The Northern Inference page reader

If your website's logs show the user agent NI-UserFetch/1.0, a person's question in a Northern Inference chat led the page reader to fetch that one page. The page reader is not a crawler: it follows no links on its own, builds no index and does not come back on a schedule.

What the page reader does

• It fetches one page at a time, only while a person is chatting: a page they linked to, or a search result the model reads for them.
• It reads only web pages and plain text (text/html and text/plain), up to 2 MiB per page.
• It sends no cookies and keeps none, and it sends no Referer header.
• It follows at most 5 redirects.
• It does not read robots.txt: each page is fetched for one person's chat, not to explore your site.

What Northern Inference keeps

No copy of your page. To answer abuse reports, we keep a coded fingerprint of the page address for 30 days, which we can match only against an address someone reports to us.

Where the requests come from

Amazon Web Services in Canada, in the ca-central-1 region.

Report abuse or ask us to stop

Write to abuse at northerninference dot ca with the page address and the time of the request. We can stop the page reader from fetching your site.