The Google Knowledge Graph is Google’s massive structured database of entities (people, places, organizations, concepts, products) and the relationships that connect them. Instead of treating the web as a pile of keywords and text strings, it lets Google understand that “Toronto” is a city in Ontario, that Ontario is a province in Canada, and that a search for “best dentist in Toronto” is really a query about a profession tied to a geographic entity.
I get asked about this constantly by small-business owners who notice a competitor has one of those tidy info boxes on the right side of a Google search and wonder how to get one for themselves. The honest answer is that you don’t really “get” a Knowledge Panel the way you order business cards. You build a credible, well-documented entity profile across the web, and Google decides whether to surface a panel based on what it has learned about you. This article walks through what the Knowledge Graph actually is, how Google builds it, and what a Toronto small business or independent professional can realistically do to earn a place in it.
A short history: “Things, not strings”
Google launched the Knowledge Graph in May 2012 with a now-famous tagline: “things, not strings.” The idea was that search had spent its first decade matching letters in queries to letters in documents, which is a clumsy way to answer questions. If you searched for “Taj Mahal” in 2011, Google had no real idea whether you meant the monument, the musician, or the casino in Atlantic City. The system was just looking for those eight characters in a row.
The Knowledge Graph was Google’s attempt to build a layer underneath search that understands real-world things and how they relate. Two acquisitions and one open project shaped it heavily. In 2010, Google bought Metaweb, which had built a community-curated database of entities called Freebase. Freebase contributed millions of entity records and a structured schema that became the early skeleton of the Knowledge Graph. Google retired Freebase as a public service in 2016 and moved much of its data and contributor energy to Wikidata, the open knowledge base run by the Wikimedia Foundation. Wikidata is now one of the most important public feedstocks for the Knowledge Graph, and Wikipedia, structured by Wikidata identifiers, sits right next to it.
Since 2012, the Knowledge Graph has grown from roughly 500 million entities to many billions, and Google has woven it into nearly every surface where it answers a question: featured snippets, the People Also Ask box, image carousels, voice answers on Google Assistant, and more recently, AI Overviews and the entity citations in Gemini.
Entities versus keywords
The single most useful mental shift for an intermediate SEO is to stop thinking only in keywords and start thinking in entities. A keyword like “Toronto SEO” is a string. It has no inherent type, no attributes, no relationships. It just sits there as eleven characters and a space.
An entity is different. “Toronto” as an entity has a type (city), attributes (population, latitude, longitude, founding date, mayor), and relationships (capital of Ontario, located in Canada, home to the CN Tower). “Zack Fenech” as an entity, if Google recognizes him, would have a type (person), attributes (occupation, location, employer), and relationships (works in SEO, based in Toronto, owns seobyzack.com).
Keywords still matter for query matching and for understanding what people are searching for. But entities are how Google actually models meaning. When someone types “Toronto SEO consultant,” Google isn’t just looking for documents containing those words. It’s also looking for entities of type “person” or “organization” whose attributes include the occupation “SEO” and the location “Toronto.” If your business is a recognized entity with those attributes, you become a candidate to surface regardless of whether those exact words appear in your content.
How Google actually builds the Knowledge Graph
Google has never published a complete spec, but between patents, official documentation, conference talks, and a decade of observation, the picture is reasonably clear. The Knowledge Graph is assembled from four overlapping types of input.
Highly structured sources. Wikidata, the CIA World Factbook, government open-data portals, licensed databases, and Google’s own products (Google Maps, Google Business Profile, YouTube) provide clean, typed data. When Wikidata says a city has a certain population and cites a statistics agency, Google can ingest that with high confidence.
Semi-structured data with schema.org markup. Schema.org is a shared vocabulary maintained by Google, Microsoft, Yahoo, and Yandex that lets website owners label content with explicit entity types and attributes. When you mark up your homepage as an Organization, declare your founder as a Person, and link out via the sameAs property to your LinkedIn and Wikidata entries, you are handing Google a clean entity declaration about yourself.
Unstructured text on the open web. Google uses natural-language processing to extract entities and relationships from billions of web pages. If a hundred reputable sites describe a person as “the founder of Acme Corp, based in Toronto,” that pattern gets noticed even without any schema markup. Wikipedia is the single most influential unstructured source, because its writing style and citation discipline produce reliable, machine-readable patterns.
Machine learning and inference. Google’s systems are constantly running classification, entity resolution, and relationship extraction at scale. Two pages that both refer to “Z. Fenech, SEO consultant in Toronto” can be resolved to the same entity. Mentions without explicit links can be tied to the right person or business through context. This is also where errors creep in, which is why consistency across the web matters so much.
