AEO for businesses with a service area

The short answer: A business that travels to its customers has deliberately removed the fact local systems anchor on, a street address, so the work is to replace that anchor with an explicit, machine-readable statement of where you go. Put the coverage in markup with areaServed, say the same boundary in plain prose on the page, and keep every directory copy of it identical. Then test whether an engine can place you, because placement fails long before quality does.

Plumbers, mobile vets, IT support firms, wedding photographers, pest control, home care, field service of any kind. The category is enormous and it has one shared structural problem that office-bound businesses never meet.

The anchor you were told to delete

Google Business Profile Help tells businesses that do not serve customers at their business address to remove that address from the profile. The same page caps coverage at 20 service areas and says the boundary of the overall area should not be more than about two hours of driving time from where the business is based. That is a sensible rule for a map product. It is awkward for everything downstream. Note the scope: this is one company's guidance for one product, not a rule other engines or directories have published.

Awkward, because the structured data specification pulls the other way. Google's LocalBusiness documentation, last revised on 8 September 2026, lists exactly two required properties, name and address, and describes address as the physical location of the business. Follow the profile advice and you have no address to publish. Follow the markup documentation and you publish an address you were told to hide.

Sites often resolve that tension by accident, and the accidental resolution is the one we run into repeatedly in audits of this category: a residential address in the footer, absent from the profile, present in two of six directories, with a slightly different suite number in the third. What we can observe when that is true is the output, not the reasoning. In the audit runs where we have seen contested addresses, the engines we test tended to answer with a hedge or with no named business rather than pick one address. We cannot see why, because we have found no published description, from any of the five engines we test, of how conflicting location claims are resolved.

areaServed is the property you are missing

Schema.org defines areaServed as the geographic area where a service or offered item is provided. It accepts four value types, AdministrativeArea, GeoShape, Place and plain Text, and it hangs off Organization, Service, Offer and ContactPoint. That is the vocabulary built for exactly your situation, and we rarely find it on the service-area sites we audit. We have not counted, so treat that as an impression from our own caseload rather than a figure.

Two things to understand before you add it. First, areaServed does not appear anywhere in Google's list of supported properties for the local business rich result, so a validator will not applaud and your agency will not report a win. Second, that is beside the point. A machine-readable sentence saying this company serves these fourteen postcodes is at least unambiguous, and it costs little to publish. Whether any given engine's retrieval layer reads your JSON-LD is not something we can verify from outside, and we have not found it documented. That is why we also ask for the same boundary in plain prose, which is readable either way.

The practical shape is a GeoShape with a GeoCircle, a centre point and a radius, or an AdministrativeArea list when your coverage follows council or district boundaries rather than a radius. Use the circle when you genuinely work outward from a depot. Use named areas when your coverage is a list of towns, because a circle drawn to include the furthest town also claims every town inside it, including the ones you refuse jobs in.

The contradiction test, and why it matters more here

Here is a check worth half an hour. Open a spreadsheet. In column A, list every public surface that states your coverage: the website footer, the areas-served page, the business profile, each trade directory, your invoices if they are public, your social profiles. In column B, paste the exact coverage claim from each. Now sort column B.

In the service-area audits we have run, the column rarely agrees with itself. The site says Greater Manchester. The profile lists nine towns. One directory, created by a staff member who left in 2023, still says nationwide. Each of those is a sentence a retrieval system can find and quote, and none of them wins outright. What we see in the answers is a claim qualified into uselessness, or a directory named instead of the business. What we cannot see is the ranking step that produced it, so read the directory explanation below as our reading of the evidence, not as a documented behaviour.

The complaint we hear most from this category is that the engine recommends a lead-generation marketplace instead of naming anyone. Our best explanation is not favouritism: the marketplace has a consistent, parseable statement of who covers what, and the business has six inconsistent ones. We have not tested that explanation against a control, so it remains an explanation. The same dynamic appears across sectors, which is why AEO by industry: where the work differs keeps returning to corroboration rather than copy.

The city page trap

The standard agency answer to geography is a page per town. Done well it works. Done at the usual volume it is a liability.

