Will ChatGPT Recommend Your Business? (GEO for Local Service Brands)
Generative Engine Optimization for small service businesses: the 5-axis method we score every site against, with a 60-day fix plan.
By Chase Weiser
When a homeowner asks ChatGPT “who’s a good roofer in Jupiter FL,” your business either comes up or it doesn’t. The same is true for Google’s AI Overviews, for Perplexity, for Claude, for the new search experiences in every major browser. Roughly 30-40% of high-intent commercial searches now run through an AI layer before the user ever sees a traditional results page. If you’ve spent a decade optimizing for Google’s blue links and you’ve ignored the AI layer, you’re optimizing for the smaller half of the market.
This post is GEO/AEO (Generative Engine Optimization / Answer Engine Optimization) for small service businesses. It’s practical, not theoretical.
The short version: AI search reads your site differently
A traditional Google crawler reads your site looking for keywords, links, and structural signals. An AI search engine reads your site looking for answers. Specifically, it’s looking for content it can lift, paraphrase, and present as the answer to a user’s question.
That difference changes everything about how you should structure your pages.
Three patterns make a site AI-friendly:
- Direct Q&A blocks that answer the questions your customers actually ask, in the same words they’d use
- Structured data (schema markup) that tells the AI what kind of business you are, where you operate, and what you do
- Citable facts (numbers, prices, service areas, hours, certifications) presented clearly enough to be lifted into a response
Sites that nail all three get cited. Sites that don’t, disappear from the AI layer.
The 5-axis GEO readiness method
We score every site we scan against five axes. Each one is independently testable. Each one has a fix you can ship in a week or less.
Axis 1: machine-readable structure
Can an AI parse your site at all? This is the foundation. A site that loads via heavy client-side JavaScript with no server-rendered fallback is invisible to most AI crawlers, because they don’t execute JS. A site with proper semantic HTML (real headings, real lists, real tables) is readable.
The test: view-source on your homepage. Can you read the page content in the raw HTML? If yes, you pass. If you see mostly empty divs and a script tag that loads the real content, you fail.
The fix: server-side rendering. For static marketing sites, that’s the default in modern frameworks. For React SPAs, it’s a one-time refactor with measurable payoff.
Axis 2: schema markup completeness
Schema markup is structured data embedded in your HTML that explicitly tells search engines and AI tools what your content is. A local business with proper schema makes itself trivially easy to cite. A site without schema forces the AI to guess, and AI tools often guess wrong or skip the citation entirely.
The minimum schema for a local service business:
| Schema type | What it says | Why it matters |
|---|---|---|
| LocalBusiness (or a specific subtype like RoofingContractor, Plumber, etc.) | This is a local business of type X with these hours, this address, this phone | Powers Google Knowledge Panel, AI citations |
| Organization | This is the parent brand, here’s the logo, the social profiles | Identity disambiguation |
| FAQPage | These are common questions and their answers | Directly fed into AI answer responses |
| Service (one per service offered) | This business offers this specific service | Enables service-specific queries |
| Review and AggregateRating | This business has X reviews averaging Y stars | Trust signal for AI tools |
The test: paste your homepage URL into Google’s Rich Results Test. It should detect at least LocalBusiness, Organization, and FAQPage. If it detects nothing, you fail.
The fix: add JSON-LD blocks to your site head. For most service businesses, it’s a 2-4 hour build.
Axis 3: question-answer content density
AI search engines reward sites that have already answered the questions users ask. The pattern is simple: a section heading that’s a real question (in plain language), followed by a 2-3 paragraph answer that stands alone.
The bad pattern is generic SEO content: “Our HVAC Services in Jupiter, FL” followed by three paragraphs of unfocused copy. An AI tool can’t lift anything useful from that.
The good pattern is:
How much does AC installation cost in Jupiter, FL?
A new central AC installation in Jupiter typically runs $4,500 to $9,500 for a standard 3-ton system, depending on whether the existing ductwork is reusable. Ductless mini-split systems for single-room installations run $3,200 to $5,800 installed. We provide free in-home estimates and include a 10-year parts warranty on every Goodman and Trane install.
That paragraph is an AI’s dream. It answers the question directly, includes a citable number range, and ends with a specific, citable commitment. It’s also good for human readers, which is the point.
The test: pick the top 10 questions your customers ask on the phone. Search your own site for each one. Do you have a clear, paragraph-length answer for each? Most service businesses have answers for 2 or 3 of them.
The fix: a content sprint. One question per blog post or per section, written the way the answer actually sounds.
Axis 4: cite-ability (numbers, specifics, source authority)
AI engines prefer content with concrete numbers, specific claims, and a clear author or organization behind them. Vague promotional copy (“we offer the best service in the area”) is unciteable. Specific factual copy (“we’ve completed 847 roof installations in Palm Beach County since 2018, with an average warranty claim rate under 2%”) is highly citeable.