Knowledge Panels: the visible tip of the iceberg
Knowledge Panels are the boxes that appear on the right side of the desktop search results, or near the top on mobile, when you search for a recognized entity. Search “Drake” and you get a panel with his photo, occupation, birthday, songs, and links to his social profiles. Search a well-known restaurant and you get hours, address, photos, and reviews. Search your favorite local plumber and, if their entity profile is strong enough, you may see a smaller panel with their phone number, hours, and a link to their site.
A Knowledge Panel is not the Knowledge Graph itself. It is a presentation layer pulled from the underlying entity data. Google decides when to show a panel, what fields to include, and what sources to draw from. You cannot directly edit the panel the way you edit a Google Business Profile. What you can do is influence the signals.
For a business with a physical location or a service area, the most direct lever is a verified Google Business Profile. The profile feeds the local panel almost wholesale: name, address, phone, hours, photos, attributes, and reviews. For a personal brand or a knowledge-worker business without a storefront, the panel is built from broader entity data, and you influence it by building consistency across Wikidata, your own site’s schema, social profiles, and authoritative third-party references.
Once a panel exists, you can claim it. Sign in with a Google account, search for the entity, scroll to the bottom of the panel, and look for “Claim this knowledge panel.” Google will ask you to verify your identity, often by linking a verified social profile or an official website that you also have authority over. Claiming the panel does not let you rewrite it, but it does give you the ability to suggest feedback, submit a preferred photo, and respond to user-submitted edits.
Why this matters more every year
Five years ago, you could make a reasonable case that entity SEO was a nice-to-have for large brands and not particularly urgent for small businesses. That argument is gone. The Knowledge Graph and its successor systems now feed almost every Google surface that matters.
AI Overviews, the generative answers Google now places above traditional results for many queries, lean on entities to decide which businesses, products, and experts to cite. Voice search on Google Assistant pulls structured answers from entity data. Featured snippets and People Also Ask blocks use entity relationships to suggest related questions. Even outside Google, the large language models behind ChatGPT, Gemini, Perplexity, and Claude were trained on web data that is itself shaped by Wikidata and Wikipedia, the same feedstocks that built the Knowledge Graph. An entity that is well-defined and consistently described on the open web is far more likely to be recognized by an LLM as a “real” thing worth mentioning.
If you are a small business in Toronto that wants to be the answer when someone asks an AI assistant “who is a good independent SEO consultant in Toronto,” the question is no longer whether you rank for the keyword. The question is whether you exist, in machine-readable form, as a recognized entity with the right attributes.
Schema markup: declaring your entity in code
Schema.org markup is the most direct way to tell Google what your business or personal brand is. It is not magic and it is not a ranking factor on its own, but it is the cleanest signal you can put on your own site, and Google reads it.
For a small service business, the foundational types are usually Organization or LocalBusiness for the company, Person for the founder or key staff, and Service or Product for what you sell. You declare those on your homepage and key pages using JSON-LD, the format Google recommends. The sameAs property is the workhorse for entity SEO. It is an array of URLs pointing to the other places on the web where the same entity is described: your LinkedIn, your Crunchbase listing, your Wikidata entry, your X profile, your YouTube channel, your Wikipedia page if you have one.
When Google’s crawler hits your homepage and sees an Organization block whose sameAs list points to half a dozen authoritative profiles, it has a strong, machine-readable hint about which entity this site represents. When the same crawler hits Wikidata and sees the official site property pointing back at your domain, the loop is closed. This bidirectional consistency is what entity SEO mostly is.
Add a Person schema for yourself if you are the public face of the business. Include your name, your job title, the organization you work for or own, your image, and a sameAs list of your professional profiles. This is how Google connects the human entity to the company entity, which is often what triggers a personal Knowledge Panel for founders, consultants, and authors.
Wikidata: the open back door
Wikidata deserves its own section because it is the most underused tool in small-business SEO. It is a free, editable, structured database of entities, run by the Wikimedia Foundation, that Google ingests heavily.
Unlike Wikipedia, Wikidata does not require general public notability for an entry to exist. The threshold is whether the entity is verifiable from at least one reliable source and “can be described using serious and publicly available references.” A local business with a registered company number, a website, and a few independent mentions can usually justify a Wikidata item. A consultant with a personal site, a LinkedIn, and a few published articles or interviews can also justify one.
That said, Wikidata is a community-edited project with its own culture. Creating an item about yourself or your business is mildly frowned upon and can attract scrutiny. If you go this route, be conservative, cite real sources, do not exaggerate, and do not engage in promotional language. Better still, focus on giving the world enough independent, citable signals (interviews, podcast appearances, articles you have written, profiles on industry sites) that a third party eventually creates the entry, or that you can point to clean external sources when you create it yourself.
Wikipedia is a separate question. Notability requirements for Wikipedia articles are genuinely high and enforced strictly by experienced editors. For most small businesses and most consultants, a Wikipedia article is not realistic and is not worth chasing. Wikidata is the practical target.
