How to Get a Google Knowledge Panel in 3 Months (Jason Barnard)

Edward Sturm

Abstract

An hour on how a machine comes to understand that a person or a company exists. The answer is unglamorous: one owned page that states the facts plainly, a footprint that says the same thing everywhere, and links out to every place that corroborates it. Confidence is the currency, and clarity is how it is bought.

• The entity home is the about page, owned rather than rented, because a rented profile can be sold or taken away.
• Consistency across the whole footprint matters more than any single strong signal.
• Linking out to corroboration beats hoarding link value, and the trade is not close.
• At indexing and grounding there is no reasoning, so the conclusion has to be written out rather than implied.
• Anything self-declared has a limited ability to survive; third-party proof is what makes the understanding stable.
• What the assistants say about you is the business card the search page used to be, and it is much harder to control.

Everything below follows the conversation in order, from clarity and the entity home through page-by-page audience design and site structure, into descriptions, naming traps and capitalisation, and out through third-party proof to why a panel is worth having at all. For anyone building a personal brand the useful part is the mechanism: the machine is not judging quality, it is checking whether the same story appears everywhere it looks.

Chapter summaries

00:23 - The first factor is being clear about the facts on the entity home

The number-one factor for a knowledge panel is being clear about the facts on your own entity home — name, birthplace, birth date, chosen subtitle, description, photos.

"being clear. You have to be clear about the facts."

— 00:23

Name, birthplace, date, chosen subtitle, description, photographs — stated plainly, in one place, without contradiction. It is the least glamorous requirement in the whole process and it is the one everything else is measured against.

00:56 - The entity home is the one page an entity lives on, and it should be owned

Your entity home is the one page where you "live online" as an entity — ideally your own site because you own it, since a company or a rented profile can be left, sold or taken away.

"you're clear about that on your own entity home website, you're already well on the way."

— 00:56

Owning it is the point. A company profile or a rented platform can be left, sold or taken away, and when that happens the entity loses the place it was reconciled from. Own the ground you stand on.

01:40 - The about section has to disambiguate one person from everyone sharing the name

The entity home's about section must be very clear about who you are, what you do, who you serve, and must disambiguate you from everyone else with your name.

"entity home is what Google call um the the the place of reconciliation"

— 01:40

Who you are, what you do, who you serve — and, critically, what distinguishes you from everyone else carrying the same name. Disambiguation is a job the about section has to do explicitly, because nothing else will do it for you.

02:11 - An entity is simply a thing that can be named

An entity is simply a thing you can name — a person, company, film, album, musician, road, house, building — and topics such as economics count too.

"can you describe what an entity is for for like people who are newer in SEO"

— 02:11

Naming is the test of an entity, and the results page for a name is that entity's page. It is a standing definition, expanded for newer practitioners at the host's request — and for practical work you can stop at nameable things, with topics as the further step.

05:01 - The engine is not fussy about fame — it grants a panel when it understands

The engine is not trying to understand only the famous; it tries to understand everything and is not fussy — if it understands, it grants a panel.

"no, it doesn't matter. Google isn't trying to understand only the most famous people, entities, companies, films."

— 05:01

It is trying to understand everything, not only the famous, and it is not selective about who qualifies. Understanding is the whole condition: when the engine understands, it grants a panel.

05:34 - The machine follows the links out and checks the story matches everywhere

The machine looks first for entity homes, follows their links to check the same entity is represented consistently elsewhere, and when it is consistent it becomes confident and grants the panel.

"it goes and gets all the information it possibly can and tries to make sense of it."

— 05:34

It finds the entity homes first, follows their outbound links, and checks whether the same entity is represented the same way elsewhere. Consistency across those destinations is what turns partial information into confidence.

06:40 - The first mistake is an entity home that is neither clear nor consistent

The first mistake newer companies make is an entity home that is not clear and not consistent.

"the first thing that people get wrong is on their own entity home they're not clear and they're not consistent."

