Hi everyone, this is Neo.

A reader messaged me yesterday: “I keep hearing that without GEO (generative engine optimization), my independent site will be left behind in the AI era. Should I spin up a GEO project right away?”

My answer: hold your horses. Before you spend money on the newest buzzword, you’ve probably got old homework you never finished.

This week I read a heavyweight piece on Search Engine Journal by Harry Clarkson-Bennett, SEO Director at The Telegraph, titled Entities Are The Foundation Of Modern SEO. There’s one line in it I agree with completely: GEO is just entity SEO with a new name — not because he’s chasing clout, but because the underlying logic of AI-driven information retrieval has never actually changed.

In this post, I’ll break down his core argument plus a framework you can copy directly — Own / Describe / Prove / Earn — and add my own take with a practical priority list for independent site sellers.

First, a Reality Check: GEO Isn’t a New Species. Entity SEO Is the Foundation That Never Moved

Over the past two years, “GEO,” “AEO,” and “AI search optimization” have been thrown around constantly, and tool vendors seem determined to invent a new protocol or concept every month. But Clarkson-Bennett’s point is to look backward:

Google’s Knowledge Graph launched in 2012 and has been around for well over a decade now — it is, quite literally, a database of entities: people, places, organizations, products, and the relationships between them. Remember Google’s famous “Things, not strings”? That was the moment search stopped “matching keywords” and started “understanding entities.” And every major algorithm shift since — Hummingbird, BERT, MUM — has been about the same thing: understanding entities and the semantics around them better.

Here’s the line that really cuts through all the GEO hype:

Search engines and AI systems have to build a picture of who you are from fragmented evidence across the web. Your job is to make that picture as clear, consistent and well-supported as possible.

That sentence strips the mystery right out of “GEO.” Whether it’s Google or ChatGPT, neither reads your homepage and calls it a day — they collect fragments about you from all over the web and stitch them into a picture. How clear that picture turns out determines whether you exist at all in traditional search and AI answers.

That’s entity SEO: reducing machine uncertainty about who your brand is. It’s not new — but it’s exactly the foundation that matters most in the AI era, and most brands have never seriously built it.

Why Entities Matter More Than Ever in the AI Era (With Data)

You might ask: if entity SEO is such an old concept, why does it suddenly matter now?

Because AI has amplified the importance of entity perception tenfold. Two hard data points:

First: Google is actively “cleaning house.” According to Kalicube’s Knowledge Graph Sensor, in June 2025 the Knowledge Graph went through its largest contraction in a decade: in a single week, entities dropped from roughly 53.1 billion to about 50.2 billion — more than 3 billion entities purged. Search Engine Land read the shakeup as Google raising the bar for what qualifies: vague, duplicate, and low-quality entities got swept out in batches, and what remains must be clear, unambiguous, and well-evidenced.

Translation: Google’s standards for “who you are” just got stricter. In the old days you could get by on volume and keyword-stuffing. Now machines scrutinize brand identity more tightly than ever.

Second: brand mentions are replacing backlinks as the AI-era “authority.” An Ahrefs study of 75,000 brands (widely cited by Onely and others) found that brand mentions correlate with AI visibility at 0.664, versus just 0.218 for backlinks — mentions carry roughly three times the weight of links for AI visibility. And brands in the top quartile of mentions earn more than 10x the AI citations of the next quartile.

Why? Because AI doesn’t evaluate you by counting links — it reads what the web says about you and in what context. Links are graph theory for crawlers; mentions are the language signal LLMs actually understand.

So the conclusion is clear: entity SEO = making everything the web says about you point, consistently and accurately, to “who you are and what you’re good at.” Until that homework is done, publishing llms.txt files and adopting MCP protocols is just pouring water into a leaking bucket.

