Almost every article about Google Knowledge Panels comes from people who have not built the thing that sits under one. Claim your panel. Add schema. Stay consistent. These are good points. They are fine. But they do not show what it is like when you start and try yourself.
We did everything on a property that we own, and wrote down all the things that did not work. We had two accounts, but one got blocked. There was a tool we could not use. In the end there was a live entity, and no improvement in rankings at all. That turned out to be the most helpful part of the story.
First, what a Knowledge Panel actually is
The box you see on the right of Google’s results is called the panel. It has a logo, a short line that tells you what it is, a few links, and some facts. Here is what most people do not say in simple words: you cannot make this box show up. Google puts it there when it feels sure it knows who you are.
That distinction is the entire game.
You can’t make a Knowledge Panel by yourself. What you can do is help Google find good reasons to make one about you. All ways to do this come down to the same thing. You have to give Google proof, using sources it already knows and trusts.
This is why “buy a Google Knowledge Panel” is a bad question with real search volume behind it. Nobody sells panels. What people sell is the work underneath, and a lot of them do that work badly.
Why we ran this on our own site
The subject here is CLT Daily. It is an independent news site and daily podcast in Charlotte. We own it. That is why this post has facts in it. There is no client to protect and no need to smooth over anything.
It was a hard case that we chose on purpose. When we began, the site had a Domain Rating of 0. There were no organic keywords. There was no real search visibility. If these things work only for brands that are already well known, we wanted to see that for ourselves.
Step one: find your notability anchor
Before we started with Wikidata, we had to ask a question that most guides leave out. What is the proof that this thing is real and that it matters?
For CLT Daily, you might think the website was the main thing. That was not right. A brand-new site with no authority is exactly the kind of subject whose entity gets challenged and deleted out from under you. What worked better was the podcast, because podcasts come with independent platform identifiers that anyone can verify without taking your word for it.
- Lesson: find the part of the business that is already in a database you do not control. Platform IDs, registry numbers, and listings on sites you cannot edit. That is your proof.
Step two: Wikidata, and the mistake that cost us an account
Wikidata is an open and structured database. It helps Google know more about people, places, and things. The database is free and open to everyone. You can add to it yourself, which is not common for Knowledge Graph inputs. Our item is Q141005278.
The first attempt got blocked, and this is a part we have never seen written down anywhere.
We signed up with the brand name. Wikimedia does not allow a company or organisation name for an account. They treat these as role accounts and block them. And a brand-named account whose first edit lands on that same brand gets noticed straight away. That is the profile they block quickly, and it draws a request to delete your item at the same time.
We left that account with zero edits and opened a new one under a personal name. The new account carries a conflict-of-interest note on the user page. The founder is named, the relationship is stated, and the abandoned account is declared so nobody could later call it sock-puppetry.
- The rule, for any brand, in any field: use a personal username and disclose the conflict of interest. Never a company name. No exceptions.
Step three: the tool you can’t use on day one
QuickStatements is the main tool for adding many entries to Wikidata at once. Every guide tells you to use it. But it needs an autoconfirmed profile, which means a minimum account age and a minimum number of edits. A profile made today gets told no.
So the item was built by hand, one statement at a time in the interface. Things went slower and felt a bit annoying. It took about twelve edits to finish, which is enough that the tool is ready for the next project. If you plan on doing this work, set up the account weeks before you need it.
Step four: what went into the entity
Ten statements. The main idea is to split what can be proven from what is interpretation, and to stay careful with the interpretation.
- Instance of — a news website and a podcast show. Two values on one property. An entity is allowed to be more than one thing.
- Country, headquarters location, official website, and language of work.
- Apple Podcasts ID, Spotify show ID, YouTube channel ID, and Facebook username. These hold the real proof.
- Inception — a year, at year precision. Not a full date.
That last one matters more than it looks.