— 06:40

Two failures at once, and they compound: the page is not clear about what the entity is, and what it says does not match what the rest of the web says. Nothing downstream can repair that.

07:39 - Linking out to the corroboration beats hoarding the link value

Then link from the entity home out to all the corroboration; the three biggest mistakes are unclear entity home, inconsistent footprint, and failing to link out because people want to hoard "link juice".

"if you can get that consistency on number one that clarity and consistency on your entity home"

— 07:39

People hoard outbound links to preserve value, and the value of being understood is larger than the value being hoarded. The trade is not close — an entity the machine cannot corroborate has nothing to conserve.

09:04 - People are easier than corporations, because that is where the attention went

The engine has focused on people for the last three or four years, so getting a panel is relatively easy for people and relatively hard for corporations.

"I'll just give you a caveat there is Google have focused on people for the last three or four years."

— 09:04

Three or four years of attention went to people, which makes a personal panel comparatively easy and a corporate one comparatively hard. It is a temporary asymmetry worth exploiting while it lasts.

09:19 - A clean new footprint resolves faster than a large messy one

A clean, new footprint is faster than a large messy one: a new company took four to five months, an established one nine months.

"we recently had an experience with a a company that was well established. It took 9 months"

— 09:19

A big existing footprint has to be de-messed before the machine can digest it, which is why starting clean is faster than starting large. Two of his own client cases are the evidence — reported from experience rather than offered as a guarantee.

10:04 - The profile pages get created at the start, while they can still be correct

Create the profile pages right at the start so they are correct from the beginning, on the platforms every new business should register on: a company-data platform, a professional network, an org directory, the major social platforms, plus at least one review platform.

"We were we were creating the the the the the profile pages so we could create them right at the"

— 10:04

Correct from the start is cheaper than corrected later. A company-data platform, a professional network, an organisation directory, the major social platforms and at least one review platform — created at the beginning, while nothing needs undoing.

11:28 - Today's panels are templates, and the templates are likely to be replaced

Knowledge panels are built from human-defined templates, and as things move to AI that system is likely to be replaced — future panels will be AI-generated, more dynamic, and less reliable in what they show.

"knowledge panels are built from templates that are human defined."

— 11:28

Human-defined templates are what today's panels are assembled from, and as the systems move toward generated output that arrangement is likely to be replaced by something more dynamic and less predictable in what it shows.

12:04 - Being in the graph is not the same as having a visible panel

Being in the knowledge graph is not the same as having a visible panel — you can be in the graph with no panel showing, and there are multiple graphs, so getting into the graph is the real key.

"being in the knowledge graph is not the same thing necessarily as having a knowledge panel."

— 12:04

You can sit in the graph with no panel displayed, and there is more than one graph. Getting into the graph is the real objective; the panel is a visible side effect that the engine may or may not choose to show.

13:50 - A panel is still worth getting even though this generation of them will die

Getting a panel today is still worth doing even knowing today's panels will die out, because the replacement will still be an informative panel of description, photos and videos.

"Getting a knowledge panel today is definitely something I would do knowing that knowledge panels as they exist today will die out"

— 13:50

The replacement will still be an informative panel of description, photographs and video, so the work transfers even though the artefact does not. Building for the current generation is not building for something that ends.

14:39 - Every page is designed around its audience

Design each page around its audience: the about page's audience is machines trying to understand you and people doing due diligence, so it must be clear, consistent, factual — who, what, who you serve, why credible.

"the about page, the audience is machines that are trying to understand you."

— 14:39

The about page's audience is machines trying to understand the entity, and people doing due diligence. That single decision determines its register: clear, consistent, factual, and answering who, what, who it serves and why it is credible.

15:04 - The homepage is a signpost rather than a destination

The homepage's audience is someone only vaguely interested, so it is a signposting page — a hop to the next step, never the destination.

