The Knowledge Graph Launch (May 2012): 'Things, Not Strings'

When I talk to business owners about "entities," I often get a polite blank look. Most people still picture Google as a giant keyword matcher: you type words, Google finds pages containing those words. That picture was mostly true until 16 May 2012, when Google announced the Knowledge Graph under a slogan that has followed SEO ever since: "things, not strings."

That launch is why your brand can have a knowledge panel, why Google can tell Apple the company from apple the fruit, and why AI Overviews can talk about your business without quoting your homepage word for word. It was not a ranking update in the Penguin or Panda sense. It was a change in how Google understood the world.

Let me explain what Google actually announced, where the data came from, what changed on the results page, who lost traffic, and what "entity SEO" means for you in 2026. You can see how this fits with everything else Google changed in 2012 on our Google algorithm updates timeline.

The Knowledge Graph launch at a glance

Item	Details
Announced	16 May 2012, in a post by Amit Singhal, SVP of Engineering
Type	New search feature and underlying entity database, not a ranking algorithm update
Confirmed by Google?	Yes, officially announced
Size at launch	Over 500 million objects and more than 3.5 billion facts and relationships
Main data sources	Freebase, Wikipedia and the CIA World Factbook, among other public sources
Rollout	Gradual, US English first, on desktop, smartphones and tablets
Who was affected	Sites answering simple factual queries (reference sites, encyclopedias, fact pages); brands and public figures gained panels

What Amit Singhal announced on 16 May 2012

The announcement post, "Introducing the Knowledge Graph: things, not strings", was short but ambitious. Singhal described a database of real-world objects (people, places, things) and the connections between them. The key idea was that Google would no longer just match the letters you typed. It would try to work out which thing you meant.

“The Knowledge Graph enables you to search for things, people or places that Google knows about”

— Amit Singhal, Google, Introducing the Knowledge Graph: things, not strings

Google said the graph would improve search in three ways:

1. Find the right thing. Many words mean several things. Singhal used "taj mahal" as the example: the monument, or the musician? The Knowledge Graph let you narrow results to the one you meant.
2. Get the best summary. Google would show the facts people most often want about a topic, based on what people searched for in aggregate.
3. Go deeper and broader. The graph surfaced related things, which is where features like "People also search for" came from.

One line from the post shows how far Google wanted to go: "we can now sometimes help answer your next question before you've asked it." In 2012 that sounded like marketing. In 2026, with AI Mode suggesting follow-ups, it reads like a roadmap.

Freebase and the roots of the graph

Google did not build the Knowledge Graph from nothing. Its biggest foundation was Freebase, an open, community-edited database of facts created by a company called Metaweb. Freebase was announced publicly in March 2007, and Google acquired Metaweb in July 2010, according to Wikipedia's Freebase article.

Freebase was later wound down. Google announced in December 2014 that it would close Freebase and help move its data to Wikidata, and freebase.com shut on 2 May 2016. That history matters for you, because it is why Wikidata is still one of the most practical places to describe a brand or person in structured form. I cover how to do that safely in my guide to Wikidata and Wikipedia for brands.

Here is how the graph grew, using only figures Google or reliable reporting published:

Date	Size reported	Source
May 2012	500 million+ objects, 3.5 billion+ facts	Google announcement
December 2012	570 million entities, 18 billion facts; expanded to seven more languages	Wikipedia, Knowledge Graph
Mid-2016	70 billion facts	Wikipedia, Knowledge Graph
May 2020	About 500 billion facts on about 5 billion entities	Google, Danny Sullivan

The December 2012 expansion covered Spanish, French, German, Portuguese, Japanese, Russian and Italian, per Wikipedia. The growth from 3.5 billion to 500 billion facts in eight years tells you the graph stopped being a curated encyclopedia and became something Google fed constantly from across the web.

Entities versus keywords, explained simply

Let me make this concrete. A keyword is a string of letters, like "mercury." An entity is a specific thing with an identity: Mercury the planet, mercury the element, Freddie Mercury the singer. Each entity has attributes (a planet has a distance from the sun, a singer has a birth date) and relationships (Freddie Mercury was a member of Queen).

Before 2012, Google mainly worked out meaning from words on pages and links between them. After 2012, it also had a map of things to check those words against. In practice that meant:

• Ambiguous searches got better. Google could ask, in effect, "which one did you mean?"
• Facts could be answered directly. Height, founding date, capital city: no click needed.
• Related ideas became connected. A search about one film could point to its director, actors and sequels.
• Brands became things, not just websites. A business could be understood as an entity with a founder, a location, products and profiles.

Singhal put the point well: "It's the intelligence between these different entities that's the key." For a deeper, practical version of this idea, see my post on entity SEO.

Knowledge panels arrive on the results page

The visible change was the box on the right side of desktop results, which the industry soon called the knowledge panel. Search for a famous person, landmark, film or company and you got a summary: a photo, a short description, key facts and related entities.

Google says panels are "automatically generated" and draw on "various sources across the web," according to its knowledge panel help page. They are updated automatically as information changes, and the subjects can get verified to "claim this panel and suggest changes." In 2020 Danny Sullivan described the graph behind them as "a giant virtual encyclopedia of facts" and stressed that "Wikipedia is a commonly-cited source, but it's not the only one," in Google's 2020 explainer on the Knowledge Graph.

If you want your own business to earn one, my guide on how to get and claim a Google knowledge panel walks through the process.

Who lost traffic when Google started answering directly

A database that answers questions on the results page means fewer clicks for some pages. This was the first big step toward what we now call "zero-click" search.

