Google keeps
a database
of people.
Behind the search results Google runs a large database of things rather than pages. People, companies, places, films, books. Each one has a record and an identifier.
It has an official name that sounds more forbidding than the idea deserves. What it amounts to is a filing system: instead of only storing pages that mention a name, Google stores the subject itself and hangs facts off it.
If you have ever wondered how Google answers a question without sending you anywhere, this is where the answer came from.
What feeds a record about a person
- An encyclopedia article, which is the single strongest input
- A structured facts database entry, considerably easier to qualify for
- Books, papers or patents carrying permanent identifiers
- Professional registers: medical, legal, academic, architectural
- Institutional staff and faculty pages
- Press coverage that names you rather than quoting you in passing
- Conference and speaker listings on the organiser's own site
- Company records where you are named as an officer
- A consistent photograph appearing across all of the above
What is actually stored
A record holds a name, a type, a short description and a set of properties. For a person that means things like role, employer, education, dates, place, and connections to other records.
Each record carries an identifier that looks like a short code. That code, not the spelling of your name, is what identifies you inside the system.
Records also point at each other. A person works for a company; the company is itself a record. Those connections are why Google can answer questions it was never explicitly told the answer to.
Where the facts come from
Structured public sources first. Encyclopedias, catalogues, registers, databases, institutional pages. Anything already organised as data rather than prose.
Then ordinary web pages, where the same facts appear repeatedly and consistently enough to be trusted.
Then licensed data from providers, which is where some dates and biographical details come from and why they can be hard to correct at the panel.
Notably absent: anything you asked to be included. There is no submission route for a person, which surprises people who assume there must be a form somewhere.
How a person gets in
Not by applying. A record appears when enough trusted sources describe the same subject consistently, and Google's own processes reconcile them into one entry.
The threshold is not published and it is not fixed. It varies by how common your name is, how distinctive your field is, and how much material exists.
Which is why honest timelines are ranges rather than dates. Anybody quoting a specific week is quoting a decision that is not theirs to make.
How to check whether you are in it
Search your name and look for a box beside the results. If one appears, a record exists and the name reaches it.
If nothing appears, that does not prove there is no record. Search your name with your city or employer added, which sometimes surfaces a record the bare name does not reach.
There is also a public lookup interface, and it is worth knowing it is partial and lags. An empty result there means the lookup index does not hold you, not that no record exists. We run that tool on a sister site and we label it that way for exactly this reason.
Why it matters more now than it did
AI assistants read structured records before they read pages. When somebody asks an assistant who you are, a record produces a confident answer and its absence produces either a hedge or a guess.
The guess is the problem. A model with no record and a common name will sometimes blend you with somebody else, and it will do it fluently.
So the value of being in the database has quietly shifted from a nice search result to controlling what gets said about you in a conversation you are not part of.
What you can and cannot change
You cannot edit a record directly. You can claim the panel it produces and suggest corrections, and Google weighs those against its sources.
You can change the sources, which is the real lever and the slow one.
And you can prevent a lot of trouble cheaply by being consistent. Most of what looks like a Google problem is a disagreement between your own profiles.
What it does and does not hold
| In the database | Not in it | |
|---|---|---|
| Your role and employer | Yes, if sourced | |
| Education and dates | Yes, from data partners | |
| Books and papers | Yes, via catalogues | |
| Your follower counts | No independent source | |
| Your services and prices | Commercial, not factual | |
| Your own claims about yourself | Not without corroboration | |
| Reviews and ratings | A different system entirely | |
| Connections to organisations | Yes, when stated consistently |
How facts get corrected, and why it feels arbitrary
A correction succeeds when the sources behind it change, and fails when they do not. That single rule explains almost every confusing outcome people report.
It is why a job title updates within weeks when the employer's own site reflects it, and why a birth year sits wrong for years when it came from a data partner rather than a web page.
It is also why suggesting a correction without touching the sources usually achieves nothing, and why the same suggestion resubmitted five times achieves nothing five times.
The connections between records matter more than the records
A record on its own is a card with facts. What makes the system useful to Google, and valuable to you, is that records point at each other.
You work for a company; the company is a record. You wrote a book; the book is a record. You studied somewhere; the institution is a record. Each connection is a fact confirmed by two parties rather than asserted by one.
Practically this means affiliation is worth engineering deliberately. Being named on an organisation's own site, in a way that matches how you describe the role everywhere else, is worth more than another paragraph about yourself.
Practical checks you can run this week
Search your name signed out and note whether any box appears, and whether the results are about you or about somebody else.
Search your name with your employer or city added, which sometimes reaches a record the bare name does not.
Ask three different AI assistants who you are, in fresh sessions, and write down what comes back. Disagreement between them is a good proxy for how thin the underlying evidence is.
Open your five most visible profiles side by side and compare name spelling, job title, city and photograph. Most people find several disagreements, and every one of them is free to fix.
Questions
Is the Knowledge Graph the same as a Knowledge Panel?
No. The graph is the database; the panel is one way its contents get displayed. A record can exist without a panel ever appearing on your plain name.
Can I submit myself to it?
No. There is no submission route for a person. Facts get in by appearing consistently in sources Google already trusts.
How many entities are in it?
Google has quoted figures in the billions over the years. The number is not useful to you; what matters is whether one of them is you.
Does being in it help my website rank?
Not directly. It is a different system. It affects how you are described, not where your pages sit.
Why does the lookup tool return nothing for me?
That index is partial and lags the live one. An empty result means the lookup does not hold you, which is not the same as having no record.
Can a record be wrong?
Yes, and usually because a source is wrong or out of date. The correction route runs through the sources, not through the record.
Everything else on this, in detail
Seven shorter pages, each answering one question properly.
If you want this handled
Three thousand dollars flat, half at the start and half only when there is something live to show you. Six to twelve months is the honest window, and about a third of the people who ask are told not to proceed.