Consistency: NAP and entity references across the web
Local SEO has talked about NAP (name, address, phone) consistency for fifteen years, and it still matters, but the entity framing is more accurate. What Google is doing is entity resolution: deciding whether two mentions on two different sites are about the same thing.
If your business name is listed three different ways across the web (“Acme Plumbing,” “Acme Plumbing Inc.,” “Acme Plumbing & Heating”), with three different phone numbers and two different addresses, Google may resolve those into two or three separate entities, none of them strong. Pick one canonical name, one canonical address, one canonical phone, and use them everywhere: your site, your Google Business Profile, your social profiles, your citations on local directories, your Wikidata entry, your invoices.
The same principle applies to personal brands. Decide whether you are “Zack Fenech” or “Zachary Fenech” or “Zack F.” in your public-facing professional identity, and stop drifting between them. Use the same headshot across your professional profiles where possible. Consistency is the easiest entity signal to control, and most businesses neglect it.
The common mistakes
Skipping schema markup entirely. A surprising number of small-business sites have no structured data at all, or have only the bare minimum injected by an SEO plugin without any sameAs links. This is leaving the most direct entity signal on the table.
Inconsistent entity references. A LinkedIn that lists the founder under one company name, a Crunchbase that uses a different legal entity, a website that uses a trading name, and a Google Business Profile that uses a fourth variation. Every inconsistency forces Google to guess.
Treating the Knowledge Graph as something you can directly edit. You cannot log into the Knowledge Graph. You cannot type your business description into it. You can only influence the upstream sources. If your panel says something wrong, the fix is almost always to correct the underlying source that Google is reading from, not to argue with the panel.
Buying citations from low-quality directories. A decade ago, citation packages from spammy directories could move the needle on local SEO. Today they mostly create entity-resolution noise and do not generate the kind of authoritative mentions Google weighs heavily. Spend the same money on one quality mention and you will get more entity benefit.
Expecting overnight results. Entity SEO is slow. Google needs to see a pattern across multiple crawls and multiple sources before it commits to recognizing a new entity. Three to six months of consistent effort is a realistic baseline.
A practical 3-to-6 month roadmap for a Toronto small business
Month 1: foundation and audit. Pick your canonical name, address, and phone, and write them down. Search yourself and your business in Google and note every variation. Pull every public profile and list them in a spreadsheet. Sign up for or claim your Google Business Profile if you have not, and verify it.
Month 2: clean up and consolidate. Update every profile in your spreadsheet to use the canonical NAP. Remove or update old profiles you no longer use. Make sure each profile links back to your website. If you have multiple websites, decide which one is canonical and 301-redirect the others.
Month 3: structured data on the site. Add JSON-LD Organization or LocalBusiness schema to your homepage, with the full canonical NAP, your logo, and a complete sameAs list pointing to your profiles. Add Person schema for the founder. Validate using Google’s Rich Results Test and Schema Markup Validator. Fix any errors.
Month 4: Wikidata and the bidirectional loop. Create a Wikidata item for your organization if one does not exist, citing your website, Google Business Profile, and any independent mentions. Add the “official website” property pointing to your canonical domain. Once the Wikidata item exists, add its URL to the sameAs list in your site’s schema. Now the loop is closed.
Month 5: authoritative third-party mentions. This is the slow part. Pursue one or two meaningful external mentions per month. A guest article on a respected industry blog. A podcast interview where the show notes link back to your site. A profile on a chamber of commerce, BIA, or industry association site. A speaking slot at a local event with a published bio.
Month 6: review and iterate. Search your name and business name again. Note any Knowledge Panels that have appeared, or any changes to existing ones. Claim any panels you can. Check Google Search Console for new query patterns. Look at your AI Overview appearances if any, and at how you are cited by Perplexity or Gemini for relevant queries.
The mindset shift
For most of SEO’s history, the goal was to get a specific page to rank for a specific query. That work is still real and still pays. But entity SEO is a different game, played on a longer timescale, with a different goal: getting Google to recognize you as a real, well-defined thing in its model of the world.
The Knowledge Graph rewards consistency, verifiability, and patience. It does not reward clever tricks or aggressive optimization. A small business that picks a clear identity, declares it everywhere in the same way, backs it up with structured data, and earns a handful of legitimate mentions every quarter will, over a year or two, become the kind of entity Google trusts enough to surface in panels, AI Overviews, and the structured answers that increasingly mediate how customers find their next service provider.
You cannot edit the Knowledge Graph. You can only feed it. The businesses and brands that feed it well, consistently, for long enough, are the ones whose names start to appear when somebody asks a machine “who should I hire in Toronto for this?” That is the prize. It is worth the work.