"The audience for the homepage is somebody who's vaguely interested in you as a person"

— 15:03

Its audience is somebody only vaguely interested, which makes it a hop rather than a landing place. Treating the homepage as the destination is the mistake — its job is to send people to the page that actually answers their question.

15:38 - The media kit is the human document, written for someone about to hire or write

The media kit's audience is someone who wants to hire you or write about you, so it is very human, adjective-rich and duplicates some about-page and homepage information.

"somebody who wants to hire you or wants more information to write information, write articles about you."

— 15:38

Each artefact is shaped by who reads it: the media kit sells where the about page states. It is described for hirers and journalists specifically, which is what makes it the human document of the set.

16:35 - The advanced move is a pillar page for every facet

The advanced move is a pillar-page system: each facet of a person or company gets its own page — talks given, podcasts appeared on, books written, patents, academic papers.

"which is what I'm calling the pillar page system where each aspect of a person"

— 16:35

Each facet gets its own page: talks given, appearances made, books written, patents held, papers published. It scales the entity without cluttering the one page that has to stay simple.

17:02 - The entity home holds the facts and links out to each facet

Keep the entity home to the fundamental facts and link out to each facet page and back, so each facet is a mini entity home reachable from the top menu.

"the reason for that is if I try and put all of that on the entity home, that single about page,"

— 17:02

The entity home keeps the fundamental facts and links out to each facet, and each facet links back. Every facet page becomes a small entity home in its own right, reachable from the top menu.

17:57 - The top menu is not the site structure

The top menu is not the site structure, because URLs do not map to the menu and the ontology that organises the information is a separate thing from the navigation that matters.

"thing that I tell my clients a lot, uh, and I've been telling them for years and years"

— 17:57

URLs do not map to the menu. The structure is ontological — how the information is organised — while the menu is what people and machines navigate. Auditing the menu usually turns up something present that should not be, or missing that should.

19:21 - Everything belongs in a category, and only the homepage sits at the root

Put everything in a category — only the homepage belongs at the root; a flat architecture is the wrong default.

"my number one rule is everything in a category."

— 19:21

Only the homepage belongs at the root. A flat architecture is the wrong default, and the rule is absolute enough to be applied without case-by-case argument, which is most of its value.

20:59 - The second factor is linking to every corroboration and keeping the links alive

The second most important factor is linking to every corroboration and keeping those links in place over time, because the web degrades.

"Linking to them all. Making sure those links are in place from your website out to each and every corroboration"

— 20:59

The web degrades: links rot, profiles move, platforms close. Linking to every corroboration is the second factor, and keeping those links alive over time is inseparable from it.

21:39 - The self-corroboration loop is the fastest route to confidence

The infinite loop of self-corroboration — entity home to a platform and straight back to the entity home — is the fastest way to make the engine confident it understands you.

"builds the pathways for the bot to follow. And it's what I call the infinite loop of self-corroboration."

— 21:39

Entity home out to a platform, platform straight back to the entity home. The loop gives the crawler a path it can complete, and completion is what produces confidence. The shortest route is a closed one.

22:04 - The "same as" declaration must never point at a contradicting profile

New sites should use "same as" schema, but the common mistake is pointing it at a URL that disagrees with the entity home — don't link to a profile you can't control that contradicts you.

"Are you using same as schema? >> Yes."

— 22:04

The declaration is standard practice for a new site and the common failure is pointing it at a profile that disagrees with the entity home. Never declare sameness with something you cannot control and that contradicts you.

22:50 - The entity home is the about page, not the homepage

The entity home is the about page, not the homepage, because the homepage is a signpost that says "look how wonderful we are" and that conflicts with a dry factual definition.

"Is the entity home always the homepage for the main website? >> No, it's the about page."

— 22:50

The homepage is a signpost that says how wonderful the business is, and that register conflicts with a dry factual definition. The about page is where the facts live, which makes the about page the entity home.

24:17 - Transparency is a different thing from clarity

Transparency is distinct from clarity: clarity is being clear about what you are, transparency is being verifiable — contact details, tax and registration numbers, ways to confirm you are a real company.

