The complete guide to local SEO

Local SEO is the discipline of getting your business identified, indexed, and recommended by the systems people use to find local services. Google Search and Maps are the obvious surfaces. Apple Maps, Bing, and AI assistants like ChatGPT, Perplexity, Gemini, and Claude are the increasingly important secondary surfaces. The mechanics overlap with general SEO but add a layer of their own: local SEO is about your business as a structured record that search engines and assistants can trust, and about the web pages that turn high-intent local searches into enquiries. Both matter, and they are strongest together.

What local SEO actually is

General SEO is about ranking pages for queries. Local SEO adds a second job: being the business a retrieval system identifies, retrieves and recommends in response to a local-intent query. So there are two things to optimise, and they work together. One is the business entity: its structured properties (name, address, phone, category, opening hours, attributes, services, products, photos, reviews), which live across sources such as Google Business Profile, Bing Places, Apple Business (formerly Apple Business Connect), directory listings, your website's structured data, and increasingly Wikidata. The other is your web pages, which earn high-intent clicks from local searches and turn them into calls, bookings and visits.

We are describing behaviour here rather than architecture, because Google does not publish how local retrieval works internally. What it does publish shows both halves at work. The factors it names for local results describe the business, and it says directly that your chosen categories "affect your local ranking on Google".¹ It also counts how many websites link to your business as part of prominence,² and your website is where people land when they click through from your profile or a search result.

When the systems behind Google Search, Google Maps, AI assistants, or third-party directories receive a query with local intent, they consult these entity sources to identify candidates, rank them, and surface the answer. That answer might be a Map Pack of three results above a SERP, a list of pins on Maps, an AI-generated recommendation paragraph, or a single Knowledge Panel for an explicit brand query. All of them rest on the same underlying entity layer.

Local SEO is making sure the world's retrieval systems agree on who you are, where you operate, what you do, and how well you do it. Everything else follows from that.

The official Google ranking factors

Google has publicly stated, in its own help documentation, that local results in Search and Maps are decided by three factors: relevance, distance, and prominence.² Almost everything else you read about how much each factor weighs is inference or observation. Useful, but not the same thing. Treat anything more specific than these three as a hypothesis worth testing rather than a confirmed mechanic.

Worth noting what Google actually puts inside prominence: it says the factors include "how many websites link to your business and how many reviews you have".² Links and reviews are named. Directory listings, posts and photos are not.

Distance is bounded by physics. You cannot move your premises. You can decide which queries you compete for, and you can set your service area honestly to match where you actually operate, which Google bounds too: a profile's service area "shouldn't extend farther than about 2 hours of driving time from where your business is based".³ Relevance is the easiest factor to influence within days of editing your GBP. Prominence is the long game where most defensible advantage is built.

How much each factor weighs: a working view

Google does not publish weights for local ranking, and nobody outside Google can measure them directly. What you can do is combine what Google does document with what reliably moves rankings in practice, and put the levers in a sensible order. The view below is our editorial reading of where effort usually pays back first. It is a guide to priorities rather than a set of measured percentages, and it shifts as Google changes. Distance sits outside it: that is set by where you are and where the searcher is.

The local search surfaces: six places to be found

"Ranking" in local SEO is not one thing. A prospect might find you on at least six distinct surfaces, and they do not always show the same businesses. Each has its own way of choosing, its own click behaviour, and its own level of commercial intent.

Google documents one ranking model for local results, not one per surface.² The differences below are what we and other practitioners observe surface by surface, not weights Google publishes. Read them as working assumptions to test against your own grid data.

These surfaces compound. Strong Map Pack presence builds brand-search volume, which feeds the Knowledge Graph, which feeds Gemini and AI Overviews. Solid traditional SEO is the floor that lets an assistant find you at all: Google requires a page to be indexed and snippet-eligible before it can appear in AI Overviews or AI Mode,⁵ and OpenAI runs a dedicated crawler to decide what ChatGPT's search can surface.⁸ Optimising one surface in isolation tends to underperform optimising the system. See the Maps ranking guide and the AI search visibility (GEO) guide for the per-surface mechanics.

The GBP layer: why it dominates

Google Business Profile is the canonical entity record for most local businesses.³⁹It feeds the Map Pack directly, populates Knowledge Panels, supplies most of the structured local data that Google's Knowledge Graph holds,¹⁰ and is the place data Google's own AI answers draw on for local queries. Other assistants use their own sources (ChatGPT, for one, now licenses Yelp's reviews and listings), so do not assume your profile is what every assistant reads.

