GBP setup (6)
Google Business Profile completeness, services, photos, categories, posts, reviews.
Field guide
The Local-Business AEO and GEO Checklist.
24 checks a multi-location HVAC, dental, salon, medical, or retail operation should run this quarter to be cited by ChatGPT, Perplexity, and Google AI Overviews.
By Marcus Chen, GEO and SEO lead at GetDigitize · Published September 20, 2025 · Updated April 18, 2026 · 3,800 words
The six groups
24 checks, grouped by discipline.
Schema markup (5)
LocalBusiness, Service, FAQPage, Review aggregate, BreadcrumbList.
llms.txt (3)
Publish, link, keep it readable. The most underused move in local AEO.
Content for AI Overviews (5)
Answer-first openings, plain headers, numeric facts, comparisons, pricing.
Citation cleanup (3)
NAP consistency, duplicate removal, regional citations per location.
Tracking (2)
Google Search Console AI-Overview impressions and Perplexity and ChatGPT citation tracking.
Group 01 · GBP setup
The Google Business Profile is the foundation.
AI Overviews for local queries pull from GBP data more than any other source. If your GBP is 70 percent complete, the AI Overview will not recommend you. Get it to 100 percent.
Complete every GBP field.
Hours, attributes (appointment required, online booking, accessibility), service area, business description. Every empty field is a lost signal.
Add all services with plain-text descriptions.
Not a one-line label. Each service gets a 2-to-3 sentence description that includes the service name, a price range if you can, and what is and is not included.
Upload 30 real photos.
Interior, exterior, team, work in progress, before-and-after. Original photos beat stock. AI Overviews sometimes surface the photo.
Claim all secondary categories.
An HVAC company is also "air conditioning contractor" and "heating contractor." A dentist is also "cosmetic dentist" and "pediatric dentist" if true. Do not over-claim; claim accurately.
Post weekly updates.
One Google Post per week. A promotion, a new-hire announcement, an answered question. Keeps the profile active in Google's freshness signal.
Collect and answer reviews.
Aim for 40 reviews per location. Answer every one within 24 hours. Negative reviews answered well are a positive signal; ignored reviews are a negative one.
Group 02 · Schema markup
Make your facts machine-readable.
Schema is the JSON-LD script block in your site head. AI crawlers parse it before they parse your prose. Five schema types do most of the work for a local business.
Add LocalBusiness schema.
One per location. Include name, address, phone, hours, geo coordinates, and a priceRange. For chains, use the branch-level schema, not one site-wide script.
Add Service schema for each service.
Each service page has its own Service schema with name, description, areaServed, and provider. Price is optional but helpful.
Add FAQPage schema.
FAQ blocks with Question and Answer markup get cited directly in AI Overviews. Five to eight FAQs per service page is the right volume.
Add Review aggregate schema.
If you have aggregate ratings from Google, pull them into AggregateRating schema on your own site. Do not invent ratings.
Add BreadcrumbList schema.
Every non-home page. This helps AI tools understand where a page sits in your site and whether it is the authoritative page for that topic.
Group 03 · llms.txt
Publish a llms.txt and link it from your footer.
llms.txt is a plain-text file at the root of your domain that tells AI crawlers what you do, where, for whom, at what price. It is the single highest-impact move we see for local operators.
Publish llms.txt at /llms.txt.
Plain text. Structured with H1 and H2 markdown headers. Sections: company summary, services, locations, pricing, FAQs, press mentions, primary contact.
Link llms.txt from your footer.
A text link in the footer ("llms.txt") tells crawlers the file exists and gives them the right fetch path. Without the link, many crawlers skip it.
Keep llms.txt under 300 lines.
Longer files get truncated. Be concise. If you have 20 locations, list them as a table, not 20 full paragraphs.
Group 04 · Content for AI Overviews
Write pages that read like answers.
AI Overviews pull from the first 150 words of a page. If your page opens with a marketing slogan, it will not cite you. Open with the answer.
Write direct-answer opening paragraphs.
First paragraph of every service page answers the query directly. Example: "AC repair in Houston costs between $150 and $650 depending on the part and the refrigerant type. Here is how we price it."
Use plain-text headers.
"AC repair cost in Houston" beats "What to expect from our HVAC excellence." Headers are what the AI indexes as topics.
Add numeric facts to every service page.
Price ranges, response times, warranty lengths, coverage areas. Facts beat adjectives. Three hard numbers per page minimum.
Publish comparison pages.
"Tankless vs. tank water heater." "Invisalign vs. traditional braces." Comparison queries drive 20 percent of AI Overview traffic for local verticals.
Publish a pricing page with real ranges.
"Call for pricing" is not AI-citable. Even wide ranges ($150 to $800, with factors) outperform hidden prices. Every pricing page lifts citation rate.
Group 05 · Citation cleanup
Get your NAP consistent across the open web.
AI tools cross-reference. If your phone number is (713) 555-4141 on Yelp and 713-555-4141 on BBB and 7135554141 on your site, the algorithm is not sure these are the same business.
Audit NAP consistency across directories.
Pull your listings from Yelp, BBB, Angi, HomeAdvisor, the Yellow Pages, and your regional Chamber of Commerce. Name, address, and phone must match exactly, including punctuation.
Remove duplicate listings.
Old addresses, previous business names, and accidental duplicates split your authority. Claim them and mark as duplicates.
Build five regional citations per location.
Business Journal profiles, Chamber of Commerce pages, local nonprofit sponsor listings, regional trade association memberships, neighborhood association pages. Five per location is the floor.
Group 06 · Tracking
Measure what you changed.
Google Search Console now reports AI-Overview impressions as a separate filter. Perplexity and ChatGPT do not report directly, but you can track citations with sample queries.
Track AI-Overview impressions.
In GSC, filter by "Search appearance: AI Overviews." Watch the trend line, not the week. Expect noise for the first 30 days.
Track Perplexity and ChatGPT citations.
Pick 20 queries your buyers use. Run them monthly in both tools, incognito. Log whether your brand was cited. A simple Notion table works. Target: 40 percent citation rate within 90 days.
Author
Who wrote this.
Marcus Chen is the GEO and SEO lead at GetDigitize. He built local-search programs for two D2C brands now over $40M ARR and runs AI-Overview citation tracking as a core KPI for every local account. He has been wrong about llms.txt twice, both times before the major crawlers announced support. He is less wrong about it now.
FAQ
Five questions operators ask.
What is AEO versus SEO?
SEO optimizes for ranking a page on a search engine. AEO, Answer Engine Optimization, optimizes for getting your facts quoted inside an AI-generated answer. The jobs overlap but the ranking signals are different. AEO rewards clear facts, schema, and citation-ready text.
How long until this starts working?
GBP and schema changes show up inside two weeks. Content changes and citation cleanup take 30 to 90 days. AI-Overview citation trails the site work by another 30 days because the crawlers re-index on their own schedule.
Do I need a developer?
For schema and llms.txt, yes. Or a CMS that supports both natively. The GBP, content, and citation work does not need a developer. A founder and a marketing lead can run it.
Can I skip llms.txt?
You can, but do not. llms.txt is the lowest-effort, highest-impact change we see. A 200-line llms.txt with your services, pricing, locations, and FAQs lifts AI-citation rate by 20 to 40 percent within a month.
Is AEO ethical?
Publishing accurate facts in a machine-readable format is the ethical baseline. What is not ethical is keyword stuffing, fake reviews, or LLM-gaming via prompt injection. None of the 24 checks here require any of that.