a retrieval step and a synthesis step, and being "recommended" means three specific things happen in sequence:

1. You are retrieved as a candidate. The assistant's search step surfaces your business as relevant to the query.
2. You are described accurately. The assistant lifts specific, correct facts about your category, service area, and offerings from your public content.
3. You are named in the synthesized answer, not buried in a footnote or hedged ("a Lexington-area roofer may be able to help").

All three are required. Most small businesses fail at step 1 (not retrieved) or step 2 (retrieved but described in generic terms the AI cannot confidently use). The ones that fail at step 3 usually have a fixable schema, NAP, or content-specificity gap.

The Five Levers That Move AI Recommendation

Lever 1: Specificity in Your Public Content

A Lexington roofer who says "we service the Midlands" gets retrieved generically. A roofer who says "we replace architectural shingle roofs in Lexington, Chapin, Irmo, Cayce, and West Columbia, with GAF Timberline HDZ as our standard upgrade and a 25-year transferable workmanship warranty on every full replacement" gets retrieved by name when the homeowner asks about exactly that. Specificity is not optional — it is the retrieval signal.

For a roofing contractor, the specifics that materially move AI recommendation include shingle brands and product lines you install, manufacturer certifications you hold (GAF Master Elite, Owens Corning Platinum Preferred, CertainTeed SELECT ShingleMaster), insurance-claim experience (do you work with State Farm, Allstate, USAA adjusters; do you handle supplemental claims), and storm response (24-hour tarping, after-hours emergency response, drone roof inspections).

Lever 2: A Complete Google Business Profile

Google Business Profile remains one of the most-cited individual sources across ChatGPT, Perplexity, and Google AI Overviews. For a Lexington roofer this means: primary category "Roofing contractor," secondary categories ("Gutter cleaning service," "Roof inspector" if applicable), service area covering Lexington and the named adjacent towns, full service list (asphalt shingle, metal, TPO, repair, inspection, gutters, leak diagnosis), attributes (free estimates, online appointments, financing available), 20+ photos including completed jobs, the crew, the truck, and a founder photo, and a recent posting cadence — at least one post per month.

Time investment: 2-3 hours for a complete first-pass; 15 minutes monthly thereafter.

Lever 3: NAP Consistency Across the Right Stack

For roofing contractors specifically, the surfaces AI assistants cross-reference are Google Business Profile, Bing Places, Apple Maps, Yelp, Angi, HomeAdvisor, BBB, Nextdoor, the GAF and Owens Corning contractor locators, the SC Department of Labor licensing-board listing, and your local Home Builders Association directory. Same name, same phone, same address, same hours, same service area — across all of them.

One stale entry showing a phone number you stopped using in 2022 is enough to introduce hedging into an AI's answer that should have been confident.

Lever 4: Reviews With Substance

AI assistants weight reviews that name a specific service, a specific outcome, and ideally a specific crew member. "Tom and the crew replaced our 3,400-square-foot roof in Lexington's Whitehall neighborhood with GAF Timberline HDZ in two days, handled the State Farm hail-claim supplemental, and cleaned every shingle off the lawn" cites differently than "great roofers, highly recommend." Update your post-job review request to prompt customers for: service type, neighborhood, crew name, and what they did not expect.

Lever 5: Schema and Author Credentials on Your Site

Add RoofingContractor schema to your homepage. Add Service schema to each service page (asphalt replacement, metal install, repair, inspection). Add Person schema to your founder/owner bio page with hasCredential pointing to your SC contractor license and any GAF Master Elite or comparable certification. Add FAQPage schema to any Q&A or "what to expect" page.

Validate with the Rich Results Test. Fix every warning.

What AI Tools Look For — In Order

The order matters. When ChatGPT or Perplexity processes "roof leak repair Lexington SC" the steps run in this rough sequence:

1. Topical match. Does your site have a page with the exact phrase or a close paraphrase, with depth? A 2,000-word "Roof Leak Repair in Lexington SC" page beats a generic "Services" listing every time.
2. Local match. Are you named in Google Business Profile and at least 6-8 secondary directories for the city in question, with consistent NAP?
3. Trust check. Do you have third-party signals — reviews with substance, a licensing-board listing, a manufacturer certification, a BBB record, a local-news mention — that confirm you are a real, locally-operating business?
4. Recommendation confidence. Is your business described specifically enough that the AI can write a one-sentence description of you without hedging?

A roofer that passes the first three but fails the fourth gets mentioned in passing. A roofer that passes all four gets named at the top of the answer.

The 60-Day Plan

Sequenced for a sole owner or owner-operator with limited weekly hours.

Weeks 1-2: Baseline + Foundation

• Run the four-assistant prompt test for the 10-12 queries your Lexington customers actually type. Document every answer.
• Complete Google Business Profile to 100%. Add the 20 best job photos you have.
• Fix NAP across the 10 highest-priority roofing directories.

Weeks 3-4: Content Specificity

• Rewrite your three highest-traffic service pages with the specifics from Lever 1: brands, certifications, named towns, warranty details, insurance-claim process.
• Write one new long-form question-shaped page — for a Lexington roofer, "What a Storm-Damage Roof Inspection Actually Looks Like in Lexington, SC" works well.
• Add schema to homepage, service pages, and bio page. Validate.

Weeks 5-6: Reviews + Authority

• Rewrite the post-job review request to prompt for specifics.
• Earn one third-party mention — a quote in a Lexington County Chronicle article about post-storm roof inspections, a chamber spotlight, or a podcast guest spot on a local home-services show.
• Add author bylines with credentials to every page that takes a position.

Weeks 7-8: Re-test + Tune

• Re-run the four-assistant prompt test with the same 10-12 queries. Compare to Week 1.
• Identify the queries where you still are not named and look at what the cited competitor has that you do not. Usually it is a single specific page, a single missing schema, or a review pattern.
• Close that gap. Re-test again in 30 days.

What Does Not Work

Common advice we see Lexington-area contractors waste time and money on:

• Submitting to 300 generic business directories. The top 10-12 roofing-specific and local directories matter. The rest are noise the AI ignores.
• Buying "AI-optimized content" packages from outsourced agencies. The output is detectably generic and frequently down-weighted by the same AI assistants you are trying to be cited by.
• Stuffing the site with "Lexington roof repair" 40 times per page. Modern retrieval models penalize this. Density was a 2014 game.
• Buying backlink packages. AI assistants do not weight purchased links the way 2016-era Google did. Wasted money.

The Bottom Line

Getting recommended by AI tools is not a trick. It is the result of being specific in your public content, consistent across the surfaces AI assistants cross-reference, trustworthy in the third-party signals that confirm you, and current in the maintenance cadence that keeps the position you earn. For a Lexington roofing contractor, eight weeks of focused part-time work crosses the threshold from "occasionally mentioned" to "named by default." The homeowner with the brown stain on the ceiling at 6:47 a.m. will not wait — but you do not have to be invisible to her either.