Answer Engine Optimization for local businesses

A customer used to search “roofers near me,” scan a page of blue links, and click two or three. Now a meaningful share of those same questions get asked in a chat window, or get answered above the links in an AI Overview, and the customer never scrolls.

They get a paragraph of prose and a small set of citations. Maybe three businesses named. Maybe a link to a directory page and a Reddit thread.

If you are not one of those citations, you did not lose a ranking position. You were not in the consideration set at all. There is no page two to be on.

That is the whole shift, and it is why “answer engine optimization” became a term. Most of what gets sold under that name is repackaged SEO with a new invoice line. Some of it is real. This post is our attempt to separate the two honestly, because we run this work for local businesses and we would rather explain the actual mechanics than sell a package.

What actually changed

Discovery used to be a two-step: the engine returned a list, the human evaluated the list. Now a chunk of that evaluation happens inside the model before you ever see anything.

Google AI Overviews, ChatGPT with search, Perplexity, and Claude all work roughly the same way for a local question. The system runs one or more retrieval passes, pulls back passages from whatever sources it can reach, and synthesizes an answer with citations attached to the passages it leaned on.

Two consequences fall out of that, and they matter more than any tactic:

• The unit of retrieval is a passage, not a page. The system is not ranking your homepage. It is pulling a specific chunk of text that appears to answer the question and quoting or paraphrasing it.
• The citation set is small and often not your website. For local queries the sources are frequently a directory listing, a review site, a local news roundup, or a forum thread, with the business’s own site appearing as one entry among several — or not at all.

That second point is the uncomfortable one for anyone who spent a decade optimizing their own domain.

Why classic local SEO is not obsolete

Here is where a lot of AEO marketing goes wrong: it implies your existing local SEO work is legacy overhead.

It is not. When you ask an assistant for a recommendation in a specific city, the substrate it reasons over is largely the same substrate local search always ran on:

• Google Business Profile. Category, service list, hours, service area, photos, and Q&A. This is the single densest structured description of a local business that exists, and it feeds Google’s own AI surfaces directly.
• NAP consistency. Name, address, and phone, identical across every place they appear.
• Reviews — volume, recency, and text. Assistants quote review sentiment constantly. Review bodies are prose that mentions your services, your neighborhood, and your staff by name, which makes them extremely retrievable.
• Structured data on your site. Machine-readable claims about what you are and what you sell.

A business with a stale Google Business Profile, four reviews from 2021, and inconsistent phone numbers across directories is not going to get cited, no matter what it does to its blog. The fundamentals are table stakes now, not maintenance work.

AEO vs SEO: what is genuinely different

The row that costs the most to fix is content shape.

A page built for dwell time buries the answer. Intro, backstory, credibility section, and the actual answer at word 800 wrapped in three paragraphs of context. That structure was rewarded for years. It works against you now, because the retrieval pass wants a chunk that stands on its own.

Practical version of the fix: for every question a customer actually asks, there should be a heading that is that question, followed immediately by a two-to-four sentence answer that would still make sense if someone pasted it into a text message with no other context. Then go deep underneath it. You do not have to choose between the two — you have to lead with the answer.

Include the specifics that make a passage worth quoting: the city, the price range, the timeline, the brands you carry, the conditions where the answer changes. “Most residential re-roofs on the North Shore run three to five days weather permitting” is quotable. “Timelines vary based on a number of factors” is not.

Structured data, and what it does not do

Schema.org markup gives a machine an unambiguous read on who you are, what you sell, where, and for how much. That is the entire job. It is disambiguation, not a ranking lever.

The types that matter for most local businesses:

• LocalBusiness (or a subtype like RoofingContractor, Restaurant, AutoDealer) — identity, address, hours, service area.
• Service — each distinct thing you sell, with an area served.
• FAQPage — question-and-answer pairs, which happen to map almost perfectly onto how retrieval wants content shaped.
• Product or Offer — when you have real inventory or fixed-price packages.

A minimal, honest example:

{
  "@context": "https://schema.org",
  "@type": "RoofingContractor",
  "name": "Example Roofing",
  "telephone": "+1-985-555-0142",
  "address": {
    "@type": "PostalAddress",
    "streetAddress": "100 Main St",
    "addressLocality": "Mandeville",
    "addressRegion": "LA",
    "postalCode": "70448"
  },
  "areaServed": ["Mandeville", "Covington", "Slidell"],
  "sameAs": [
    "https://www.google.com/maps/place/...",
    "https://www.facebook.com/..."
  ]
}

Two things people get wrong. First, they mark up claims the page does not support, which is a guidelines violation and does nothing for you. Second, they treat schema as the whole strategy. Adding FAQPage markup to a page whose answers are vague does not make the answers less vague.