People often feel like they can find the launch date by looking at the oldest file dates on a site. But those are modification dates, not publication dates. On our site one of them turned out to belong to an entirely different owner of the domain, years earlier. So we set the year and left the precision loose rather than claim something we could not defend. A wrong fact in a public database is worse than a vague one, and much harder to walk back.
Step five: make the site agree with the entity
An entity in an outside database is only half the work. The site also has to say the same things, in a form machines can read. Three fixes did most of it, and the first was a mistake we had been shipping for months without noticing.
The publisher was declared as a Person. The schema listed the publisher as both an Organization and a Person, using a person-style identifier. So every article on the site was naming an individual as its publisher. Google does not accept a Person as an article publisher. Our SEO plugin prints that warning on its own settings screen, which is a slightly humbling place to find it out. We rebuilt it as a proper Organization with a consistent name, location, logo, and description.
The sameAs array was empty. sameAs is where you declare your official profiles. It tells Google that these accounts are the same entity as this website. Ours had nothing in it. We filled it with six URLs: three social, three podcast platforms.
None of those profiles were linked anywhere a crawler could see. This is the step most people miss. A sameAs claim inside your markup is one side making an assertion. Google corroborates it against links that are actually visible on the page. So the same six URLs went into the footer as real server-rendered links, matched character for character against the markup: no short links, no tracking parameters, no redirects, trailing slashes identical. That last part sounds picky. It is not. A mismatch breaks the corroboration you did all the work to create.
- Worth planning early: use the same handle on every platform you can. Handle consistency across platforms is itself a corroboration signal, and it costs nothing if you decide it when you create the accounts instead of three years later.
What we deliberately didn’t do
No Wikidata item for the founder as a person. If someone is not notable, their item is likely to be deleted, and a failed deletion discussion permanently attached to your name is worse than having no entry at all.
No Wikipedia article. Wikipedia has real rules about what deserves a page. Trying to make one for a small business leads to problems with editors and a bad look for the business.
No Google Business Profile, even though it would have been simple to set up. The business is online only and has no in-person contact with customers, which means it is not eligible. Filing anyway risks a suspension that is much harder to undo than simply not having a profile yet. We will file when it qualifies, not before.
The honest part: what this does, and what it doesn’t
Entity work is comprehension infrastructure. It changes how accurately machines can describe you. On its own, it will not make you rank.
A month after the build, organic traffic had not moved. Not even a little. That is normal. Organic traffic needs links and content to grow, and the site did not have either yet.
But one thing did change. The entity now resolves. Google, and the AI systems reading the same structured sources, can answer “what is this?” with facts we supplied, instead of guesses assembled from whatever was lying around.
That distinction should drive your sequencing.
If your problem is that nobody links to your site, do this second. If your problem is that AI assistants describe your business wrongly, credit your work to a competitor, or cannot surface you at all when someone asks for options in your category, do this first. The same infrastructure decides whether you appear in AI-generated answers, and that is a fast-growing way people find local businesses.
If you want to run this yourself
- Find your notability anchor. This is the part of your business that sits in a database you do not control.
- Make a personal Wikimedia account, disclose the conflict of interest, and do it several weeks before you need it.
- Audit your current schema before adding anything. Check that the publisher is not a Person and that the sameAs array is not empty. Ours was wrong on both counts.
- Make your visible links match your markup exactly. Character for character.
- Stay conservative with facts. Vague and right beats specific and wrong.
- Expect gains in comprehension, not in rankings. Plan the rest of your work around that.
Where this fits
Entity identity is what connects the schema markup describing your business to machines and the question of whether you show up in Google’s AI Overviews. Schema is the way things get written down. The entity is what is being written about. AI Overviews is where the two produce something a person actually sees.
If you want someone else to handle this — the audit, building the entity, rebuilding the schema, and checking the work — we offer it as a service. It does not happen quickly. It is not a magic fix for rankings, and we would rather say so now than have you find out in month three.