"that's a something that people fail to or often fail to to implement is transparency."

— 24:17

Clarity is being clear about what you are; transparency is being verifiable. Contact details, registration and tax numbers, ways to confirm a real company exists behind the site — those are transparency, and they are a separate obligation.

25:37 - Keywords never get aimed at the homepage

Keywords are never aimed at the homepage, because the homepage is the brand page and should rank first for the brand itself.

"do you have thoughts on whether or not brands should be targeting keywords with the homepage"

— 25:37

The homepage is the brand page and should rank first for the brand, which is what visitors expect to find there. Aiming keywords at it competes against that job rather than adding to it.

26:16 - A company named after what it does loses to the generic term

When your company name is also what you do, the engine treats the generic term as the dominant interpretation rather than you, and ranking first for a generic word is a huge job.

"So if your company is called um Barn >> Austin Plumbers"

— 26:16

The generic term becomes the dominant interpretation and the business becomes the minority reading. Ranking first for a generic word is an enormous undertaking, and the name chose that fight before anyone else did.

27:32 - An invented brand name scales better than a descriptive or geographic one

A unique, invented brand name scales better than a descriptive or geographic one, because a real brand is easier to grow than a name tied to a trade or a city.

"perspective on on naming a company is I named"

— 27:32

A real brand grows in a way a trade-and-city description cannot. The invented name has no incumbent meaning to displace, which is why it scales — distinctiveness is cheaper than displacement.

28:56 - Ambiguity is the biggest problem, in both directions at once

Ambiguity is the biggest problem: with a common name the machine must both disambiguate you from everyone sharing it and merge your own multiple facets into one person.

"100% I don't and then you have to think of another business model."

— 28:56

The machine has to do two opposite jobs at once: separate you from everyone sharing your name, and merge your own several facets into one person. With a common name both jobs get harder simultaneously.

31:22 - Which of two same-named entities shows comes down to three things

Which of two same-named entities shows depends on three things: famousness/notability, relevancy to the searcher, and the engine's confidence in its information.

"The first would be famousness."

— 31:22

Famousness, relevance to the searcher, and the engine's confidence in what it holds. Click-through rate plays no part in it — with entities the question is confidence, not behaviour.

32:57 - One executive summary sized to the tightest platform solves the description problem

Manage per-platform descriptions with one executive summary of about 150 words that fits the tightest platform, then append the longer description wherever a platform allows it.

"platforms because there's rules for like how long each description can be."

— 32:56

One executive summary of roughly a hundred and fifty words, sized to fit the tightest platform, then extended wherever a platform allows more. One artefact solves a problem people usually solve platform by platform.

33:48 - The description is built modularly and trimmed per platform

Build the description modularly — executive summary plus heading-and-paragraph chunks, each covering one facet — then on each platform remove the chunks that do not fit that audience.

"we've done a modular system as well where if you create it with multiple headings"

— 33:48

An executive summary plus heading-and-paragraph chunks, each covering one facet, so that fitting a platform becomes a matter of removing the chunks its audience does not need. Assembly rather than rewriting.

35:19 - At indexing and grounding there is no reasoning

The biggest current mistake: at the moments of indexing and grounding there is no reasoning — the engine uses the cheapest, "pretty stupid" model that only digests and categorises.

"I'll give you the biggie."

— 35:19

At the moments of indexing and grounding there is no reasoning at all — the cheapest model runs, and it digests and categorises rather than infers. That single fact explains most of what follows.

36:41 - So the reasoning gets spelled out rather than left to be inferred

So do the reasoning for it — spell out the conclusion ("...has done this since 1998, therefore is an expert") rather than leaving it to be inferred, and it gets indexed.

"So you need to do the reasoning for it."

— 36:41

So the conclusion gets written out: the years of practice, therefore the expertise. Leaving the inference to be drawn means it is not drawn, because nothing at that stage is drawing inferences.

