Tarun Bulchandani
← All insights
by Tarun Bulchandani

Local discoverability in the era of AI search: considerations for businesses with a physical presence

AI assistants handle local queries differently from classical search engines, and the divergence is sharper than the literature has yet caught up to. Considerations for firms whose business depends on being found in their local market, and where the conventional 'local SEO' toolkit now sits.

The local-search market (the queries that drive footfall to a physical location or surface a service provider serving a specific town or region) has been served, for fifteen years, by a mature set of conventions. Google Business Profile (or its predecessors), local citations, NAP consistency, structured LocalBusiness markup, and a handful of platform-specific optimisations were the toolkit.

That toolkit isn't obsolete. Most of its components still matter. But it's no longer sufficient, and in a small number of places it has become structurally misaligned with how AI assistants handle local queries.

This note sets out where the divergence sits, what it means for firms whose business depends on local discovery, and how the practical workplan should adjust.

How local queries are now handled

A user asking ChatGPT, Claude, Perplexity or a Copilot variant a local query ("good chartered accountant in Hammersmith for a small consultancy", "family dentist near Highgate that takes new NHS patients", "wills and probate solicitor central Bristol") is processed differently from a user typing the same query into Google.

Three differences matter.

First, the assistant is more aggressive about parsing the intent of the query. A query that uses approximate location terms ("near", "central", "around"), or that mentions the user's other constraints obliquely, is more likely to be correctly resolved by an assistant than by a classical search engine, which tends to default to a literal-keyword match.

Second, the assistant is more selective about which sources it cites. A classical Google query returns a list; the assistant returns one to three recommendations. The candidate set is smaller, and the cost of being outside it is correspondingly higher.

Third, the assistant draws from a wider source pool than the classical local-search index. Google Business Profile data still matters, but the assistant is also pulling from LinkedIn, Companies House, professional body listings, the firm's own website, third-party review aggregators, local press coverage and (where applicable) Wikipedia. A firm that has invested heavily in Google Business Profile alone and neglected the broader sources is, in this layer of the market, materially worse off than the same investment would have made them five years ago.

Where the conventional toolkit still works

Most of the conventional local SEO discipline transfers across cleanly.

Google Business Profile remains foundational. The assistants pull from it directly. A complete profile, with current hours, accurate categories, recent photos and a substantive description, is still a baseline expectation.

NAP consistency still matters. Name, address and phone number must be consistent across the firm's website, its Google Business Profile, its Companies House listing, its LinkedIn page, professional body directories, and the local citation sources. Inconsistency is a corroboration failure in the framework of the six signals.

LocalBusiness schema is still the right markup type for firms with a single physical premises serving local clients. The markup should include the full structured address, hours, accepted payment types where relevant, and service area declarations.

Local citations on recognised directories (Yell, FreeIndex, Cylex in the UK; equivalents in other jurisdictions) still function as corroboration sources, although the relative weight of any individual directory is lower than it was a decade ago.

Reviews on the recognised platforms (Google Reviews, Trustpilot, sector-specific platforms like Doctify or Vouched For) feed the model's selection. Volume matters, but specificity matters more. A small number of detailed, substantive reviews is materially more useful than a large number of one-line ratings.

Where the conventional toolkit is now incomplete

Three categories of work weren't, ten years ago, part of the local SEO discipline but are now necessary.

Substantive content that demonstrates the firm's actual practice. A solicitor whose website lists "wills and probate" as a service line but has no underlying material on the specifics of probate in their jurisdiction is materially less likely to be recommended than a solicitor whose website has 2,000 words on the practicalities. The model is, in effect, asking the firm to demonstrate competence rather than declare it. The implications are substantial: a firm that treated its website as a digital business card now needs to treat it as a publishing platform.

A llms.txt that explicitly covers the firm's local positioning. The conventional llms.txt summary should mention the firm's primary location, the area served, and the specifics of its local practice. (See Understanding llms.txt.)

Principal-level visibility. The assistant's selection, particularly in professional services, draws on the principal's individual presence: LinkedIn, professional body profiles, any publications, any teaching, any expert-witness or commentary work. A firm whose principal has a strong individual presence outperforms a firm of equivalent size whose principal is invisible.

A note on multi-location firms

For firms operating from multiple locations, the structural question is whether each location is a distinct LocalBusiness entity (with its own URL, its own structured data, its own Google Business Profile) or whether the firm is operated as a single entity with multiple physical addresses. The right answer depends on the firm's structure, but the wrong answer (a single page listing all locations with no per-location URLs and no per-location markup) is, in my work, a persistent source of local-discoverability problems.

The recommended pattern: per-location pages, per-location structured data, per-location Google Business Profile linkage. The investment is higher, but the return is proportionate.

The practical workplan

For a firm whose business depends on local discoverability, the recommended sequence is:

  1. Audit the conventional baseline. Google Business Profile, NAP consistency, citations, LocalBusiness markup, reviews. The audit framework covers this, with the local-specific signals called out explicitly.
  2. Address the structural gaps. Schema.org markup, llms.txt, FAQ markup, principal Person markup. The Retrofit package covers this work.
  3. Invest in content depth. The firm's website should substantively demonstrate the practice. The Authority package covers the ongoing investment.
  4. Invest in principal-level visibility. The principal's LinkedIn, professional body profile, any commentary or teaching work, any contributions to industry publications. This is outside the scope of any single website engagement; it's the firm's ongoing positioning work.

The order matters. A firm that invests in content depth before addressing the structural gaps is publishing into a void that the model can't read.

What this means for the firm

The local-search market is still, in 2026, recognisable to practitioners who worked in it in 2016. The conventions are familiar; the tooling is mostly familiar; the underlying question of who appears when a prospect asks for a local recommendation is the same. The mechanism by which that recommendation is generated has shifted, and the firm's discoverability programme needs to shift with it. The shift is modest in cost relative to the cost of being missed at the moment a prospect is researching.