Entity SEO,
without the
mystique.
Search used to be about matching words. You wrote a page containing a phrase, somebody searched that phrase, and the two were matched on the letters.
That is still part of it, and it stopped being the interesting part around the time Google started keeping files on people, places and organisations rather than only on pages. Entity SEO is the work of getting into those files correctly.
For a company it is fairly well understood. For a person it is barely practised, which is the entire opportunity.
Page SEO against entity SEO
| Ordinary SEO | Entity SEO | |
|---|---|---|
| Unit of work | A page | A fact about a subject |
| Measured by | Ranking position | Whether the record exists and is correct |
| Main lever | Content and links to your site | Independent sources agreeing |
| Your own site is | The asset | The reference point |
| Keywords matter | Yes | Barely |
| Timescale | Weeks to months | Six to twelve months |
| Survives a site rebuild | No | Yes |
| Read by AI assistants | Sometimes | First |
What an entity actually is, in plain terms
A person, a company, a place, a book, an event. Something that exists independently of any web page describing it.
Google keeps a record for each one it is confident about, with an identifier attached. Your name is text; the record is the thing. Two people can share the text and still have separate records, which is the whole reason the distinction exists.
Everything below is about getting a record created, keeping it accurate, and making sure the plain version of your name points at it.
How it differs from the SEO you already know
Ordinary SEO optimises a page so it ranks for a query. The unit of work is a URL and the measure is position.
Entity SEO optimises a set of facts across sites you mostly do not own, so a machine can reconcile them into one subject. The unit of work is a claim, and the measure is whether the record exists and is right.
Which means most of the familiar levers do very little here. Keyword density, internal anchor text, page speed and content length are page-level concerns. They matter for ranking a page. They do not persuade Google that a human being exists.
The lever that does work is corroboration: several independent sources saying the same thing about the same subject, in a form a machine can parse.
The three things a machine needs before it commits
Existence. Something outside your own control has to assert that you exist and what you are. One directory entry, one press mention, one catalogue record.
Consistency. Those assertions must not contradict each other. This is where most people fail, and it fails silently, because a contradiction produces no error message.
Connection. The assertions have to be linkable to one another. A name in isolation is a coincidence; a name attached to the same role, city, employer and photograph across five sources is a person.
Where structured data fits, and where it does not
Marking up your own site helps. It labels the facts so nothing has to be inferred, and it removes a class of ambiguity for free.
It is not evidence. Markup on your own domain is you making a claim about yourself in a tidier format, and a tidier claim is still a claim.
The correct mental model: markup is how you speak clearly. Independent sources are what make you believed. People who only do the first half wonder why nothing happened.
The order the work goes in
Audit before you build. About half the people who assume they have nothing already have fragments, and extending a fragment is faster than starting clean.
Kill contradictions before you add anything. Adding new claims on top of an inconsistent base makes the inconsistency worse.
Build the reference page next, because everything after it gets checked against it.
Then place structured facts, then earn independent coverage, then wait. The waiting is not optional and it is not a sign of failure.
How to tell whether it is working
Search the identifier if you have one. A card coming back means the record exists even if the plain name does not reach it yet.
Ask an AI assistant who you are, in a fresh session. Assistants read structured records early, so they often reflect progress before the search page does.
Watch whether the description Google shows about you changes. Movement there means reconciliation is happening.
What not to watch: rankings and traffic. They measure a different system and will mislead you in both directions.
Doing it, in order
Find what already exists
Search your name, your name plus city, your name plus employer. Note every source you did not create.
List every contradiction
Name spellings, titles, dates, photographs, cities. Put them in one document. Most people find eight to fifteen.
Fix the contradictions you control today
Profiles, author boxes, speaker bios. Free, fast, and the base everything else sits on.
Write one reference page and mark it up
On your own domain. Plain facts, structured labels, links out to everywhere you genuinely appear.
Place facts in structured sources
Registers, catalogues, indexes, institutional pages. Dull and heavily weighted.
Earn coverage from people with nothing to gain
The slow half. It cannot be automated, because the value is that it is not yours.
Wait, and measure the right things
Identifier lookups and assistant answers, not rankings.
Where people waste money
Paid placements in publications that print anything. They are recognised for what they are and contribute close to nothing.
Beautiful personal websites with nothing independent behind them. The site is the reference point, not the evidence.
Forcing an encyclopedia entry that does not qualify, which is the one mistake that leaves a permanent public record working against you.
And changing everything repeatedly, which prevents a consensus from ever forming.
What a name collision does to the whole plan
If somebody more written about shares your name, none of the standard advice applies cleanly. You are no longer establishing that a person exists; you are arguing about what a word means, against somebody with a head start.
Three routes exist. Compete, which is slow and sometimes not winnable. Differentiate, by adopting a fuller or middle-initial form of the name consistently from today. Or narrow, by building the record around a name-plus-field combination and accepting the bare name may never resolve to you.
The mistake is discovering which of the three applies in month ten. It takes an afternoon to work out at the start: search the name, count the distinct people on page one, and look at what is backing the strongest of them.
Why AI assistants changed the value of this work
Until recently the payoff was a nicer search result. A person could reasonably decide it was cosmetic and spend the money elsewhere.
Assistants changed the arithmetic, because they answer rather than list. Asked who somebody is, a model with a structured record to draw on produces a confident, sourced answer. With nothing, it produces a hedge, or worse, it produces a fluent blend of you and somebody else with your name.
The blend is the genuinely new risk. A wrong search result is visibly a wrong search result. A wrong sentence in an assistant's answer arrives with the same confidence as a right one, and the person reading it has no reason to check.
So the work has quietly moved from reputation polish to something closer to record-keeping: making sure the version of you that machines repeat is the accurate one.
What to measure, and what will mislead you
Measure whether a record exists, by searching the identifier if you have one. That is binary and it is the thing you are buying.
Measure what assistants say, in a fresh session, on a fixed set of questions, once a month. Write the answers down. Movement in how you are described is the earliest reliable signal there is.
Measure whether the description Google shows about you changes, because that reflects reconciliation happening underneath.
Do not measure rankings or traffic. They move for a hundred unrelated reasons, they will show gains in months when nothing happened and losses in months when everything did, and they are the single most common reason people abandon this work early.
Doing it for a company instead of a person
Most of it transfers, and the difficulty inverts. Organisations leave a heavy structured trail without trying: company registers, funding databases, trade press, customer coverage.
So a company that has filed accounts, raised money and been written about usually has most of what it needs already, and the work is tidying rather than generating.
The two do not substitute for each other. A well-known company does not confer a record on its founder, and a well-known founder does not confer one on a company nobody has covered. If you want both, they are two projects, and the personal one is the longer.
Questions
Is entity SEO the same as personal branding?
No. Personal branding persuades people and works on story and consistency of taste. This persuades a machine and works on independent corroboration. You can be excellent at one and invisible in the other.
Do I need a Knowledge Panel for entity SEO to be working?
No. The panel is the visible end. A record can exist and answer AI assistants for months before it surfaces on your plain name.
Does this replace normal SEO?
No, it sits alongside it. They optimise different systems and neither substitutes for the other.
How long before anything changes?
Six to twelve months from the point the evidence is in place, which is not the same as six to twelve months from when you started thinking about it.
Can I do it myself?
The consistency work, yes, and it is worth doing whoever you hire. The independent coverage is the part that takes either relationships or time.
Is there a tool that does it?
There are tools that check what exists. There is no tool that makes other people write about you, which is the part that matters.
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.