37:38 - When the name does not say what the company does, the description has to

When the brand name doesn't say what you do, you must say it — "[Brand] is an electrical services company in [city] serving households since [year]" — naming the audience and giving history and a reason to trust, then repeat it across every platform.

"I'll give you a brand name. You're called um Palm Tree Wires."

— 37:38

The description carries what the name does not: what the company does, where, for whom, since when, and why that is a reason to trust it. Then the same description is repeated across every platform.

38:32 - The description opens with a semantic triple

Start the description with a semantic triple — subject, verb, object — then extend to who you serve, where, and why to trust: name, what you do, who you serve, why trust.

"Palm Wires is an electrical company. >> That's a way better name."

— 38:32

Subject, verb, object — then extended outward to who is served, where, and why to trust. The triple is the load-bearing part, and everything after it hangs off a sentence the machine can parse cleanly.

39:21 - Naming related entities improves performance, because the machine already knows them

The more related entities you name, the better you perform — a city is a place entity, a customer type is a concept entity — because naming things the AI already knows lets it associate your meaning.

"the more you name related entities, the better you will perform."

— 39:21

A city is a place entity; a customer type is a concept entity. Naming things the machine already knows lets it locate your meaning in relation to them, so the density of related entities is itself a performance factor.

40:03 - A word that is both verb and noun confuses the indexing model

A word that is both a verb and a noun is hugely confusing to the indexing model, which struggles to tell which sense you mean.

"I found something really interesting the other day as well is using a a verb that can also be a noun"

— 40:03

The indexer struggles to tell which sense is meant, and the ambiguity propagates into everything built on top. It is an avoidable problem that is usually created at the naming stage and paid for permanently.

42:01 - A name that doubles as an activity hits a capitalisation trap

A company named after what it does hits a capitalisation trap — at the start of a sentence the capital is read as sentence case, so the proper noun is lost; you may need to force it ("at [Brand]").

"the problem that companies have when they've called themselves what they do, like Car Sales Limited,"

— 42:01

At the start of a sentence a capital letter reads as sentence case, so the proper noun disappears into the activity. Forcing the construction — putting a preposition in front of the name — is sometimes the only fix.

44:17 - Anything self-declared has a limited ability to survive

Anything self-declared has limited ability to survive: you can self-declare your way to a panel through your own site and profiles, but it will not be stable or authoritative.

"So anything self-declared has limited ability to survive, let's say."

— 44:17

Self-declaration is unverified, so the engine holds it loosely however much of it there is. Directories, your own site and rented profiles all fall into that class — you can reach a panel this way, and it will not be durable.

44:51 - Third-party proof is what makes the understanding stable

Third-party proof — other media writing about you — makes the panel and the machine's understanding significantly more stable and authoritative.

"As soon as you get third party proof uh other media writing about you,"

— 44:51

Other media writing about you changes the character of the understanding, not just its volume. The panel becomes stable and the machine's confidence becomes durable, which is what self-declaration cannot produce.

45:25 - A podcast appearance plus show notes is third-party corroboration

A podcast appearance plus a show-notes description is third-party corroboration: the host vouches that you exist and are the specific person, and each time you then repeat that description it carries more weight because someone else said it first.

"one thing is coming on your podcast and then you write a a show notes"

— 45:25

The host vouches that you exist and are that specific person, and the show notes carry the description. Each time the description is repeated afterwards it carries more weight, because somebody else said it first.

46:53 - Two footprint strategies both work: small and controlled, or large and self-sustaining

Two footprint strategies both work: keep it small and tightly controlled, or "splurge" on a big uncontrollable presence that then sustains itself.

"experiment uh the other day when I was thinking about to what extent machines reason."

— 46:53

One is small and tightly controlled; the other is a large presence you cannot police, which then sustains itself. Both work, and the choice is about temperament and capacity rather than correctness.