Own: Clean Up Everything You Control (Entity Hygiene)

The first step of Clarkson-Bennett’s framework is Own — take inventory of every asset you can edit, everything fully under your control:

• Your main site: About Us, Contact, privacy policy, footer — and don’t forget structured data and sameAs markup
• Your corporate site, if you have one
• Profile pages for key people
• Social channels
• Google Business Profile
• WikiData and Wikipedia entries (if you qualify)
• Your storefronts and profiles on marketplaces and industry platforms

Then run an “entity hygiene” pass:

• Is your brand name, logo, and description identical, character for character, across every channel?
• Is your NAP (Name / Address / Phone) fully consistent? — An old-school local SEO concept that still works in the AI era
• Do all your links point to the right endpoints?
• Are descriptions of founders, authors, and key employees unified?

His point: skip the grand plans — start with Own. It won’t cause a traffic spike overnight, but it’s the starting line for everything else, because your “entity” in the machine’s eyes is defined, bit by bit, by the properties you control. If those are inconsistent, everything downstream is wasted effort.

He calls the discipline entity hygiene and the consequence of skipping it entity drift — your identity slowly warping out of shape across the web.

Describe: Make Who You Are — and Your Web of Relationships — Explicit

The second step, Describe, is page-level work: making every page tell the machine, unambiguously, which entity it’s about and what that entity relates to.

Self-audit checklist:

• Does this page explicitly target the right entity? (Product pages highlight the product entity; author pages highlight the person.)
• Does the surrounding information — headlines, navigation, breadcrumbs — corroborate that targeting?
• Does the heading structure answer the questions users actually care about?
• Are internal and external link anchors consistent and relevant?
• Does the page answer topically relevant questions?

Three operational notes from the article:

First, structured data is the language of relationships. Organization schema defines who your brand is; Article and Person schema corroborate individual pages; sameAs ties your site, socials, Wikipedia, and industry listings into one evidence chain that screams “these are all the same entity.”

Second, internal linking is the most scalable win for medium-to-large sites. Articles about mortgages link to the mortgage hub page; the hub houses the hundreds of posts you’ve written so the machine can trace your topical map. Author pages work the same way.

Third, make your experience explicit. How you ran a study, what your testing protocols were, where your data came from — these aren’t just trust signals for readers. They’re evidence for machines that this brand knows what it’s talking about.

Prove: Demonstrate That You and Your People Are Real and Credible

Step three, Prove, tackles the hardest part of E-E-A-T head-on: how do you prove you’re not an AI-generated shell brand?

Clarkson-Bennett’s answer is refreshingly specific:

• Every person who matters — authors, founders, executives, named experts — needs one consistent, verifiable identity: an author page linking to a real LinkedIn profile, a real publication history, valid structured data, contact links, and a bio that says who they are and why they’re trustworthy.
• Show the credentials: won awards? Highlight them. Have a Master’s in economics and a decade at Bloomberg? That’s useful for both your audience and machine retrieval.
• Market your partnerships actively: if you’ve collaborated with or been endorsed by an authority in your space, don’t keep it quiet — third-party endorsement is proof.
• Publish your research process and methodology: how your data was gathered, how conclusions were tested. These transparency moves are pure E-E-A-T gold.

In one line: prove you exist, prove you’re real, and prove your claims have receipts.

Earn: You Can’t Fully Define Yourself — You Must Earn Corroboration

Step four, Earn, is the most counterintuitive and arguably the most important part of the framework: your definition of your own brand doesn’t count. What the rest of the web says about you does.

The article draws a crucial distinction:

• Entity corroboration answers “who are you?” — accurate brand mentions, consistent third-party descriptions, correct directory and association listings.
• Topical corroboration answers “what are you known for?” — expert commentary, interviews, bylines, and citations that keep connecting you to the subjects you claim to specialize in.

He puts it bluntly: you can call yourself the world’s foremost Bitcoin expert all day, but if the web sees you as a chef from a small town, you’re swimming against the entire internet. For independent sellers this is a gut punch — is your category positioning and expert persona something you claim, or something others actually corroborate?

