We list companies that never asked to be listed. That is deliberate — and it produced the only number in our directory I find genuinely interesting: 622 of 819 listings have never been claimed by their owner.
This post is about the gap between being listed and being verifiable. Measured, not asserted.
The numbers
Read from our own public API today — snapshot October 6, 2026.
Never claimed — 622 listings (75.9%):
Average GEO score: 52.4
Evidence items per listing: 0.01
Domain verified: 0.0%
Claimed — 197 listings (24.1%):
Average GEO score: 70.3
Evidence items per listing: 0.52
Domain verified: 2.5%
Across the whole directory:
731 of 819 listings (89.3%) have no evidence item attached at all.
814 of 819 (99.4%) have no verified domain.
129 listings (15.8%) carry an editorial pick.
GEO score: median 52, mean 56.7, range 41–99.
That 17.9-point gap is the entire argument of this post. It is not a judgement about any company in that list. It is arithmetic about evidence nobody has supplied.
What we do, and what we can't
We index companies and products from public sources. That is how a directory reaches 819 listings without a sales team: we read the web the way an assistant does, and we structure what we find.
What we cannot do is claim a profile on someone's behalf. A listing reaches its full form only when its owner claims it and brings first-party, checkable facts. That rule does not bend — not for a two-person startup, not for the largest company in the directory. A directory whose operator writes the facts is not verifiable; it is one more page of claims.
A search for Y Combinator in our directory returns no listing. The invitation stands, and it is the same invitation everyone gets: listing costs nothing, the claim has to be yours.
Why we can't close it for them
CitableHub publishes two trust signals that are never merged: domain ownership and editorial status. A listing can have either, both, or neither. Both require the owner:
Domain ownership is proven by an artifact on the domain itself — a meta tag, a file under /.well-known/, or a DNS TXT record. We can't place that for you. If we could, it would prove nothing.
Evidence is first-party material: the product page, a public repository, a case study, a review. We can link it. We can't invent it.
Editorial status is our own review, kept separate on purpose. It is an opinion, and opinions should not be dressed up as verification.
The pattern in our inbox
Since we launched, companies have written to ask us to list them — and more often, to ask us to fill in the profile for them. We say no, the same way every time.
I can't show you our inbox, and I'm not going to name anyone in it. What I can show you is the directory, which is public: 75.9% of it is unclaimed, and unclaimed listings average 0.01 evidence items.
That is the honest version of that paragraph. The claim you can't verify, I don't get to use.
What this is not
A listing is moderation. It is not domain verification, and it is not an editorial pick.
None of it makes any AI crawl, index, cite or recommend you. We measure reading. Our public counter reads 12,338+ verified AI crawler reads as of October 2026, from first-party server logs — and reads are not citations. We track citations separately, and we publish that number too, including when it is zero.
We don't sell placement. A boost changes ordering, not the score.
If you are one of the 622
Claim the listing. Free, no card, you keep control of the profile.
Attach at least three evidence items — a product page, a repository, a case study, a published review.
Verify your domain: meta tag, /.well-known/ file, or DNS TXT record.
Complete the profile: privacy, terms and contact URLs, category, audience, use cases.
Answer the questions people actually ask an assistant. Those become FAQ data on your profile.
Publish something new occasionally. The score carries a freshness dimension; a profile that never changes gives a crawler nothing new to read.
Claimed listings average 70.3 against 52.4. That is what the work is worth — and it is the only lever we can't pull for you.
If you run a directory, publish your own distribution too. And if you find a better way to measure this gap, tell me where my numbers are wrong.
Top comments (0)