48:43 - The first benefit is a brand-search page that works as a business card

The first benefit of a panel is the brand-search page as your "online business card" — people look at it before your website, and an impressive panel makes them assume you are important and authoritative.

"the first one is Google brand search. The search engine results for your name is your online business"

— 48:43

People look at the search page before they look at the website. An impressive panel makes the reader assume importance and authority before a single claim has been read, which is what a business card is for.

49:44 - A panel stops the leak at the bottom of the funnel

A panel stops leakage at the bottom of the funnel, which is your biggest enemy — you can lose a deal you had already won because of what the search page shows.

"it gets people over the line when they're trying to decide"

— 49:44

Losing a deal you had already won is the expensive failure, and it happens at the last check. A panel closes that leak, which makes it a bottom-of-funnel asset rather than an awareness one.

50:56 - A panel also signals that the machine is confident, which improves what it says

A panel also signals that the AI understands you and is confident, so the answers AI assistants give about you are more accurate, more stable and more positive.

"will help to stop that. Another thing is ego makes you feel good about yourself."

— 50:55

Demonstrated understanding upstream produces better and steadier machine answers downstream. What assistants say about a well-understood entity is both more accurate and more stable — and the stability matters as much as the content.

51:55 - The due-diligence rabbit hole is the new brand-search page

The "AI resume" or due-diligence rabbit hole is the new equivalent of the brand-search page: an AI assistant invites a conversation, suggests follow-ups and sends people deeper, so that path must represent you as you want all the way down.

"the last thing that I wanted to add here is the AI rumé rabbit hole"

— 51:55

An assistant invites a conversation, suggests follow-ups and sends people deeper, which means the path has to represent you the way you want it to all the way down. It is the new version of the brand-search page, with more surface area.

52:42 - What the assistants say is the business card now, and it is harder to control

What AI assistants say about you is now the business card the search page used to be, and controlling that is much harder than controlling a brand page — a panel is a sign you are positioned to take control.

"now your AI resumeumé is what"

— 52:42

Controlling what the assistants say is much harder than controlling a brand page ever was. A panel is a sign of being positioned to take that control rather than a guarantee of having it.

53:12 - A panel missing from a search does not mean it does not exist

A panel not appearing where you search does not mean it does not exist — the engine may simply be choosing not to show it; you have to check the graph directly.

"I searched a bunch of brand names just now and so this was crazy."

— 53:12

The engine may simply be choosing not to display it. Checking the graph directly is the only way to know, and the absence of a panel on a search is weak evidence about anything.

54:31 - The engine decides what a great result looks like for each query

For any query the engine decides what a great result looks like — videos, images, news, a social feed or a knowledge panel — and can choose to omit the panel.

"I'm just doing a typical a typical Google search"

— 54:31

For each query it decides what a good result looks like — video, images, news, a social feed, a panel — and a panel can lose that competition for reasons having nothing to do with you.

55:52 - A panel can be withheld for size, for confidence, or for an update in progress

A panel may not show because it is too small to be worth the space, because the engine is not confident the term means that company even at rank one, or because a knowledge-graph or panel update is in progress.

"if you look at the the knowledge panel I just sent you"

— 55:52

Value, confidence and update state each independently gate whether the module is shown at all. A large software brand can rank first and still show an unremarkable or absent panel — and the explanations are offered partly as informed guesses.

56:56 - The closing workflow is a few months long and entirely mechanical

The closing workflow to a panel in a few months: follow this episode; use a text-analysis sandbox to check which entities the engine recognises and rewrite until your name and company are recognised at the top; track appearance and a confidence score in a graph explorer; generate simple schema.

"this is my last question for the 15 to 30 minute episode that I pitched you on"

— 56:56

Writing against the engine's own recognition and then verifying in the graph closes the loop between drafting and confirmation. A sandbox runs the engine's natural-language API, and a knowledge-graph explorer returns a confidence score for what was written.

