On this page
What you are building
To build a Google Knowledge Panel yourself, publish one page about you on a domain you own, add Person code that matches it, make every profile say the same thing, and earn outside pages that confirm it. Google shows a panel when those sources agree. You cannot apply for one, and nothing below forces Google's hand. It removes the reasons Google has to doubt you.
- Source page
- The page on your own domain that states who you are. Everything else copies from it.
- Person code
- A block of structured code, in a format Google publishes, that describes the source page in a way a machine can read without guessing.
- sameAs
- The list inside that code of every profile that is you. It tells Google these accounts belong to one person.
- Drift
- When two profiles describe you differently: an old title, a different spelling, another photo. Drift is what makes Google hesitate.
| Layer | What you do | Hours |
|---|---|---|
| 0. Baseline | Search your name signed out, ask three AI assistants, save everything with the date | 2 |
| 1. Source | Write the identity sentence, build the page about you | 9 |
| 2. Description | Add the Person code, list every profile, test it | 4 |
| 3. Sync | Rewrite every profile to match, run the drift check | 10 |
| 4. Confirmation | Earn the first round of outside pages that name you correctly | 15 |
| Claim and monitor | Claim from your own Google account once a panel shows, re-run the baseline monthly | 2 |
Put in the hours you can actually protect each week. Be honest: the date is only useful if it is true.
Layer 0: take the baseline before you touch anything
Engineers measure before they change a system, and the same rule holds here. Open a private window, search your full name, and screenshot page one. Then ask ChatGPT, Claude and Perplexity the prompt below, and save each answer with the date. In three months you will want to know what changed, and memory will not tell you. Our guide on taking a baseline before you automate explains why this step gets skipped and why that costs later.
Who is [Full Name], [what they do, in a few words]?
List every fact you state about them, and for each fact name the page you got it from.
If you are not sure who I mean, say so instead of guessing.Read the answers for three things: facts that are wrong, facts that belong to someone else with your name, and facts with no source at all. Each one points to a page in your record that needs fixing or creating.
Layers 1 and 2: the source page and its code
Write one sentence that says who you are and what you do, for whom. This is your identity sentence, and you will paste it onto every profile for years, so get it right once. Then build a page on a domain you own (your own name is best, a page on your company site also works) with that sentence as the first paragraph, your photo above the fold and links to every profile that is you.
Then add the code. Copy the block below into the page, replace every placeholder, and delete any line you cannot back up with visible text on the same page. That last part matters more than any field: code that claims more than the page shows is the quickest way to be ignored. If structured output is new to you, our guide to getting structured output from language models explains the same idea from the other direction.
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "ProfilePage",
"mainEntity": {
"@type": "Person",
"@id": "https://yourname.com/#person",
"name": "Your Name",
"jobTitle": "Founder and CEO",
"description": "The same identity sentence you use on every profile.",
"url": "https://yourname.com/",
"image": "https://yourname.com/headshot.jpg",
"worksFor": { "@type": "Organization", "name": "Company Name", "url": "https://company.com/" },
"sameAs": [
"https://www.linkedin.com/in/yourname/",
"https://www.youtube.com/@yourname",
"https://x.com/yourname"
]
}
}
</script>Test it in Google's Rich Results Test before you move on. A clean result means Google can read it. It does not mean Google believes it, which is what the next two layers are for.
How search and personal branding connect, and why your own page is the anchor for both.
Layer 3: sync every profile, then check for drift
Paste every profile link that is you into the checker below. It flags duplicates, missing major profiles and links that will not work in the code. Then open each profile and rewrite it from the source page: same name string, same photo file, same opening sentence. LinkedIn, YouTube, X, GitHub, Crunchbase, conference speaker pages, author pages, anything that carries your name and that you control.
Paste one link per line. The check runs in your browser and nothing is sent anywhere.
When the profiles are done, run the drift prompt with the text of each one. It is tedious, and it catches the old job title on the profile you forgot existed, which is the one Google tends to find first.
Here are my profile texts, one block per site.
Compare them line by line. List every place where the name spelling, job title,
company name, city or the first sentence differs between sites.
Do not rewrite anything. Only list the differences, as a table: field, site, what it says.If you use a middle initial, use it on every profile or on none. Google treats small spelling differences as a reason to wonder whether two pages describe the same person.
Layer 4: the part code cannot do
Everything so far is under your control. The confirmation layer is not, and it carries the most weight. Google wants independent sites that say who you are in their own words: interviews, podcast pages, conference listings, guest articles, industry directories, a client's case study that names you. Ahrefs' study of 75,000 brands found mentions across the web track AI Overview visibility far more closely than backlinks do, which is the same lesson from a different angle.
Aim for a first round of five to eight pieces placed where your field actually reads. Each one should use your exact name string and, wherever the host allows, link back to your source page. Log each piece with its URL, date and the facts it states. That log becomes your record of what Google has had a chance to read.
Paid listicles of the 'top 10 founders' kind, invented awards, review swaps, and Wikidata entries for people who do not meet its rules. Never edit a Wikidata item that belongs to someone with your name. These shortcuts tend to work briefly and then undo the rest of the record.
Claim it, then keep the monitor running
When a panel appears for your name, claim it from your own Google account. Google verifies the person, so this always happens in your hands, and you need the same account for every later edit, so choose one you will control for years. The walkthrough below shows the real flow, screen by screen.
The final part of a three-part walkthrough of claiming a verified panel, recorded on the actual claim screens.
Then put the monitor on the calendar: once a month, re-run the baseline prompt, search your name signed out, and compare with the first save. Our guide on monitoring AI systems in production is written for software, but the habit is the same. A record that nobody checks drifts back to where it started.
The honest timeline
On our own record of 23 panels, built across five countries, the build itself finishes inside a quarter and the panel tends to follow within 6 to 12 months on a strong name. A DIY build can match that if the hours are real. Where it usually slows is layer 4, because outside pages depend on other people's calendars.
If it is the second column, the service page shows what we take on and the panels our team has built.