The Visibility Gap in Home Services SEO: Why "Near Me" Still Needs an Entity Behind It

Home services SEO now has two jobs, not one

A homeowner with water on the floor does not open a spreadsheet of contractors. They ask their phone who to call and act on what comes back. For years, what came back was a map pack: three local businesses, ranked, with reviews and a call button. Today the same query often returns a written answer above that pack, assembled by a generative engine, naming a handful of businesses inside a sentence. Two surfaces now answer the one question that has always driven this trade.

It is tempting to treat these as two separate marketing problems requiring two separate vendors. The evidence points the other way. The local pack and the AI answer are different presentation layers over a shared foundation: a structured representation of businesses as entities, not as strings of text. An engine that cannot confidently decide which real-world company your website, your Google profile, and your directory listings all refer to cannot rank you reliably in the pack, and it cannot name you reliably in the answer. The gap that used to cost one listing now quietly costs both.

This is why "home services SEO" is no longer a synonym for "ranking." It is the work of making one business legible to machines that increasingly reason over identities rather than keywords. The rest of this piece traces that entity layer from its origins to the two surfaces it now feeds, and states plainly where the evidence is solid and where the vertical-specific numbers do not yet exist.

Why "near me" local SEO always resolved to an entity

The phrase "near me" implies the whole problem is geometry: find the closest business and return it. Local ranking has never been that simple. Google's own account of local results describes them as a product of relevance, distance, and prominence together, and every one of those inputs presupposes a prior step the machine has to get right first. Before an engine can score how close or how prominent a business is, it has to decide which business this is, and whether the profile it found, the website it crawled, and the citation it read on a third-party directory are all the same company or three different ones.

For a storefront with one fixed address, that resolution is usually easy. For a home-services company it is genuinely hard, and that difficulty is the root of most local visibility gaps in the trade. A service-area contractor has no single storefront to anchor to. Crews travel across a dozen towns, the address on the truck differs from the address on the license which differs from the address a directory scraped years ago, and reviews pool at head office instead of spreading across the neighborhoods actually served. Each inconsistency gives the engine a reason to hedge. "Near me local SEO" is, underneath the geography, an identity problem: the businesses that win are the ones a machine can place without guessing.

The knowledge graph: what "entity based search" actually means

The shift from matching strings to understanding entities is not a reaction to generative AI. It is more than a decade old and was announced in plain language. In May 2012, Google introduced its Knowledge Graph under the explicit thesis of indexing "things, not strings," launching with 500 million entities and 3.5 billion facts about the relationships between them, assembled in part from sources such as Freebase, Wikipedia, and the CIA World Factbook. A search engine stopped treating "plumber" as a token to match on a page and started treating a specific plumbing company as a node with attributes, a location, a category, a set of relationships, and a reputation.

Entity based search is simply that: the engine's attempt to map the words on a page back to a real thing it already knows about, or to learn a new thing and connect it to the graph. For a local business, being "in the graph" as a confidently resolved entity is what makes the difference between being a candidate the engine can reason about and being an ambiguous mention it sets aside. This matters now more than it did in 2012 for a specific reason. The generative engines that write AI answers are built on retrieval-augmented generation, an architecture that pairs a language model with a live index queried at answer time. That retrieval step reaches into the same structured, entity-organized substrate that classic search has been building for years. The knowledge graph was never the destination. It was always the ground floor, and the AI answer is just a newer room built on top of it.

The sameAs problem: being the same business everywhere is harder than it looks

If entity resolution is the whole game, the obvious question is: how does a machine decide two records describe the same company? The mechanism the web offers is a declaration of identity, encoded in Linked Data as owl:sameAs and surfaced for businesses today as the sameAs property in schema.org and JSON-LD, where you list the canonical profiles that all point to the one entity. It sounds like a solved lookup. The foundational research says it is not.

In 2010, Halpin, Hayes, and colleagues examined how owl:sameAs was actually being used across the published Linked Data web and found it applied inconsistently, with publishers encoding at least four looser, non-equivalent senses of "same." Things that were genuinely identical were conflated with things that merely shared a context, a role, or a close-enough resemblance. The strict logical meaning of identity was, in practice, not what people were writing down. That looseness has never been fixed at the standard level; platforms work around it with their own heuristics, which means identity across the web is corroborated, not declared once and trusted forever.

A home-services company sits at the hard end of this problem. It is one business that must resolve cleanly across Google, Apple Maps, and Bing, plus Angi, Thumbtack, Yelp, Nextdoor, the Better Business Bureau, and the trade and licensing bodies, each with its own record of your name, address, and phone. This is why NAP consistency SEO, the unglamorous work of making name, address, and phone read identically everywhere, is not a checkbox chore but the practical core of entity clarity for a traveling contractor. Every source that agrees is a vote the engine can count. Every source that disagrees is a reason for it to hesitate, on both surfaces at once.