The fields Google has progressively added to GBP over the years (services, products, attributes, posts, social links, AI-generated descriptions) are the ones the algorithm has reason to read. Empty fields are missed signals. Most businesses fill the obvious ones (NAP, hours, a few photos) and stop. The depth and accuracy of the less-obvious fields is one of the main differentiators between listings that rank and listings that do not.

The full optimisation playbook for GBP, in priority order, is in the GBP optimisation guide. The single biggest field is the primary category, and this is a rare case where Google's own documentation is direct about it. Its category help page states plainly that "the categories you select affect your local ranking on Google".¹ If you change one field this month, change that one.

The website layer: where high-intent clicks land

Your website does work your profile cannot. It ranks in the organic results under the Map Pack, it is where people land when they click through from your profile or a search result ready to act, and its pages are among the sources AI answers read.¹¹ The profile gets you considered, and the page is often what turns that into a call or a booking. The mechanisms that matter:

Schema.org structured data

Schema.org markup is how you make your website's claims about your business machine-readable.¹² The minimum viable stack for a single-location business:

• LocalBusiness on your homepage (or a more specific subtype like Restaurant, Dentist, Plumber, FinancialService). Use the most specific subtype Schema.org defines for your sector. Subtypes inherit all parent properties and add their own.
• Full set of properties: name, image, address (broken into streetAddress, addressLocality, addressRegion, postalCode, addressCountry), telephone, geo (with latitude and longitude), openingHoursSpecification (one entry per day-of-week range), priceRange, paymentAccepted, areaServed.
• A sameAs array linking your domain to your GBP listing, your Wikidata Q-number if you have one, and your major social profiles. This is the explicit declaration that those URLs all describe the same entity.
• On service pages: Service entries with serviceType, provider referencing your LocalBusiness, and areaServed.
• On any FAQ section: FAQPage, but know what you are and are not buying. Google restricted FAQ rich results in August 2023 to "well-known, authoritative government and health websites",¹³ and has since withdrawn the feature altogether: its changelog records that "the FAQ rich result feature is no longer shown in Google Search results" and that the documentation for it has been removed.¹⁴ Any guide still promising you FAQ rich snippets is out of date. Markup keeps the question-and-answer structure machine-readable, which is worth something on its own, but Google states plainly that there is "no special schema.org structured data that you need to add" to appear in its AI features.⁵ Mark up FAQs because you genuinely publish them, not as an AI play.
• For multi-location businesses: Organization on the corporate domain, with each location's LocalBusiness node pointing back to it through parentOrganization (and subOrganization on the Organization if you want the link both ways). See the multi-location guide for the architecture.

Site architecture for local intent

For a single-location business, the site architecture is straightforward. For multi-location businesses, the URL structure becomes the single most important decision (covered in the multi-location guide).

For a single location, the patterns that work:

• Homepage: positions the entity, contains the canonical NAP, carries the LocalBusiness schema, embeds a Google Map.
• One service page per major service: each ranks for service-plus-location queries, carriesService schema, has location-specific copy, links back to the homepage and to other service pages where relevant.
• About / team page with Person schema for staff or experts where credentials matter (especially for YMYL sectors like medical, legal, financial).
• Contact page with the canonical NAP, the Google Map embed, and contact methods.

Page experience and Core Web Vitals

Google words this carefully, and the wording is worth copying. There is no single page-experience ranking signal: Google's documentation says its core ranking systems "look at a variety of signals that align with overall page experience". On the same page, it says Core Web Vitals "are used by our ranking systems".¹⁶ Those are Largest Contentful Paint, Interaction to Next Paint, and Cumulative Layout Shift.¹⁷ That is a weaker and more useful claim than "Core Web Vitals are a ranking factor", because it tells you speed work is a floor rather than a lever.

So run your homepage and top service pages through PageSpeed Insights. If any are failing Core Web Vitals, fix the worst one, then go back to the entity work. Do not spend a fortnight shaving milliseconds while your primary category is wrong.

The reviews layer

Reviews are simultaneously a ranking input, a click-through input, and a conversion input.¹⁸¹⁹ The full mechanics, scripts, and the operational playbook are in how to get more Google reviews. The summary for this guide:

• Count and velocity matter. A business accumulating reviews steadily over years signals genuine ongoing activity. A sudden spike or extended drought reads as suspicious or stale.
• Specific phrasing in review text helps, but it has to come from the customer: Google's policy bars asking for particular content.¹⁹ Asking soon after the job, while it is fresh, tends to produce reviews that describe it. Google does not document review text as a relevance input, so treat any ranking effect as a working hypothesis. What is not a hypothesis: a review describing the service someone else is about to search for is the most persuasive sentence on your listing.
• Response rate and time matter. Replies are public, evaluated for sentiment and recency, and influence the next reader.
• Recency matters. A 4.8-star average from reviews ten years old carries less weight than a 4.6-star average from active recent reviews.

