AEO, GEO and SEO for Local Businesses: How AI Assistants Decide Who to Recommend

Definitions: SEO vs AEO vs GEO

The vocabulary is messy, so here is how we use it.

For a local business the questions are concrete: "best dentist near me that takes my insurance," "who fixes tankless water heaters in Racine," "is there a good vegetarian taco place open late." Each is answered by a system that retrieves information and then writes a short list. Your job is to be retrievable, matchable and believable.

What does not change: eligibility

Google is explicit that its AI features run on the same foundation as Search. Its guidance for site owners says there are no additional requirements to appear in AI Overviews or AI Mode, and that no special optimizations are needed. To be eligible, a page must be indexed and eligible to show a snippet. It adds that standard practices still apply: let crawlers in, link pages clearly, provide text content, keep structured data matching visible content, and keep your Business Profile current.

That is good news and a warning. Good, because you do not need a secret file. A warning, because anyone selling a "GEO hack" that skips fundamentals is selling hope. Start here:

• Pages return 200, are not blocked by robots.txt, and are not marked noindex by accident.
• Every important page is in your XML sitemap and linked from your site.
• The page's main content is real HTML text, not text baked into images or built only after a click.
• Title, headings and the first paragraph say plainly what the page is about, with the city or service area where it applies.

A working model of how assistants choose a local business

Label: working hypothesis. Vendors do not publish a ranking recipe for local recommendations. The model below is consistent with their public documentation and with how retrieval-based assistants are built, but it is our practical framing, not a quote from any vendor.

1. Interpret the question into a category, a place and constraints (open late, kid friendly, takes my insurance).
2. Retrieve candidates from web search results, business listing or map data, and pages the assistant can fetch.
3. Check evidence for each candidate: do several independent sources agree on what it is, where it is, and that people like it?
4. Write a short list and often cite the sources it leaned on.

Two consequences follow. A business with thin, inconsistent or contradictory public information is hard to match at steps 2 and 3. And because the answer is generated, the same prompt can return different lists on different days. You are improving odds, not buying a slot.

The six signals to build

1. Entity clarity: one business, described the same way everywhere

Your name, address and phone number (NAP), primary category, hours and service area should match across your website, Google Business Profile, Apple and Bing business listings, and major directories. Pick one canonical version, then audit for old addresses, tracking numbers and name variants such as "Joe's Pizza" versus "Joe's Pizzeria LLC." Inconsistency is the easiest problem to fix and one of the most common.

2. Reviews: volume, recency and specifics

Reviews are the public record of what customers say about you, in their words. Recent, specific reviews ("fixed our furnace the same night, explained the quote before starting") give a system concrete language to match against a question. A steady flow beats a spike. Owner replies add signal and show you are active. How you collect them matters legally and for trust, which is why we wrote a separate guide to doing it the right way.

3. Answer-ready content

For each service and location, write a page that answers the questions customers actually ask, with the answer first and the detail after it. Use headings phrased as questions, a direct answer in the first one or two sentences, then specifics: prices or price ranges, timelines, what is included, who it is for, service area, hours. Concrete beats clever. If a human skimming would find the answer in five seconds, so can a machine.

4. Structured data that matches the page

Mark up what is visibly on the page: LocalBusiness (or a more specific type) with name, address, phone, hours, geo coordinates and sameAs links to your official profiles; Article for posts; BreadcrumbList; FAQPage where you really have Q&A. Keep it accurate, since mismatches between markup and visible text work against you. Do not expect visual rich results from FAQ markup in classic Google results, because Google has sharply limited them. The value is clear, machine-readable structure. Also avoid marking up reviews of your own business on your own site as star-rating rich results, since Google treats that kind of self-serving review markup as ineligible.

5. Corroboration from other places

Assistants trust claims that more than one independent source supports. Earn mentions where real customers and locals look: your chamber of commerce, industry associations, local news, sponsorships, supplier or partner pages, and reputable local directories. Quality and relevance beat quantity, and purchased links or spammy directories do harm.

6. Crawl access

If a crawler cannot fetch your pages, it cannot cite them. See the next section.

Crawler access: who to allow and why

AI companies run several different crawlers with different purposes, controlled separately through robots.txt. Here is what OpenAI and Anthropic document:

Google and others have equivalents. Perplexity documents PerplexityBot for its search index, and Google offers the Google-Extended robots.txt token to control use of content for its Gemini models without affecting Search inclusion. Check each vendor's current page before relying on a name.

Our recommendation for local businesses that want to be recommended: allow the search and user-request crawlers. Whether to allow the training crawlers is your call. We allow them on reviewkathy.com, and you can see our robots.txt at reviewkathy.com/robots.txt. Two cautions: do not block by IP address, because that can stop crawlers from reading your robots.txt at all, and check that your host or CDN is not blocking AI bots by default. Cloudflare, for example, offers a one-click AI bot blocking setting.

Do you need llms.txt?

llms.txt is a proposed convention: a plain Markdown file at /llms.txt that summarizes your site and links to your best pages for language models. It is cheap to publish, and we publish one at reviewkathy.com/llms.txt. But be realistic. Google says no AI-specific files are needed for its features, and we are not aware of a major assistant confirming it relies on llms.txt. Treat it as a tidy index, not a ranking lever. Put your effort into the six signals first.

How to measure it

1. Build a prompt panel. Write 20 to 30 realistic questions customers ask, covering your categories, neighborhoods and comparison phrasing.
2. Run it monthly across the assistants your customers use, with a clean session each time. Record whether you are named, in what position, what is said about you, and which sources are cited.
3. Repeat runs. Answers vary, so run each prompt several times and report a rate, not a single result.
4. Track referrals. Look for visits from chatgpt.com, perplexity.ai, claude.ai, gemini.google.com and copilot.microsoft.com in your analytics. On this site, if you accept cookies, we count those as AI referrals.
5. Watch the inputs. Review count and recency, your rating, response time to reviews, Business Profile views and calls, and branded search.

A 30-day plan

What to avoid

• Fake or purchased reviews, or conditional incentives. The FTC's rule on fake reviews and testimonials bans fake reviews, including AI-generated ones, and bans offering incentives conditioned on a particular sentiment. It also covers review suppression through groundless legal threats or intimidation.
• Review gating, which means steering only happy customers to public review sites.
• Mass-produced AI pages that say the same thing for 200 towns. They add no evidence and can hurt quality signals.
• Hidden text or stuffed keywords. Systems built to read language read through it.
• Promising guaranteed AI rankings. Anyone who does is guessing or selling.

Frequently asked questions

Sources and further reading

• Google Search Central: AI features and your website
• OpenAI: Overview of OpenAI crawlers
• Anthropic: Does Anthropic crawl data from the web, and how can site owners block the crawler?
• FTC: Final rule banning fake reviews and testimonials (press release, Aug 2024)
• IndexNow documentation

Honest limits. Nobody outside the AI companies and Google can see exactly how recommendations are ranked, and answers change from run to run. This guide separates what the vendors document from what is practical experience. We do not guarantee any ranking, review count or AI mention. Check the linked sources for current wording, because policies and crawler names change.