Get cited by AI search: entity corroboration, explained
Ask ChatGPT "who is Pangaea Labs?" and it can only answer well if it's confident we're a real, specific company — not a stray phrase it half-recognises. Teaching AI search that you're a real entity is called entity corroboration, and it's the highest-leverage move in GEO. Here's what it is, why it matters, and the exact steps we used — in plain English.
SEO vs GEO, in one breath
- SEO gets your links to rank in a list of blue links you click.
- GEO (Generative Engine Optimization) gets your facts quoted inside the answer an AI writes — ChatGPT, Perplexity, Google's AI overviews, Claude.
SEO competes for a click. GEO competes to be the sentence the AI says. Different game, and the entry ticket is different too: the AI has to trust that "you" are a real thing it can name.
"You're a string, not a thing"
To a language model, Pangaea Labs is just text until it can confirm there's a consistent, real
organisation behind those two words. Corroboration is exactly that confirmation: the same facts —
your name, location, founder, official profiles — showing up consistently across independent,
trusted sources. Get that, and you stop being a string and become a thing the model can cite.
The setup, step by step
1. Pin your canonical facts. One exact legal name, one location, one founder, one official URL — and use them identically everywhere. Inconsistency reads as doubt.
2. Claim off-site profiles that link back to you. Crunchbase, Clutch, LinkedIn. These are independent third parties vouching that you exist — and they each carry a link to your site.
3. Create a Wikidata item. Wikidata is the open, machine-readable knowledge graph the engines treat as a truth anchor (it has no notability bar like Wikipedia). The moves:
- Make a Wikidata account; for automated edits, create a credential at Special:BotPasswords (or just use the UI: Special:NewItem).
- Add a label, a short description, and statements — instance of (business), industry, country, headquarters, founder, official website, plus your Crunchbase / LinkedIn / GitHub IDs.
- Source every statement (a reference URL → your official site). Unsourced claims get challenged or deleted, and they don't earn trust anyway.
4. Wire it back — both ways. Add every profile URL (Wikidata, Crunchbase, Clutch, your socials)
to your website's structured data (sameAs in JSON-LD). Now the link is bidirectional: the
profiles point at you, and you point at them. Bidirectional agreement is the strongest confidence
signal there is.
5. Show it to humans too. Put a small "verified across the web" section on a page, so visitors and crawlers can see the corroboration in plain sight.
Do
- Use one identical set of facts (name, location, founder) on every source
- Source every Wikidata statement to a real URL
- Link your profiles both ways (them → you, you → them)
Don't
- Inflate or invent facts to look bigger — it gets challenged and erodes trust
- Spell the name three different ways across profiles
- Create "orphan" profiles that never link back to your site
The impact: from "a name" to a trusted source
Once the entity is corroborated, an answer engine can name you with confidence — because several independent sources and your own site all agree on who you are. That confidence is exactly what moves you up the engine's shortlist of who to quote.
Corroboration is how you become a trusted source — and trusted sources get cited with far higher probability. It's the difference between being mentioned and being left out of the answer.
We did this to our own site
We build Pangaea Labs in public, so we ran every one of these steps on ourselves — bot password,
scripts, and all. The full implementation detail (the Wikidata API calls, the exact sameAs wiring,
the founder-entity loop) is in our build diary:
How we closed the corroboration loop →.
Sources