The customer side is just as clear. People read reviews before they choose a local business, and they look at three things: how many there are, how recent they are, and the overall rating. Treat those as three separate gates. Clearing two of them is not enough.

Google points the same way on the ranking side: it names how many reviews you have, and says more reviews and positive ratings can help your local ranking.²

The citation layer

Citations are mentions of your business across the web, with or without a link. Google names links and reviews when it explains prominence, and does not name directory listings.² The practitioner consensus, which is what we are relying on here, is that their direct ranking weight has fallen over the past decade as entity resolution improved. They still matter for two reasons.

• Entity corroboration. When the same NAP appears across your company registry, GBP, Apple Business, the major directories in your country, and sector-specific directories , and your website, Google has multiple independent confirmations of your existence and identity. This strengthens entity resolution, which feeds everything else.
• Inconsistency damage. The reverse is also true: a single outdated phone number on a long-running directory can split your prominence signals across two perceived entities. Fixing inconsistency unlocks signals that were already there but were not converging.

The full directory list, audit process, and duplicate-handling playbook are in NAP consistency & local citations. The summary: do the top 15 to 25 directories well rather than chasing volume. Nobody has ever shown that raw listing count moves rankings, and the failure mode of chasing it is real, because every extra listing is another record that can drift out of date.

AI search and the entity convergence problem

AI search is a surface of its own, and the honest framing is that it takes clicks before it gives them. Pew Research Center tracked the actual browsing of 900 US adults and found that when a Google results page carried an AI summary, users clicked a search result 8% of the time, against 15% when no summary appeared, and that they very rarely clicked the sources the summary cited.⁶ That is the mechanism that makes this surface matter: being mentioned in the answer is increasingly the whole of the visibility, because the click may never come. The mechanics are different enough to warrant their own guide. Briefly:

• Every assistant runs a retrieval step followed by a synthesis step that picks which entities to mention and which sources to cite. What each retrieval step reads is only partly documented, and it has changed more than once: Google grounds its AI answers in its own Search and Maps data, OpenAI says ChatGPT search uses third-party search providers alongside partner data such as Yelp's reviews and listings, and Claude's web search runs on Brave Search. Treat anything more specific about who is powered by whom as unconfirmed, and measure each engine separately instead.
• Your visibility depends on entity convergence: your Knowledge Graph node, Wikidata entry (if any), GBP record, and domain all referencing the same entity, with explicit sameAs links between them.
• Retrieval crawlers need access, and each vendor runs several agents that do different jobs. OpenAI separates GPTBot (training) from OAI-SearchBot, which exists to "surface websites in search results in ChatGPT's search features", and ChatGPT-User, which fetches a page a user asked about.⁸ Anthropic separates ClaudeBot (training) from Claude-SearchBot, which "navigates the web to improve search result quality", and Claude-User.²⁰ Perplexity runs no training crawler: PerplexityBot surfaces sites in its search results, and Perplexity-User fetches pages when someone asks a question.²¹ Blocking a training agent does nothing to a search agent, and the reverse. Robots.txt is not a hard switch for user-triggered fetches either: OpenAI says its rules may not apply to ChatGPT-User, and Perplexity says Perplexity-User generally ignores them. Audit robots.txt against the current list rather than a remembered one.

Full detail in the AI search visibility (GEO) guide. The point for this guide: traditional local SEO foundations are the floor. Without them, GEO efforts struggle.

Measuring local SEO honestly

Local rankings are personalised by location, device, and search history. Checking your rank from a single point gives you a flattering, mostly useless number.

Geo-grid ranking

The honest measurement is geo-grid ranking: defining a grid of points across your service area, checking your rank from each point for a given keyword, and plotting the result on a map. The boundary between where you win and where you lose tells you where to invest. The change in that boundary over months tells you whether your work is landing.

That competitor list under each pin is where the diagnosis usually is. If the same rival tops every point you lose, you have a prominence problem against a specific business and can go and read their listing. If the winners change from pin to pin and each one is simply nearer, you are looking at distance, and the honest answer may be that the query is not winnable from where you are.

Geo-grid resolution varies by tool, typically somewhere between 5×5 and 21×21, with the higher resolutions giving more accurate boundary detection at the cost of more API calls and higher pricing. For most single-location businesses, a 7×7 or 9×9 grid run monthly on the top five commercial keywords is sufficient.