How local pack ranking and AI answers share one substrate

Once you see the entity layer, the relationship between the two surfaces stops looking like a coincidence and starts looking like plumbing. Local pack ranking runs on relevance, distance, and prominence, and all three read the corroborated entity: the categorized profile, the consistent facts, the reviews attributed to the right company. The AI answer runs on retrieval-augmented generation, which fetches candidate passages and structured facts at query time and composes them into a recommendation. Both start by resolving who you are. A business the engine cannot confidently identify is weak input to the ranker and weak input to the generator.

This is also the limit of the analogy. Classic ranking and AI citation are measurably decoupled, not identical: being retrieved and named in an answer is a different outcome from placing in the pack, and one does not guarantee the other. The controlled research on generative-engine visibility shows that the levers which get a source cited in a synthesized answer are content-level and corroboration-level, not raw rank. So the shared substrate is the entity, but the two surfaces read it through different mechanisms and must be measured as two distinct things rather than folded into a single blended number. That decoupling is precisely why a serious home-services program treats classic search and AI answers as separate pillars with separate readings.

What the click data says about the near-me moment

The stakes of being named in the AI answer, versus merely ranking below it, are not a matter of opinion. The strongest independent evidence comes from the Pew Research Center, which tracked the real browsing of 900 U.S. adults across 68,879 Google searches in March 2025, of which 12,593 contained an AI summary. When an AI summary was present, users clicked a traditional organic link in about 8 percent of searches, versus 15 percent when no summary appeared, and clicked a link inside the summary itself only about 1 percent of the time.

Read that against the "who do I call" moment. If the homeowner reads a summary that names three contractors and one of them is not you, the older comfort that you still rank on the page below carries far less weight than it used to, because the click to that page is roughly half as likely to happen. Being the entity the answer names, rather than a link beneath it, is the outcome that increasingly decides the job. This is a general finding across search, not a home-services-specific measurement. But the mechanism it describes, attention concentrating on the named answer, applies with full force to an urgent local decision made in seconds.

What changes the outcome, and what the evidence does not yet say

It would be easy to end here with confident vertical numbers, but the evidence does not support them yet. There is strong, peer-reviewed evidence for the general levers, and a real gap where the home-services-specific figures should be.

On the established side: the founding controlled study of generative-engine optimization tested roughly 10,000 queries and found that adding citations to credible sources, including direct quotations, and replacing vague claims with specific statistics produced a meaningful relative lift in a source's visibility inside generated answers, with citing authoritative sources the single strongest lever. Structured data is a durable foundation, not a trick: schema.org has been a jointly governed vocabulary since Google, Bing, Yahoo, and Yandex founded it in 2011, which is why service-area schema and sameAs markup are safe, standards-based investments rather than hype-cycle tactics. And a widely misread point worth correcting before anyone spends budget on it: E-E-A-T is a human-rater evaluation standard Google uses to assess its own results, not a machine-computed score a page can be tuned to hit directly. A great deal of "authority" advice quietly conflates the two.

On the emerging side: one large-scale 2025 study found AI answer engines systematically favor earned, third-party media over brand-owned pages, more so than classic Google does. That is a single study, not yet replicated, and we flag it as emerging rather than settled. And on the missing side: the vertical-specific numbers, how often independent home-services businesses actually appear in AI answers for their own emergency queries, and by how much a clean entity moves that appearance rate, are not yet established evidence. That data has to be measured, per business, against a frozen panel of real buyer questions, with the engine, locale, and date stamped on every reading. Anyone quoting you a precise home-services citation rate as settled fact is ahead of the evidence.

A rigorous read for a home-services business

The practical work follows directly from the science, and none of it depends on a promised rank. It depends on making one business unambiguous to machines that reason over entities, then measuring both surfaces carefully. In priority order for a service-area contractor:

• Resolve the entity first. Make name, address, and phone read identically across Google, Apple, and Bing and the contractor-specific directories, so the corroboration an engine needs to place you actually agrees with itself.
• Configure the profile as a service-area business, not a storefront: correct primary category, defined towns and neighborhoods, address handled appropriately for a company that travels rather than one a buyer visits.
• Encode identity where machines read it. Apply service-area schema on your site with sameAs links tying your verified profiles into one entity graph, accepting that corroboration, not a single declaration, is what earns trust.
• Spread real reviews across the crews and neighborhoods you serve, attributed correctly, from genuine customers only, because reviews are both a prominence signal for the pack and evidence a generative engine reads.
• Measure classic search and the AI answer as two separate pillars against a fixed panel of your real urgent buyer questions, sampling each engine many times and reporting appearance as a range with the engine, locale, and date on every reading.
• Treat any vertical-specific citation number as a measurement to take, not a promise to accept, and re-read it on a cadence, because the answer surface is volatile and non-deterministic.