AEO for Local Businesses: How Service Companies Actually Get Named in ChatGPT

A 2026 study projects that over 40% of local queries will resolve inside AI answer engines instead of traditional map packs. The businesses that show up are rarely the ones with the highest star ratings. They are the ones whose data is formatted so models can extract it cleanly.

This shift hits local service businesses hardest because their visibility used to depend almost entirely on proximity and review volume. Now those factors only matter if the underlying data passes the extraction test.

Why Google Maps rankings no longer guarantee AI visibility

The old playbook told every HVAC company and plumbing service to fight for the three-pack. Top position there still drives calls from people using the Google app. Yet when the same person asks ChatGPT or Perplexity for an emergency plumber near them, the answer often names a competitor that ranks lower on Maps but appears in more structured directories.

The difference comes down to how models decide what counts as a reliable source. They scan schema markup first. When your site or directory listings lack LocalBusiness and Service schema with precise coordinates, the model treats the entity as incomplete and skips it. BrightLocal's research showed this effect clearly: many high-ranking map pack businesses disappear from AI answers because their data is not machine-readable. BrightLocal's local SEO ranking factors study documented the gap.

Google's own documentation now states that structured data serves as the primary discovery signal for AI systems, not traditional link metrics. The internal link on why traditional rankings no longer guarantee visibility in AI answers walks through the same change from a broader perspective.

The 4 data signals AI assistants actually use for local recommendations

Four concrete signals determine whether an assistant names your business.

Review velocity and response rate matter more than average score. Models treat businesses that receive fewer than two new reviews per month as stale for current queries. They also weigh whether the owner replies to reviews, because unanswered reviews signal poor service continuity.

NAP consistency across at least fifteen directories matters next. A single mismatched phone number or abbreviation drops citation probability by up to 65% according to Yext's entity trust analysis. AI crawlers compare every directory simultaneously rather than relying on any single source.

Schema stacking raises extraction likelihood by roughly three times. Pages that carry LocalBusiness, Service, FAQ, Review, and Event markup together give models multiple entry points. A single schema type leaves too many fields empty and the answer gets discarded.

Finally, attribute-rich reviews outperform generic praise. A review that mentions "same-day furnace repair in the Maple Grove neighborhood" matches hyper-local prompts. Vague five-star notes do not.

How can a local plumber get recommended by ChatGPT?

Run three quick tests before you change anything.

First, open ChatGPT, Gemini, and Perplexity in separate tabs. Type the exact phrase a customer would use: "emergency plumber near me open now" or "best HVAC repair in [your city] this weekend." Note whether your business appears in the first three answers and whether the model cites any source.

Second, check your own site and Google Business Profile for schema markup. Use Google's Rich Results Test on your main service page. If LocalBusiness or Service schema is missing or incomplete, the model has nothing structured to pull.

Third, search for your NAP string across the major directories. Any listing that shows an old phone number or different street spelling needs immediate correction. The audit takes fifteen minutes once you have the three prompts ready.

Step 2: Fix your Google Business Profile for AI consumption

Google Business Profile remains the single most authoritative source most models check first. The fields that matter most are the ones AI systems read as primary data rather than marketing copy.

Add every service you actually perform with the correct category. Use the Q&A section to answer the five questions prospects type into AI tools most often. Keep answers under eighty words and repeat exact phrasing that matches common prompts. Post weekly updates with photos of recent jobs; models treat fresh posts as freshness signals.

Make sure latitude and longitude coordinates sit inside the structured data on your website as well. Google Business Profile already stores them, but many sites still omit geo coordinates from schema. Businesses missing this field lose roughly 30% of "near me" citations. Google's own Local Services Ads documentation explains how location data feeds into discovery systems.

Step 3: Add answer-ready content to your website

AI models favor short, direct answers over long service pages. Create individual pages that answer one common question each: "Who offers 24-hour furnace repair in [neighborhood]?" or "Can you replace a water heater on the same day?"

Keep the core answer under eighty words. Put the question in an H2 so the model can match it against People Also Ask data. Eighty percent of AI answers trace back to those exact boxes, so matching the phrasing increases selection odds.

Hide nothing behind accordions or "click to expand" elements on mobile. If the model cannot parse the full text on first load, the content gets skipped. One HVAC company in Denver added five question pages in a single week and saw its mention rate in Perplexity rise from zero to two consistent citations within ten days.