The test: open any service page on your site. Count the concrete numbers (years in business, jobs completed, response time guarantees, warranty terms, pricing ranges). Count the vague adjectives (“best,” “leading,” “trusted,” “premier”). Healthy ratio is 3 numbers to 1 adjective. Most sites have the inverse.
The fix: a copy rewrite. Replace adjectives with numbers wherever possible. The numbers don’t have to be huge; they have to be specific and true.
Axis 5: AI discoverability layer (llms.txt, sitemap, robots)
The newest axis. AI crawlers respect a small set of standard files that tell them what your site is and what they’re allowed to use. Three matter:
| File | What it does |
|---|---|
robots.txt | Tells crawlers (including AI crawlers like GPTBot, ClaudeBot, PerplexityBot) which paths they can access |
sitemap.xml | Lists every URL on your site, with freshness signals |
llms.txt | Emerging standard. A plain-English summary of your site for AI tools, with priority pointers to your most important pages |
The test: navigate to yoursite.com/robots.txt, yoursite.com/sitemap.xml, yoursite.com/llms.txt. The first two should exist and be readable. The third is optional but a real edge for early adopters.
The fix: 30 minutes of editing the three files. The llms.txt template is open and easy.
Worked example: a typical Jupiter service business
Imagine scoring a typical Jupiter exterior cleaning business on the 5 axes. A realistic mid-tier profile for this vertical might look like this:
| Axis | Score | Common pattern |
|---|---|---|
| Machine-readable structure | 7/10 | Server-rendered, good semantic HTML, lazy-loaded images causing minor parse delays |
| Schema completeness | 3/10 | Generic Organization schema only. No LocalBusiness, no FAQPage, no Service schemas. |
| Q&A content density | 4/10 | Services pages are generic. No question-format content. Two blog posts in 18 months. |
| Cite-ability | 5/10 | Phone, address, hours present. Pricing absent. Warranty terms vague. Reviews not surfaced on site. |
| AI discoverability layer | 6/10 | robots.txt present and allows AI crawlers. Sitemap healthy. No llms.txt. |
Total: 25/50, the score that gets a business partially cited (showing up sometimes for branded queries, rarely for commercial queries). This is the most common profile we see on first scans for this vertical.
The 60-day fix plan:
- Add LocalBusiness, FAQPage, and 5 Service schemas (week 1)
- Rewrite the 5 service pages with Q&A blocks and concrete numbers (weeks 2-3)
- Publish 8 question-format blog posts answering the top phone-call questions (weeks 4-8)
- Ship llms.txt (week 1)
- Re-run the scan and verify the score moved (week 9)
Realistic post-fix score: 40/50, the score that gets you cited for most commercial queries in your service area.
What this means for the local pack vs. the AI layer
The local pack (covered in why the Maps local pack decides who gets the call) and the AI layer are not the same channel. They’re driven by overlapping but distinct signals, and they reward different patterns.
| Channel | Primary signals | Best practice |
|---|---|---|
| Local pack | Prominence (reviews, citations, profile activity), relevance, distance | Operate the Google Business Profile like a living asset |
| AI layer | Machine-readable structure, schema, Q&A content, cite-ability, discoverability files | Operate the site like an answer machine |
A business that wins both channels has a Google Business Profile that’s healthy and active AND a website that AI tools can read and cite. Most service businesses today win neither.
The cost of not being citable
When a homeowner asks ChatGPT for a roofer recommendation in Jupiter and you’re not in the response, you didn’t lose a click. You lost a recommendation. Click loss is measurable; recommendation loss is invisible until your phone stops ringing as much as it used to.
A 30% revenue stream that disappears silently is more dangerous than a 30% revenue stream that disappears loudly, because nobody on your team will flag it.
FAQ
Will optimizing for AI search hurt my regular Google ranking?
No. Every fix in this post (semantic HTML, schema, Q&A content, llms.txt) is either neutral or positive for traditional Google ranking. The five axes are reinforcing, not competing.
Do I need a different site for AI search vs. traditional SEO?
No. One site, optimized for both. The same page that ranks #3 in the local pack and gets cited by ChatGPT is the site you want.
How fast does AI search visibility change after a fix?
Faster than traditional SEO. Schema and llms.txt changes can show up in AI responses within days. New Q&A content typically gets cited within 2-4 weeks. The traditional SEO benefit of the same changes takes 30-60 days.
Skip the manual scoring
Our free 15-point Growth Scan includes the full 5-axis GEO scoring plus competitive AI visibility (we check whether ChatGPT, Google AI Overview, and Perplexity recommend you for your top 5 commercial queries). The deliverable is plain English with a prioritized fix list. No sales call required to receive the report.
Book your free Growth Scan at /scan
Related reading: why the Maps local pack decides who gets the call for the foundation of how local search works.