ChatGPT has never heard of your brand. Entity SEO is how you make it remember.

Open ChatGPT right now. Type the name of your business. Just the name.

If it comes back with "I could not find specific information about that company", you do not have a content problem. You do not have a backlink problem. You have an identity problem. The machine does not know you exist as a thing. You are a string of characters to it, not a business.

That is what entity SEO fixes. And most of the people selling it to you are selling the wrong half of it.

What entity SEO actually is

An entity is a specific, distinct thing that a machine can point at and say "that one". Apple the company. Apple the fruit. Blaydon the town. Your business, if you have done the work.

Keywords are words. Entities are things. Search engines stopped being pure word matchers years ago. Google's own documentation talks about using Organization markup to "disambiguate your organization in search results", which is a polite way of saying: we need to know which one you are, because there are probably fourteen other businesses with a similar name and we are not guessing.

So entity SEO is three jobs, in this order:

1. Existence. The machine knows you are a thing.
2. Disambiguation. The machine knows which thing you are, and does not confuse you with the plumber in Ohio who shares your name.
3. Association. The machine knows what you are for, so when someone asks for that, your name is in the sentence.

Nobody skips straight to three. Everybody tries.

Why this matters far more in 2026 than it did in 2023

Because the answer has no page two.

In a classic search result you could scrape a living from position seven. In an AI answer you are named or you are absent. There is no consolation ranking. And the thing that decides whether you get named is not what your homepage says about you. It is what the rest of the internet says about you.

Ahrefs put numbers on this. They analysed 75,000 brands and measured which search factors correlated with brand mentions in Google AI Overviews, using Spearman correlation. Here is what came back.

Read that top row again, then read the bottom half. Mentions of your brand name across the web correlated three times harder than the number of backlinks pointing at you.

Two more numbers from the same study that should make you sit up. Brands in the top quartile for web mentions averaged 169 AI Overview mentions. The quartile below them averaged 14. And 26% of the brands studied had zero AI Overview mentions at all.

Ahrefs were straight about the caveat and so will I: correlation is not causation. These are relationships, not proof of a lever. But when the top three factors are all off-site and all about your name being spoken elsewhere, the direction of travel is not subtle.

The bit nobody selling you "entity SEO" wants to read out loud

Google published a guide on optimizing for generative AI search. In it, they list things you can stop worrying about. Two of them land directly on this topic.

One: "Structured data isn't required for generative AI search, and there's no special schema.org markup you need to add." They still recommend using it, because it makes you eligible for rich results. But schema is not the magic switch.

Two, and this one is sharper: they explicitly warn against "seeking inauthentic mentions" across the web, and point out that their spam systems exist precisely to deal with that.

So put the two datasets together and you get the honest version. Real mentions correlate hardest with AI visibility. Fake mentions are a spam policy waiting to fire. Schema will not save you, but it will stop the machine mixing you up with someone else. That is a narrower job than the hype implies, and it is still worth doing, because it is free and it takes an afternoon.

Anyone telling you a JSON-LD block will get ChatGPT recommending you by Friday is selling you a dream. We say the quiet part out loud, which is why we called llms.txt a five minute nice-to-have instead of a religion.

The entity stack: build it in this order

Order matters. Doing step five before step one is how people burn six months and nothing moves.

1. Pick one name and never deviate

One legal name. One trading name. Spelled the same way, cased the same way, everywhere. "The KillerEdge", "Killer Edge" and "KillerEdge Ltd" are three entities to a machine until something tells it otherwise.

Same for your address, your phone number, your founder's name and your one line description. Write those five things down in a document. That document is now the only source of truth for anyone who touches your web presence. Sounds trivial. It is the single most common reason a business has no knowledge panel.

2. Build an entity home

One page that is unambiguously about you. Usually your homepage or your about page. It states your name, what you do, where you are, when you started, and who runs it, in plain text a human can read. Not a hero image with the facts buried in a PDF.

The AI systems do not have a secret channel into your business. They read the page. Give them something to read.

3. Add Organization schema with sameAs

Google's docs are clear that there are no required properties for Organization markup, and that you should add as many relevant ones as you can. They recommend placing it on your home page or a single page describing your organization, not on every page of the site.

The properties that do the identity work: name, alternateName, url, logo, description, address, telephone, foundingDate, legalName and above all sameAs.

sameAs is the one people skip. It is a list of URLs on other sites that are also about you: your LinkedIn company page, your Companies House or state registry listing, your YouTube channel, your X profile, your Crunchbase entry, your Wikidata item. Every one of those is you saying "and that is me too". That is how a machine resolves ambiguity.

Use the most specific schema subtype that fits. Ecommerce should use OnlineStore. A physical business should use the right LocalBusiness subtype. If your business has a front door and a postcode, read our schema markup guide and then go and fix your local setup properly, because entity signals and map pack signals are the same signals wearing different hats.

We walk through this exact JSON-LD block, line by line, with a fill-in template, inside the Dojo.

4. Claim and verify the profiles you already have

Google Business Profile. LinkedIn company page. YouTube. Companies House or your national registry. Your industry bodies. Trustpilot or your sector's equivalent. Each one is a corroborating source that repeats the same five facts from step one.

Then, if you have already got a knowledge panel showing, claim it. Google lets verified representatives of an entity suggest changes to their panel. Most business owners never look.

5. Get a Wikidata item, if you legitimately qualify

Wikidata is the structured sibling of Wikipedia, and it is read by Google's Knowledge Graph. Its notability bar is lower than Wikipedia's: an item can refer to "an instance of a clearly identifiable conceptual or material entity that can be described using serious and publicly available references".

Lower is not zero. You need real, citable references for the statements you add. A company registration record, press coverage, an official filing. If you cannot cite it, do not add it. Wikidata editors remove unsourced promotional items, and getting your item deleted is a worse outcome than never creating one.