The clearest public example was Wikipedia. In January 2014, The Register reported that knowledge panels had caused significant declines in Wikipedia readership, and Wikimedia's Dario Taraborelli said the lack of attribution "undermines people's ability to verify information," as summarized in Wikipedia's Knowledge Graph article. By May 2016, knowledge boxes appeared for "roughly one-third" of 100 billion monthly searches, according to reporting cited there.

I treat these figures as direction rather than precise measurement, because nobody outside Google can see all the data. But the pattern was clear, and it hit a predictable kind of page:

Type of page	Effect of the Knowledge Graph	Why
Simple fact pages ("how tall is…", "when was … founded")	Lost clicks	The answer appears on the results page
Reference and encyclopedia sites	Lost some clicks, still cited	Panels summarize them
Brands and public figures	Gained visibility	Panels give them a prominent box for their name
In-depth guides and opinion	Little direct impact	Panels cannot replace analysis or experience
Local businesses	Gained a panel through Google's business listings	Hours, reviews and maps shown directly

A simple example of the math, using hypothetical numbers: imagine a trivia site that got 10,000 monthly visits for "Eiffel Tower height" type queries. If a panel answers half of those searchers on the page, the site might drop to around 5,000 visits without losing a single ranking position. Rankings stayed, clicks went. That is exactly the pattern many publishers now see again with AI Overviews.

From the Knowledge Graph to Hummingbird and AI search

The Knowledge Graph was the foundation for several later changes. The most direct one was Hummingbird, rolled out around August 2013 and announced on 26 September 2013. Google's ranking systems guide calls Hummingbird "a major improvement to our overall ranking systems made in August 2013," and it is widely understood to have made better use of the graph's entity understanding. I explain it in my post on Google Hummingbird.

The line of development since then is easy to follow:

1. 2012: Knowledge Graph gives Google a database of entities.
2. 2013: Hummingbird uses meaning, not just keywords, to interpret whole queries. In the same year Google tested other entity-friendly features such as in-depth articles.
3. 2010s: Featured snippets and voice answers rely on knowing what a query is about.
4. 2023–2026: AI Overviews and AI Mode combine language models with Google's index and its entity knowledge. My AI search rollout timeline covers that stage.

The Knowledge Graph also arrived in a very busy year. Just months earlier, Google had launched Search Plus Your World, and a week later it ran the first Penguin data refresh. Google was changing both what it showed and how it judged links at the same time.

Entity SEO in 2026: a practical checklist

You do not need to be a famous brand to benefit from the thinking behind the Knowledge Graph. You need Google to be confident about who you are and what you do. Here is how I suggest you approach it:

1. Pick one canonical name for your business and use it consistently on your site, social profiles and directories.
2. Write a clear About page that states who you are, where you are, what you do and who runs the business.
3. Add Organization structured data with your logo, address and sameAs links to official profiles. Our Organization schema guide shows the code.
4. Mark up people and articles. Google's Article structured data documentation lists author name and author URL as recommended properties, which connect content to real people.
5. Keep your Google Business Profile accurate if you serve local customers.
6. Earn mentions on trusted third-party sites, because the graph is built from many sources, not from what you say about yourself.
7. Consider Wikidata carefully, only if you meet its notability rules and can cite independent sources.

If structured data is new to you, start with our lesson on structured data and schema markup for beginners.

Questions about the 2012 Knowledge Graph launch

Was the Knowledge Graph a ranking update?

Not in the usual sense. It was a new database and search feature. It changed what appeared on results pages and how Google understood queries, which later fed into ranking systems like Hummingbird.

Where does the Knowledge Graph get its information?

At launch, from public sources including Freebase, Wikipedia and the CIA World Factbook. Today Google says it uses hundreds of web sources, licensed data and structured markup, with Wikipedia as one source among many.

Can I get a knowledge panel for my small business?

Local businesses usually get a panel through Google Business Profile. A brand or person panel depends on Google having enough trustworthy information about you across the web. You cannot buy one.

Does schema markup put me in the Knowledge Graph?

Not directly. Markup helps Google understand your pages and connect them to your entity, but Google decides what goes into the graph based on many sources.

Is "things, not strings" still relevant with AI search?

More than ever. AI answers need to know which business, person or product a query is about. Clear entity signals make it easier for Google to mention you correctly.

Making Google confident about your brand as an entity

The Knowledge Graph launch was the moment Google started treating brands as things with identities, not just websites with keywords. Fourteen years later, that idea sits under knowledge panels, AI Overviews and much of how Google decides who to trust.

The good news is that the work is not mysterious: consistent naming, clear About and author pages, solid structured data and real mentions elsewhere. If you would like help making your brand easier for Google to understand, our SEO team can audit your entity signals and fix the gaps. Get a quote and tell us about your business.

Google updates around the same time

Before this update:

• Google Penguin (2012): The Update That Changed Link Building (Apr 2012)
• The 2012 Blog Network Deindexing: How Google Took Down BuildMyRank (Mar 2012)

After this update:

• Penguin 1.1 (May 2012): The First Penguin Data Refresh (May 2012)
• The Pirate Update (2012): How Copyright Notices Affect Rankings (Aug 2012)

Or browse every Google update on one timeline, newest first.

Sources for this Knowledge Graph launch history

1. Google: Introducing the Knowledge Graph: things, not strings (16 May 2012)
2. Google: Danny Sullivan's explainer on the Knowledge Graph and knowledge panels (May 2020)
3. Google Help: About knowledge panels
4. Wikipedia: Google Knowledge Graph
5. Wikipedia: Freebase (database)
6. Wikipedia: Timeline of Google Search
7. Google Search Central: A guide to Google Search ranking systems
8. Google Search Central: Article structured data