58:18 - The trusted-sources list has nothing to do with domain authority

Use a "trusted sources" list — the sources the knowledge graph trusts, which has nothing to do with domain authority — and getting a profile on one helps; start with the obvious authoritative directories.

"you can look in the trusted sources list."

— 58:18

The sources the graph trusts are not the sources with high domain authority, and conflating the two sends effort in the wrong direction. Starting with the obvious authoritative directories is the practical entry point.

59:14 - Schema on the entity home is supporting evidence rather than a requirement

Schema on the entity home is helpful but not necessary — it is supporting evidence, not the be-all; the simplest solid schema is best, and in a common CMS you can paste it into an HTML block mid-page where the machine reads it but people don't see it.

"If you add schema to the entity home, that's helpful, but it is not necessary."

— 59:14

It is supporting evidence, not the requirement people treat it as. The simplest solid version is the best one, and in a common content system it can be pasted into a block mid-page, where the machine reads it and nobody sees it.

Personal Branding Lessons

An episode about being legible to a machine that does no reasoning at the moment it reads you. The moves below are all about removing the need for it to infer anything.

Own the entity home rather than renting it

The one page where an entity lives online should sit on a domain you control. A company profile or a rented platform can be left, sold or taken away, and when that happens the entity loses the place everything else was reconciled from. Owning it is not a preference — it is what makes the rest durable. 00:56

Keep every mention of the name identical across the whole footprint

The discipline a local business applies to its name, address and phone number applies to everything else too. Every mention of the brand or the person, identical everywhere the machine can find it. Variation reads as uncertainty, and uncertainty is precisely what stops the confidence accumulating. 07:11

Link out to every corroboration instead of hoarding link value

People withhold outbound links to preserve something, and the value of being understood is larger than the value being withheld. The trade is not close. An entity the machine cannot corroborate has nothing worth conserving in the first place, and the links also have to be kept alive as the web degrades. 07:39

Design each page around one audience

The about page is written for machines and for people doing due diligence, so it is clear and factual. The homepage is written for someone vaguely interested, so it signposts. The media kit is written for someone about to hire or write about you, so it is human. Read the first paragraph of any page and the audience should be obvious. 14:39

Put everything in a category and leave only the homepage at the root

A flat architecture is the wrong default. Every page belongs inside a category, and the only page at the root is the homepage. The rule is blunt enough to apply without arguing each case, which is most of what makes it useful. 19:21

Write one executive summary sized to the tightest platform

Roughly a hundred and fifty words, built to fit wherever the limit is strictest, then extended where a platform allows more. Build it modularly — a summary plus heading-and-paragraph chunks, one per facet — so adapting it to a platform means removing chunks rather than rewriting. 32:57

Spell out the conclusion rather than leaving it to be inferred

At the moments of indexing and grounding there is no reasoning at all: the cheapest model runs, and it digests and categorises. So the conclusion gets written out — the years of practice, therefore the expertise — because an inference left to be drawn is an inference that does not get drawn. 36:41

Ask for a factual description rather than praise

People are uncomfortable asking others to describe them a particular way, and the discomfort comes from imagining they are asking for praise. Asking for a plain factual description is a small request, and the goal is consistency rather than flattery — the same words, from other people, in other places. 46:48

Questions

Each answer ends at the moment in the recording where it is given.

How do I get a knowledge panel for a new brand, and how long does it take?

Be clear and factual about who you are, what you do and who you serve on a page you own, keep every mention of yourself identical across the web, and link that home page out to all the corroborating profiles. A clean new footprint built correct from the start is reported to reach a panel in roughly four or five months, while a large messy one took around nine — the delay is untangling the mess, not the size. It is offered as experience, not a guarantee. 00:23

What is an entity home and where should it live on my site?