And note this: mentions are easier to get than links and harder to fake. For a decade, SEOs treated unlinked mentions as a chore to convert into backlinks — in his view, largely a waste of time. The logic has now flipped: a pure mention with no link is an asset by itself, because it feeds the language signals AI uses to evaluate your brand.

So stop agonizing over whether a piece of press “didn’t earn us a backlink.” Any opportunity that gets an authoritative third party mentioning you in the right context is a premium link in the AI era.

Monitor + Entity Map: Make Entity SEO a Continuous Discipline

Own / Describe / Prove / Earn isn’t a one-off project — it’s a discipline you maintain. The article offers one very practical anchor: build an Entity Map.

You can start with nothing more than Google Search Console query and page data:

1. Identify the primary entity. Who or what are you mapping — the brand, the founder, a product line?
2. Identify connected entities. The people, products, services, locations, topics, organizations, and events associated with it.
3. Map the relationships. Founder of, employee of, headquartered in, specializes in, writes about, produces, owns…
4. Attach identifiers and evidence. Official URLs, LinkedIn profiles, Wikidata IDs, author pages, authoritative sources.

You’ll probably be startled by the result: the brand you think you are and the brand your evidence describes are rarely the same person. That’s precisely the point of the exercise — it forces you to confront the gap between “how you’re perceived” and “who you think you are.”

Power moves:

• Use an LLM to run NER (named entity recognition) over tens of thousands of GSC queries, extracting and classifying them into entities — people, places, organizations, products — in minutes instead of weeks.
• Filter your GSC data by country and draw one map per market. How the same brand is perceived in different countries is rarely what you’d expect.
• Track mentions with a tool like Alertmouse and watch for entity drift: when employees leave, does the corporate site update? Are third parties still describing you with outdated language? Is the AI’s picture of you still the brand you want to be?
• The simplest test of all: every few weeks, ask ChatGPT, Gemini, and Perplexity “what do you know about [your brand]?” LLMs are the best mirror of how the whole web sees you. Read their answers and you’ll know exactly how far your entity work still has to go.

Neo’s Take: What Independent Site Sellers Should Actually Do

What I love about this article is that it drags “GEO anxiety” back down to earth. Too many people today make a living selling new concepts: GEO tools, AI visibility audits, llms.txt deployment services. But Clarkson-Bennett has it right — protocols and tools are downstream; knowledge and consistency are upstream. If you haven’t figured out who you are in the machine’s eyes, chasing the newest format just means patching a different hole in the same leaking bucket.

For independent sellers, here’s your execution priority:

First, do Own tomorrow (half a day). Sweep every platform you control: your site, socials, Google Business Profile, marketplace storefronts. Are the brand name, bio, links, and contact details consistent? It’s free, it’s fast, and it’s the foundation of everything else.

Second, draw an entity map this week (one day). Export your GSC queries and run the four-step method above. You’ll discover a lot of “keywords I thought I owned” that the machines never connected to you at all. That map is now your guide for content direction and external partnerships.

Third, make “how you’re described” a KPI. We used to watch rankings and traffic. Add a new question to the mix: what words do AI and third parties use to describe you? Ask an LLM “who am I” every month, and ask yourself before every piece of content: does this sentence make the machine’s picture of me sharper?

Fourth, don’t buy GEO tools yet. For the vast majority of independent sites, entity consistency plus high-quality third-party mentions is worth more than any AI optimization tool. Tools are amplifiers — they only work once you have a clear, consistent, well-supported brand entity to amplify. Otherwise you’re just amplifying noise.

Here’s the closing thought: Google and AI never assume who you are. Who you think you are doesn’t matter. Who you are on the evidence chain decides whether you get any place in search and recommendations in this era.

Entity SEO isn’t sexy. But it works. Don’t let new-concept anxiety push you forward — go lay the foundation that hasn’t moved in a decade and a half, properly, once and for all.

I’m Neo, and I write about independent site SEO. If you’ve built an entity map of your own, come back and tell me what you found — I’d love to hear it.