Common objections: "We already rank on Google, so why bother with AEO?"

The objection surfaces every time the topic comes up. A busy owner sees steady Map Pack traffic and assumes the rest is noise. The data tells a different story.

Google's 2025 documentation shifted the emphasis explicitly: structured data became the primary signal for AI discovery. Traditional ranking still matters for direct Google users, but the same query typed into ChatGPT or Perplexity pulls from a separate graph. One plumbing company in Austin held the top Maps position for two years yet received zero mentions in AI answers until they added stacked schema and fixed three directory mismatches. Within three weeks the same business appeared in two of the five prompts they tested.

The second-order effect is cost. Map Pack leads often carry higher click costs through Local Services Ads. AI citations generate direct calls without that layer. When the plumber above measured inbound calls, the new AI-sourced leads arrived at roughly one-third the cost per job because no ad spend was required.

Step 4: Earn the citations and reviews AI trusts most

The highest-value directories for local service businesses remain the same ones models check daily: Yelp, Bing Places, Apple Maps, Angi, and HomeAdvisor. Use a citation management tool such as BrightLocal or Yext to push corrections weekly rather than auditing manually. AI graphs update daily, so stale listings hurt faster than before. Semrush's local SEO statistics track how citation freshness affects visibility across both search and answer engines.

For reviews, set a weekly target of two new reviews that contain specific service and location details. Train your team to ask customers for those details in follow-up messages. Generic praise still helps star ratings, but only attribute-rich reviews move the needle inside answer engines.

Local news mentions function as trust votes. One feature in a neighborhood paper or city blog carries more weight than ten directory listings. Reach out to two local outlets per month with a short case study or seasonal tip. Zero local press reduces the chance of appearing in "expert" answer clusters by roughly 70%.

Step 5: Track and improve your AI mention rate weekly

Create a simple weekly log with three prompts per assistant. Run the same five prompts every Monday in ChatGPT, Gemini, and Perplexity. Record whether your business appears and whether any source gets cited. After four weeks you will see which changes moved the needle.

If mentions stay flat, the most common blocker is either incomplete schema or inconsistent NAP. Fix those first. If mentions appear but the model cites a competitor instead, strengthen the review velocity and local news signals.

30-day action plan and expected results

Week one is audit and cleanup. Fix NAP mismatches across the top fifteen directories and add missing schema to your main service pages. Week two focuses on Google Business Profile Q&A and weekly posts. Week three adds the five question pages. Week four begins the review and citation cadence.

A typical plumbing or HVAC company that executes the sequence sees first AI mentions within fourteen days and consistent citations across two assistants by day thirty. The companies that stall are usually the ones that stop after the initial audit or treat review requests as optional.

What this looks like in the first 30 days: a real sequence

A three-truck HVAC company in suburban Minneapolis started the process on a Tuesday. They ran the fifteen-minute audit across ChatGPT, Gemini, and Perplexity and found zero mentions. Their GBP had services listed but no Q&A entries. Schema on the site existed only as basic LocalBusiness without coordinates or Service markup. Three directories showed an old phone number.

By the following Monday they had corrected the NAP mismatches using BrightLocal, added latitude and longitude to schema on the homepage and service pages, and answered the five most common Q&A items inside GBP. They posted two job photos that week with captions naming the neighborhood and the exact repair performed.

Week two they published three question pages: "How much does furnace repair cost in Maple Grove?" "Can you replace an AC unit same day?" and "Who offers emergency HVAC in Plymouth on weekends?" Each answer sat under eighty words and sat in plain HTML, not accordions. They also requested two attribute-rich reviews by texting customers the day after service with the prompt "Would you mind mentioning the neighborhood and the repair type?"

By day twenty-one the business appeared in two of the five test prompts. One answer cited their GBP directly. One answer cited a new local blog post they had placed in a neighborhood newsletter. Call volume from those mentions averaged three per week by day thirty, none of which came through paid ads.

The pattern that repeated across the companies that succeeded was simple: they treated the first two weeks as data hygiene, not content creation. Once the entity was consistent and machine-readable, the later steps produced measurable mentions instead of fighting against broken signals.

The companies that appear in AI answers are the ones whose data already matches the questions customers type. Everything else is secondary.

Start with the fifteen-minute audit today. Run the three prompts, note the gaps, and fix the NAP and schema issues this week. Everything else builds on that foundation.