Your entity home is the one page where you live online as an entity — the place the search engine treats as its point of reconciliation for everything it finds about you elsewhere. It should be your about page, not your homepage: the homepage exists to welcome and signpost, and that celebratory job conflicts with a dry factual definition. Keeping them separate stops you fighting yourself over the message, which is really about nailing a clear statement of who you are. Ideally it sits on a site you own, since a rented profile can be closed or sold. 00:56

Which platforms should a new business register on to get a knowledge panel?

A small, specific set covers most of it: a company-data platform, a professional network, an org directory, the major social platforms, and at least one review platform — whichever one is standard in your country. The advantage of a new brand is that these can be created correct from the very beginning, before any cleanup is needed. Held consistent with your entity home, that handful plus your own page is enough to reasonably expect a panel within a few months, though where you start depends on where your audience actually is. 10:04

What is the difference between the knowledge graph and the knowledge panel?

The panel is the visible box; the graph is the stable understanding beneath it. You can sit in the graph with no panel showing, and you can even have a panel without being in one particular graph, because there are several. Getting into the graph — marked by a unique identifier that means the engine has understood you and is confident — is the real goal. Today's panels are built from human templates and are expected to be replaced by AI-generated ones; the graph underneath is what endures. 12:04

How should my about page, homepage and media kit each be different?

Each is built for one reader. The about page's audience is machines trying to understand you plus people doing due diligence, so it is clear, consistent and factual — who, what, who you serve, why credible. The homepage's audience is someone only vaguely interested, so it is a signpost to the next step, never the destination. The media kit's audience is someone who wants to hire you or write about you, so it is human and adjective-rich. The test: read the first paragraph and you should know exactly who it is for. 14:39

Should my company name describe what I do, or be a unique brand?

A unique, invented name scales better than a descriptive or geographic one. When your name is also what you do, the engine treats the everyday word as the dominant interpretation rather than you, and ranking first for a plain generic term is an enormous job only globally dominant brands have pulled off. A trade-plus-city name also cannot follow you to another city without becoming confusing. A coined word has no everyday meaning to fight and no geographic ceiling, which is a real way to stand apart rather than blend in. 26:16

How do I write an entity description an AI can understand?

At indexing time the engine does little or no reasoning — it uses the cheapest model, which only digests and categorises — so anything that needs inference is lost. Do the reasoning for it: state the conclusion explicitly, as in a track record "therefore an expert", rather than leaving it implied. Open with a semantic triple — name, what you do, object — then extend to who you serve, where, and why to trust you. Name related entities the machine already knows, like a city or a customer type, to give it footholds. 35:19

Do podcast appearances and press really help my knowledge panel?

Yes, and they do something your own pages cannot. Anything self-declared reaches a panel but stays fragile and unauthoritative. A third party describing you — a host's show notes, an article — is corroboration you do not control, vouching that you exist and are the specific person, and every time you then repeat that description it carries more weight because someone else said it first. Asking to be described with a plain factual label is a small ask, and consistency is the whole goal. 44:17

Why does my knowledge panel not show up when I search my name?

Not showing does not mean not existing — the engine may simply be choosing not to display it, so check the graph directly before assuming it is gone. For every query it decides what a great result looks like and assembles modules — videos, images, news, a panel — and can omit any of them. A panel might be withheld because it is too small to be worth the space, because the engine is not confident the term means that company even at rank one, or because a graph update is in progress. 53:12

Sources

How to Get a Google Knowledge Panel in 3 Months (Jason Barnard)

The sections above follow the episode's own order. Each timestamp is the point in the recording where the idea is discussed, so the post reads as a map of the conversation rather than a rearrangement of it.

Quotations come from the episode transcript with spoken filler ("okay", "right", "you know", "um", "actually", "kind of", "sort of") and false starts removed. No words are added: square brackets mark anything inserted for sense, and an ellipsis marks any cut. British spellings are restored, and speech-to-text errors were repaired against context only where the intended word is unambiguous. Where a line appears both in a cold-open teaser and in its place in the conversation, the timestamp cites the second — the point at which it is actually said.

Quotes are stamped with the moment they are said rather than with a speaker's name.