This is the step where honest businesses hesitate and chancers charge in. Be the honest business.

6. Go and earn the mentions

Everything above is plumbing. This is the part that actually correlates with getting named.

Mentions, not links. Unlinked brand mentions do very little for classic SEO and a lot more for AI visibility, because language models learn a brand from words on a page and the company those words keep. So the plays are:

• Get into the listicles. "Best [what you do] in [your town]" roundups are what AI answers quote. Getting into someone else's list beats writing your four hundredth blog post.
• Be genuinely useful in forums. Reddit and niche communities get cited constantly. Answer properly, disclose who you are, never drive-by spam.
• Do actual digital PR. Real data, real commentary, real journalists. Our digital PR playbook covers the version that works without a retainer.
• Podcasts and video. Transcripts get indexed. Your name gets spoken next to your topic. That is association, step three, happening in public.
• Directories and industry bodies. Boring. Effective. Most rivals cannot be bothered.

This is the bit agencies charge $2k a month for, and about seventy percent of it is a spreadsheet, a shortlist and a polite email.

How long this takes

Nobody can promise you a knowledge panel or an AI citation on a date. Anyone who does is guessing with your money. What I can give you is the shape of the work.

Steps one to five are a week of evenings. Step six is the rest of your business's life. That ratio is the whole lesson.

What it costs

Straight talk on money. These are the ranges you will see quoted in the market, not a survey and not a quote from us.

Who this suits, and who it does not

Do this now if: you sell something people ask AI assistants about, you have a real trading history, you operate in a named place, or you have any press or third party coverage at all sitting unused.

Deprioritise it if: you are pre-launch with no verifiable facts to publish, your entire revenue comes from one enterprise contract nobody googles, or your site is not indexed yet. Fix indexing first. An entity nobody can crawl is not an entity.

Tools that actually help

• Google Rich Results Test. Free. Tells you if your markup parses.
• Google Search Console. Free, and it now has a generative AI performance report so you can see how content performs in AI features rather than guessing.
• Wikidata's own query service. Free. Shows you what the graph already believes about your sector.
• Brand mention monitoring. Paid tools track this properly. A saved search and an alert gets you eighty percent of the way for nothing.

Google also flag this in their guide and they are right: be wary of any third party tool claiming access to internal Google ranking or AI systems. None of them have it. Our honest take on the paid options lives in The Armory.

A worked example

A single location trade business. Name appears three different ways across the site, the invoices and the LinkedIn page. No Organization schema. Companies House listing has the old address. Two local news mentions from a charity job three years ago that nobody ever linked to.

The entity work is not "write more blogs". It is: pick one spelling, fix the registry record, publish an about page that states the founding year and the owner's name in text, add Organization schema with a sameAs array pointing at Google Business Profile, LinkedIn, Companies House and the two news pieces, then ask the two publications to correct the business name in their archive copy.

None of that is glamorous. All of it is the machine going from "no idea" to "specific business, this town, this trade, these facts, corroborated in four places". That is the entire game. If you want the sentence-level version of how the answer gets built, read our guide on getting AI to recommend your business.

The mistakes that keep you invisible

• Schema on every page. Google says put Organization markup on your home page or one page about your organization. Not sitewide.
• sameAs pointing at profiles you do not control. If it is not verifiably yours, it is noise.
• Three versions of your own name. Still the number one killer.
• Buying mentions. Google's guide calls out inauthentic mentions specifically. The spam systems are not decoration.
• Creating a Wikidata item with no citations. It gets deleted and you have taught the graph nothing.
• Treating schema as the finish line. It is the label on the tin. Somebody still has to want what is in the tin.
• Facts that contradict each other. Founded 2011 on your about page, 2013 on LinkedIn. You just told the machine not to trust either.

What to do instead, if you are not doing this

The alternatives are not really alternatives, they are the other half of the same job. Classic technical and content SEO gets you indexed and retrievable, which is the precondition for everything here. Link building still drives classic rankings and drags branded anchors along with it, which was the number two correlated factor in the Ahrefs study. Paid search buys attention but the same study found branded ad traffic correlating at just 0.216 with AI Overview mentions, so do not expect a budget to buy your way into the answer.

Entity work is the cheap layer underneath all of it. That is why it is first.

The risks, stated plainly

You can do every step here and still not get a knowledge panel. Google decides what enters the Knowledge Graph and there is no application form that guarantees entry. You can also get resolved perfectly and still not get quoted, because the model chose a bigger name. Correlation, not causation, remember.

What you are actually buying with this work is the removal of an excuse. Right now, if the machine ignores you, one plausible reason is that it does not know what you are. Do this and that reason is gone. Everything after that is coverage, quality and time.

Where this goes next

Two things are pulling in opposite directions and you should watch both.

Google keeps insisting that optimizing for generative AI search is just SEO, and that special files and special markup do not get you into AI answers. Meanwhile every independent dataset keeps finding that off-site, name-level signals are what separate the brands that get named from the 26% who get nothing.

Both are true at once. Google is telling you there is no cheat code. The data is telling you that being a well-known, well-described, consistently-named business is the closest thing to one. Which is not a hack. It is just brand building with a machine as the audience.

The window is open because most of your competitors have not spelled their own name the same way twice. That is a low bar. Go and clear it.

Your move

Do the audit tonight. One name, five facts, one page, one schema block, four profiles saying the same thing. It costs nothing but an evening and it removes the dumbest reason for being invisible.

Then start on the mentions, because that is the part that compounds. If you want the templates, the exact JSON-LD, the mention-hunting tracker and the order to do it in, that is what the membership is for.

Join The Dojo. $67/mo. Or if you would rather we just built the whole entity stack for you, that is what Done For You is.