The sameAs array is quietly the most useful part for AEO. It is you telling a machine that the Google Maps listing, the Facebook page, and this website are one entity.

Entity consistency is the boring thing that decides it

An AI system answering “who should I call for a UTV service in this area” is reconciling several sources that may disagree. Your site says one phone number, an old directory says another, your Google Business Profile lists hours nobody updated after you extended Saturdays.

Inconsistency reads as low confidence. Low confidence means the system reaches for a source it trusts more — usually a directory or a review platform.

The audit is unglamorous and mostly a spreadsheet:

1. Pick the canonical version of your name, address, phone, and hours. Write it down. One version, exactly, including whether it is “St” or “Street.”
2. List every place your business appears: Google, Apple Maps, Bing Places, Yelp, industry directories, chambers, supplier locators, old sites you forgot about.
3. Fix or remove every mismatch. Duplicate listings are worse than missing ones.
4. Make your service list identical everywhere. If your site sells six services and your profile lists three, you are invisible for three of them.
5. Re-check quarterly, because aggregators reintroduce old data on their own schedule.

Third-party surface area matters more than another blog post

This is the strategic reframe most local businesses miss.

For a large share of local recommendation queries, the sources cited are not business websites. They are directory pages, review platforms, “best X in Y” roundups, local publications, and forum threads where a real person answered the same question.

You cannot control those sources. You can influence your presence in them:

• Reviews, continuously. Not a one-time push to 50. A steady flow with recent dates, on Google first and then the platforms specific to your industry. Ask customers to mention what they had done and where — that language is what gets retrieved.
• Complete directory profiles, including the vertical-specific ones your industry actually uses. Complete beats numerous.
• Real local mentions. Sponsorships, chamber listings, supplier and manufacturer locator pages, local news. A manufacturer’s dealer locator carries more weight than a self-published page saying the same thing.
• Answer questions where they are asked. If people ask about your service category in local forums or community groups, a genuinely useful answer from an identifiable local business is both good marketing and retrievable text on a domain with authority you will never have.

We have watched an assistant recommend a powersports dealer in Louisiana almost entirely on the strength of review text and a manufacturer locator page, while barely touching the dealer’s own site. That is the normal case, not the exception.

How to measure this without lying to yourself

There is no rank tracker for “what ChatGPT said.” Anyone selling you an AEO rank report is selling you a sample of one, dressed up.

What actually works is three imperfect signals, run together:

Manual prompt testing on a schedule. Write down the ten to twenty questions your real customers ask before they buy — in their words, with the city in them. Run them monthly across the major assistants, logged out, in a fresh session. Record whether you were mentioned, who else was, and which sources got cited. The competitor and source lists are the valuable part. Do it the same way every time or you are comparing noise.

Referral traffic from AI sources. Assistants send real clicks and they show up in analytics with identifiable referrers. Volume is usually small compared to organic. Watch the trend and the landing pages, not the absolute number.

Search Console impressions. AI Overviews impressions land in your normal Search Console data. You will often see impressions holding steady while clicks soften on informational queries — that is the answer being consumed without a click, and it is a signal worth watching even though it is not attributable cleanly.

None of these are precise. Say so out loud rather than building a dashboard that implies a precision that does not exist.

What nobody can promise you

We would rather lose the sale than pretend here.

• No one can guarantee a citation in ChatGPT, an AI Overview, or Perplexity. There is no submission process, no index you can pay into, and no support ticket. Any guarantee is either ignorance or a lie.
• “AEO packages” that promise rankings in AI assistants are selling something they do not control. There is no ranking to sell. There is a retrieval and synthesis process that changes without notice.
• The ground truth changes every few months. Model updates, retrieval changes, and interface changes have repeatedly moved what gets cited. A tactic that worked in the spring may be irrelevant by fall. Anyone presenting AEO as a settled discipline has not been watching it.
• There is no shortcut around being a real business. Every durable input here — reviews, accurate information, clear answers, genuine local presence — is the same input that made a business findable before any of this existed. That is inconvenient for people selling a new service line, and it is the actual answer.

The honest summary: do local SEO properly, restructure your content so the answers stand alone, mark up your entity so machines are not guessing, build presence on the third-party sources that get cited, and check your real buying questions across the assistants every month. That is the whole program. Everything beyond it is speculation, and it should be priced as speculation.

We do this work for local businesses across Louisiana and the Gulf South — local SEO, structured data, content restructuring, and the monthly prompt testing that tells you whether any of it is landing. If you want a straight read on where you currently stand, let’s talk.