Other metrics worth tracking

• GBP profile views, calls, direction requests, website clicks (from the profile's Performance report or via the Business Profile Performance API)
• Review velocity (new reviews per week or month) and average rating
• Brand-search volume in Google Search Console (queries containing your business name)
• Localised organic ranking for non-Map Pack keywords (your service-page rankings)
• Citation accuracy across the top 15 to 25 directories (annual or quarterly audit)
• AI visibility for a defined query set across ChatGPT, Gemini, Perplexity, Claude (monthly)
• Conversion data tied back to GBP and search source (calls answered, leads booked, customers served)

The 90-day starting plan

For a business starting from a low or unmaintained baseline, this is the order of work that produces the most visible movement in the shortest realistic time.

1. 1

   Days 1 to 14: Claim, verify, and complete GBP

   Search Google for your business. Click "Own this business?" if you do not already have access. Verify: Google offers phone or text, email, live video call, mail, and instant verification via Search Console, but note that the options "available to you depend on your business type, public info, region, or business hours" and are "automatically determined by Google and can't be changed", so you get what you are given.⁹ Add a backup owner. Then fill every field: primary category, a small number of genuinely accurate secondary categories, 10 to 30 services, every applicable attribute, 20+ photos, a description using the full 750 characters allowed,²³ opening hours, special hours for the next 12 months of public holidays. Google retired the profile Q&A box in late 2025 and now answers customer questions itself from these fields, so the services and attributes lists are where that work goes.

2. 2

   Days 14 to 21: Audit citations on the top 15 to 25 directories

   Decide your canonical NAP format. Visit each major directory (Bing Places, Apple Business, Yell, Yelp, Facebook, FreeIndex, Tripadvisor or Trustpilot, local council registries). Fix any inconsistencies. Hunt for duplicates and request removal where they exist. Confirm your company registry record, where it is publicly indexed (Companies House, a Secretary of State filing, Corporations Canada, ASIC/ABN Lookup) matches.

3. 3

   Days 21 to 30: Wire up reviews

   Generate your direct GBP review link. Add it to receipts, post-service SMS messages, post-purchase emails, in-store QR codes. Train staff on when to ask. Set a 48-hour SLA on review replies. Document the brand voice for those replies.

4. 4

   Days 30 to 45: Technical website work

   Add LocalBusiness JSON-LD to your homepage. Add Service schema to each major service page. Add FAQPage schema to FAQ sections you actually publish, expecting machine-readability rather than a rich result. Validate everything, because invalid markup fails without telling you.¹⁵ Audit Core Web Vitals on the homepage and top service page. Fix the worst issue.

5. 5

   Days 45 to 60: Entity convergence

   Add sameAs to your homepage Organization or LocalBusiness schema. Wikidata's notability bar is not press coverage, which is the common misreading: an item qualifies if it is "an instance of a clearly identifiable conceptual or material entity that can be described using serious and publicly available references", or if it fills a structural need.²⁴ A registered company with a public filed record often clears that. If yours does, create or claim a Q-number and link it back via sameAs. Audit robots.txt for AI crawlers per the GEO guide.

6. 6

   Days 60 to 90: Measure and iterate

   Run your first geo-grid scan on five commercial keywords. Note where you rank from each grid point and where the boundary sits. Re-run monthly. Identify gaps where competitors appear and you do not, and target the underlying signal. By day 90 you should be seeing measurable boundary movement on at least two of your five keywords.

Common patterns we see go wrong

Quick-reference quarterly checklist

Run through this once a quarter. If any line is "no", that is a leak.

• GBP is verified and has a backup owner who is not an ex-employee
• Primary category is the most specific accurate match Google offers for your sector
• Services and products are populated with current offerings
• Every applicable attribute is ticked
• Photos uploaded in the last 30 days, including at least one new exterior or interior shot
• Description uses the full 750 characters and reflects current positioning
• Top 15 to 25 citations have matching NAP, with no known duplicates
• Review velocity averages at least a few new reviews per month, depending on customer volume
• Every review in the last 90 days has a reply within 48 hours of being posted
• LocalBusiness schema is on the homepage and validates cleanly
• sameAs array is on the schema and links to GBP, social profiles, and Wikidata if applicable
• robots.txt has been audited for AI crawler access
• Core Web Vitals on the homepage and top service page pass
• Geo-grid scans on top 5 keywords have been run in the last 30 days
• GBP Performance has been reviewed for trend changes (sudden drop in calls, direction requests, website clicks or profile views)

Where to go next

Each layer above has its own deep guide. Start with whichever is most broken in your business right now.

And when you are ready to measure what is actually working, that is what SearchOps is built for.