Google’s Great Clarity Cleanup: 3 Shifts Redefining the Knowledge Graph and Its AI Future
Google’s Great Clarity Cleanup: 3 Shifts Redefining the Knowledge Graph and Its AI Future
Published on Search Engine Land August 18, 2025 by Jason Barnard
Billions of entities vanished from Google’s Knowledge Graph in June 2025. Here’s how it signals a bold new direction in AI-powered search.
Google’s Knowledge Graph saw its largest contraction in a decade in June: a two-stage, one-week drop of 6.26% - over 3 billion entities deleted.
Since 2015, we’ve tracked the Knowledge Graph and have never witnessed a cull of this magnitude.
We’re calling it Google’s great clarity cleanup.
If you recently noticed issues with a Knowledge Panel for yourself or a client, this update may seem like the culprit - but in most cases, it wasn’t.
The real story is much bigger.
What we just witnessed offers a clear signal about Google’s direction.
The Knowledge Graph is the fact-checking core of Google’s algorithmic trinity, giving it a competitive edge.
The way Google refines this advantage tells us exactly where it’s headed in the AI assistive engine race.
Google’s summer Knowledge Graph tradition - and 2025’s twist
Since 2015, Google has rolled out major Knowledge Graph changes each summer - almost like hitting a big red “update” button.
Some stand out more than others.
The “Budapest update” of summer 2019 expanded Google’s ability to verify facts in the Knowledge Graph.
The summer 2023 update zoomed in on people, especially content creators.
History has taught us that when a Knowledge Graph update signals a clear strategic shift, it’s a preview of the next stage in SEO, AEO, and GEO.
Those insights have kept us consistently ahead of the curve.
This year, Google hit the button twice - first on June 13, then again on June 20.
The message was unmistakable: the June update was about one thing - clarity. It marks the start of what I call the age of algorithmic clarity.
A strategic contraction in the Knowledge Graph
From May 2024 to May 2025, the Knowledge Graph expanded at a steady 2.79% - healthy, incremental growth by our tracking.
Then, in June, everything flipped: over two closely timed updates, the graph contracted by 6.26%, wiping out more than 3 billion entities in a single week.
That’s twice the net additions of the entire previous year, erased almost overnight.
The scale and speed point to a deliberate “anti-hoarding” move - trading sheer volume for clarity and confidence.
The goal is a leaner, higher-quality dataset that will underpin AI features like AI Overviews, AI Mode, and Google Learn About.
No other Big Tech player currently has a comparable knowledge base, so refining it is a direct investment in Google’s AI advantage.
And this was no blanket purge. The data shows three specific areas where Google cut decisively, each revealing a different facet of its clarity-first strategy.
1. The massive drop in event entities is a post-pandemic reset
The “event” category dropped 76.91% as Google wound down the temporary, pandemic-era surge in event tracking.
At the start of the COVID-19 pandemic, our data showed an explosion of event-related entities, presumably to help users navigate cancellations and online events.
For the last five years, turnover in event entities has been incredibly fast.
We saw this with our own webinars: online events were often added as entities within 15 minutes of being published, then deleted shortly after.
The average lifespan of an event entity dropped from 839 days pre-COVID to just 124 days after March 2020.
Now that the world has stabilized, Google is switching off event reactivity.
The resources required to maintain that level of real-time event data are no longer justified, so the system has been reset.
The lesson is clear: take advantage of temporary opportunities while you can, but don’t build your digital strategy around chasing the latest algorithmic updates.
Build your AI assistive engine optimization strategy on stable corporate, personal, and product brands.
2. Google is cleaning house by removing ambiguous thing entities
The focus on events was a temporary five-year glitch we can now safely forget.
The more significant cleanup comes from the “thing” category - Google’s most generic classification.
This update reduced the number of entities labeled as “thing” by 15.27%, which is about 8 billion entities.
Multityping - when Google assigns multiple categories or types to the same entity - is common, and “thing” is often used as one of those extra types.
In our dataset, about half of all entities are multityped, and 27.83% - roughly 13 billion entities in Google’s Knowledge Graph - have “thing” as one of their types.
While the total number of thing entities dropped sharply, the proportion of entities that are unityped - given only a single, definitive type - rose from 23.9% to 28.7%.
This points to a shift toward single, unambiguous typing for concepts, topics, or things.
After manually checking 10,000 unityped thing entities (a process that left my eyes and brain aching), the pattern is clear - Google is moving away from using it as a lazy label, and toward reserving it for concepts and topics that cannot be classified in any other way.
Over the last decade, the Thing entity type has been an easy, sometimes spammy, entry point into the Knowledge Graph.
That entry point is closing fast, and getting a foothold as a “thing” will soon be viable only for entities with no alternative classification.
Get the newsletter search marketers rely on.
Sign me up!
3. Google is removing ambiguity by focusing on unityped person entities
This focus on clarity for person entities continues a long-term trend we saw in 2023 and again in 2024.
The 2023 update increased the number of person entities threefold, while also concentrating on subtitles that signal E-E-A-T.
Roles like “writer” and “author” rose by 21% while others were removed. It was a clear move to highlight expertise.
The 2024 update continued that focus, particularly on identifying person entities with a “content Creator” role.
The June update takes it a step further, significantly changing the proportion of person entities that are unityped.
Google’s confidence that an entity is unambiguously a person rose from 70.16% to 76.78% during this update.
It is actively removing other classifications from person entities to ensure there is no doubt in its algorithmic mind.
These very clearly point to a huge focus on person entities for N-E-E-A-T-T.
Clarity is now the only point of entry
The June update confirms a principle I’ve emphasized since Hummingbird in 2013: in the Knowledge Graph, clarity is the only point of entry.
You and your client may have invested heavily in a digital marketing strategy to be visible across every channel for awareness and consideration.
Your brand may be seen and evaluated.
However, the final step, conversion, comes down to a bottom-of-the-funnel moment of truth for both humans and AI assistive engines: confidence in understanding the brand.
“Do I know who they are, and can I trust them?”
The solution is algorithmic knowledge - the Knowledge Graph (and that means the Knowledge Graph of every Big Tech player).
Your identity, offering, and credibility must be clearly and confidently understood by the algorithmic encyclopedias that make the final recommendation to your perfect client.
Most brands fail here, and yet the fix is straightforward.
AI is like an eager-to-please child - present it with a messy, contradictory digital footprint, and it gets confused.
This update shows Google actively tidying that bottom-of-the-funnel “room” and discarding any brand that creates ambiguity.
A solid, unityped, confident place in the Knowledge Graph is your only route to being at the top of the algorithmic mind at that zero-sum, bottom-of-the-funnel “perfect click” moment.
Breaking: Google strikes again on Knowledge Graph clarity
On Aug. 11, in a break from its decade-long tradition of a single summer Knowledge Graph update, Google launched a second clarity cleanup less than two months after the first.
This one-day update showed an even sharper focus on corporation, organization, and brand entities.