The tell is mechanical. Take two of your location pages, diff them, and count the words that differ. If the delta is the town name, a postcode and a stock photo, you have published the same page many times. A retrieval system that chunks documents into passages then has near-identical passages competing for the same query, and in the audits where we have checked which of a client's town pages gets cited, it is usually one page, not the set, and often not the one the client would have picked. We are describing a pattern in our own cases rather than a published retrieval rule.

Fewer pages, each carrying something only that page can carry, beats forty clones. Real job references. Travel time from your base. Local constraints a resident would recognise, a parking restriction, a permit office, a type of housing stock. The test is whether a reader in the next town over would notice the page is not about them. If they would not, merge the pages.

Resourcing this is its own question. In a five-person field service company, nobody owns the directory listings and nobody owns the markup, which is the small-business version of the problem described in Who owns AEO inside a company.

Reviews are your corroboration layer, and they are rented

For a regulated company the third-party layer is press, regulators and reference data. For a service-area business it is reviews and local directories, and almost all of it sits on platforms you do not control.

Notice what that means when an answer goes wrong. A fintech can get a wrong fact corrected at a reasonably authoritative source, which is the long game described in AEO for fintech. You often cannot, because the wrong fact is in a four-year-old review on a platform with no correction path. What you can do is out-publish it: make the current, correct statement of coverage abundant and consistent enough that it is the version a system keeps seeing.

Terminology gets in the way here too, since local marketers were sold the same work under three names in as many years. The distinctions in AEO vs GEO vs SEO: what the three terms actually mean are worth ten minutes before you buy anything labelled with any of them.

Running the test yourself

Pick six prompts a customer would actually type, in their words, not yours. Emergency phrasing for one. A town at the edge of your coverage for another. A town just outside it, because a false claim of coverage is as damaging as absence and it burns a real lead.

Run each one at least five times per engine, on different days. Record three fields: were you named, was the coverage claim attached to your name correct, and what did the engine cite. A single run tells you little, because in our own repeat runs the same prompt on the same day has returned different sources on the engines we test. The self-check is a longer version of this you can run in an afternoon without tooling.

Honest limits

This post establishes less than it may read like. We cannot show that areaServed markup changes how often a business is named, because in every case where we have added it, it shipped alongside the consistency work, and we have run no version with one and not the other. The claim that inconsistent coverage statements cause hedged answers is our reading of audits we ran, not a controlled test and not anything an engine documents, and the same goes for the directory explanation and the town-page pattern. Our evidence comes from client audits in a handful of markets, mostly UK and India, which is a sample and not a census, and local answer behaviour is the area where we see the widest variation between countries. The only claims here that rest on published sources are the Google Business Profile limits, the required properties for Google's local business rich result, and the value types schema.org lists for areaServed.

FAQ

Should I publish my home address if I work from home? Usually not on the business profile, since the guidance there is explicit. On the site you can state the city you operate from without a street line, which gives systems a centre point without putting your front door on a map.

Does areaServed get me a rich result? No. It is not in Google's list of supported properties for the local business rich result. It is vocabulary for describing an entity. We cannot verify from outside whether a given engine reads it, so pair it with the same boundary written in plain prose.

How many town pages should I have? As many as you can make genuinely different. If two pages differ only by place name, they are one page with a find and replace applied, and they compete with each other.

Why does the engine name a directory instead of my business? We do not know, and we have not found it documented. In the cases we have audited the directory stated coverage consistently and the business stated it several different ways, so reconciling every public copy of your coverage is the cheapest thing to try first.

Sources

All three were opened and checked on 12 October 2026.

• Google Business Profile Help, managing service areas: the instruction to remove your address if you do not serve customers there, the limit of 20 service areas, and the boundary of about two hours of driving time. Checked 12 October 2026: support.google.com/business/answer/9157481
• Google Search Central, local business structured data, page last updated 8 September 2026: address and name are the required properties, and areaServed appears nowhere in the supported property list. Checked 12 October 2026: developers.google.com/search/docs/appearance/structured-data/local-business
• Schema.org, the areaServed property, its expected value types and the GeoCircle example. Checked 12 October 2026: schema.org/areaServed