For 25 years, SEO meant winning at the link graph. Earn the right citations, point them at the right pages, and Google would do the math. The page with the strongest backlink profile usually won.
That model is breaking.
ChatGPT, Perplexity, Gemini, and Google AI Overviews don’t rank pages — they answer questions by stitching together facts from multiple sources. To do that, they need to know what you are, not just where you sit on a SERP. The new substrate underneath the answers is the entity graph: people, companies, products, places, and concepts, cross-referenced for overlap, consistency, and reliability. Links still help. They are no longer the main currency.
The mechanic
When Perplexity gets asked “which CRMs are best for solo consultants in 2026?”, it isn’t ranking ten URLs. It’s asking: which entities show up consistently across reputable sources for this query? What attributes of those entities are agreed upon? Whose name keeps appearing next to “solo consultant”? The system retrieves passages that talk about the same entity from different angles, weights them by source authority and internal consistency, then writes a synthesis.
Two things follow. First, the model has to recognize you as an entity at all — that means a stable, machine-readable identity (your business, your founder, your products) that reads the same in twelve different places. Second, your association with the topic has to be reinforced from outside your own domain. Self-published claims aren’t a signal. Pattern-matched mentions across the open web are.
The practical mechanics: name-address-phone consistency (still — and now more than ever), Wikidata and Wikipedia presence, schema entity references with `sameAs` pointing to authoritative profiles, structured Person and Organization markup, and consistent founder/team bios across LinkedIn, Crunchbase, podcast appearances, guest posts, and your own About page. Each of these is a vote that says “this entity is real, this is what it does, this is who runs it.” When Gemini cross-checks, it finds agreement. That’s what gets cited.
The numbers already reflect the shift. Sites with 32K+ referring domains are roughly 3.5× more likely to be cited by ChatGPT than thin-profile sites. That stat isn’t really about links per se — it’s about the entity having enough surface area for the retrieval system to lock onto it. The link graph and the entity graph are correlated; they aren’t the same thing.
What to do this week
Pick the top three entities you want to own — usually the company, the founder, and the flagship product or service category. For each one:
1. Audit the bio. Pull your current About copy, your LinkedIn summary, your Crunchbase blurb, and three guest-post bios. They should agree on what you do, who you serve, and the descriptive phrase you most want to be associated with. Most founders have four different versions written years apart. Pick the one you want, propagate it everywhere.
2. Ship Person and Organization schema. On your homepage and About page, mark up Organization with `name`, `url`, `logo`, and `sameAs` (LinkedIn, Crunchbase, X, Wikidata if you have it). On founder/author pages, mark up Person with `sameAs` pointing to the same external profiles. Schema isn’t the citation lever. The point is that you’re handing AI engines a clean entity card so they don’t have to guess.
3. Get on Wikidata. Even a barebones Wikidata item with your company, founder, founding date, and a couple of `sameAs` links costs an hour and feeds into nearly every retrieval system. If you can earn a Wikipedia page later, great. Wikidata first.
4. Audit five external mentions. Search your brand name. Find the top five non-owned references — directories, partner pages, podcast notes, press hits. Are they describing you correctly? Wrong category, wrong founder name, wrong tagline? Fix the ones you can edit. Ask politely on the ones you can’t.
This work is unsexy and almost entirely off your own site. That’s the point. The link graph rewarded what you could pull onto your domain. The entity graph rewards what the rest of the web says about you when you’re not in the room.
If you’re a brand that wants to be the answer LLMs reach for (not just rank on Google), Paris Roussos has been engineering search visibility for 30 years and now runs done-for-you AI SEO. Flat-rate, no-fuss. Email parisroussos@gmail.com.
The brands that win the next two years won’t have the most pages. They’ll have the cleanest, most consistent entity record across the open web — and the AI engines will quietly start naming them before the competition figures out what changed.