Billuminate

BillBot

Our crawler

BillBot fetches public knowledge from corporate websites and authority sources so Billuminate can answer with verified, sourced content. It follows the protocol, identifies itself openly, and respects opt-out without negotiation.

01 · Identity

How to recognize us

BillBot identifies itself in the User-Agent header on every request:

BillBot/1.0 (+https://www.billuminate.io/billbot)

If you see that string in your server logs, it is us. The URL points back to this page so any operator can quickly check who we are and what we do.

02 · Behavior

How BillBot behaves

  • Follows robots.txt per RFC 9309. We do not crawl anything a site has disallowed.
  • Maximum one request per second per domain. You should never notice us in your load.
  • Uses ETag and If-Modified-Since for conditional GETs. Unchanged pages cost you no bandwidth.
  • Bounded crawl depth per source, configured per customer. We do not spread across your site without reason.
  • HTTPS only. No cookies, no forms, no logged-in views.

03 · Opt-out

Don't want us crawling?

Two options. Either works. You don't have to contact us to opt out, but we welcome an email so we can confirm.

Option A: robots.txt

Add the following to your robots.txt:

User-agent: BillBot Disallow: /

On the next crawl, we see the directive and stop. No exceptions, no negotiation.

Option B: Email us

Send a request to privacy@billuminate.io with the domain you want excluded. We confirm within one business day and remove any existing material from our knowledge base within seven days.

04 · Why

Why we crawl

Billuminate is building a verified knowledge layer for consumer-facing financial questions. End-users should be able to ask things like "what is a payment remark?" or "what rules apply when I move my savings?" and get answers traceable to the underlying source.

That requires us to keep current content from authority sources (Swedish Tax Agency, Consumer Agency, Central Bank, Financial Supervisory Authority, ...) and from companies operating consumer products (banks, lenders, insurers). BillBot keeps that data fresh.

We do not train models on what we crawl. Content is used as a grounding layer for retrieval and cited answers, not for weight learning.

05 · Data

What we do with the content

Content is stored within the EU. PII is stripped at ingestion where possible. We do not republish text verbatim — we synthesize answers with source citations. Takedown requests are honored, within 48 hours for logos and within seven days for text.

Full policy lives on the transparency page.

To the transparency page →Contact us →