About
We deleted our own product
Xither spent two years building a directory of enterprise AI tools: thousands of vendor profiles, buyer's guides, comparisons and ranked lists. In August 2026 we withdrew all of it. The pages now return a “gone” response rather than a redirect, because the content did not move somewhere else. We took it down.
Why we withdrew it
Two reasons, and the second one is the one that mattered.
The first is that nobody needs it. Describing what AI tools exist was genuinely useful in 2023. Today you can ask any assistant and get a better answer in seconds, and our own traffic said so plainly.
The second is that a large amount of it was not true. We had generated content at a scale no one could check, and when we finally checked, we found statistics attributed to reports that do not exist, quotes from practitioners who were never interviewed, and security certifications asserted for vendors on no evidence at all. Some of it had been live for months. We are not going to describe that as a content quality issue. We published things that were not true, about real companies, and the only responsible response was to take it down rather than try to repair it piece by piece.
What we do instead
We read a named set of public pages that AI vendors publish about how they handle data — sub-processor lists, data processing addenda, retention terms, model cards — and we keep a dated copy of each one. When a page changes, we can show you exactly what changed, on what date, with the vendor's own words and a link to the page they published them on.
This is a deliberately smaller promise than the one we used to make. We are not telling you which vendor is best, or whether a vendor is trustworthy. We are telling you what they said and when it changed.
What we hold ourselves to
Every claim shows its source
You can see the page we read, the date we read it, and the exact words. If we cannot show you those, we do not publish the claim.
We say when we do not know
If a vendor publishes nothing on a subject, we record that, with the date and how many pages we looked at. A blank is never left to be read as a yes.
We publish our mistakes
Our methodology page carries a dated list of what we have got wrong and what we changed as a result. It will keep growing.
Our scanner identifies itself
Our scanner identifies itself and honors robots.txt. If a vendor blocks it, we record that we were blocked rather than disguising ourselves to get in.
The full detail is on our methodology page, including the seven rules the system enforces and what each one is there to prevent.