Last updated 1 October 2026

For publishers and authors

If you publish or write something we may have used, this page says what Walking Points does with sources, which opt-outs we read, and how to reach a person. You can write to us at hello@walkingpoints.fm at any time.

What Walking Points is

Walking Points makes short audio episodes to listen to on a walk. Each episode is researched and written by AI and voiced by two AI voices we call Mia and Adam. That is said in every episode and in its show notes. It is a personal project run from Denmark by Lars Rothschild Henriksen.

How an episode uses sources

• Facts and our own commentary. An episode retells what happened in its own words and adds its own view. It is not meant to stand in for reading the original.
• Copied wording is checked for. Before the audio is made, where we can read a cited page, the script is compared with it. A run of 20 or more words in a row that matches the page is sent back for a rewrite, and if the rewrites keep failing, the last draft is used and the match is reported to us. A short quote that is attributed to its source (under 40 words) is allowed. The check is best-effort: If it cannot read a page, it does not run for that page.
• Not on one article. A news item is asked to rest on a primary source, or on reporting from more than one outlet, and this is checked before the audio is made.
• Nothing behind a paywall. A source on a site from our list of paywalled sites is rejected before the audio is made. The list is one we keep by hand, so a paywalled site missing from it can slip through. Tell us if you find one.
• Every source is linked. The show notes of every episode list the sources it was built from, with links.
• Credit on air. When a news item rests on an outlet's reporting, the hosts say the outlet's name once, in the sentence that carries its first fact. We do not name an outlet for a claim our fact-check questioned or could not read on its page.

The opt-outs we read

Before an episode is made, every source it cites is checked against the opt-outs below. This is what the check reads, exactly.

• robots.txt. A rule that disallows a page for any of these agents: ClaudeBot, Claude-User, anthropic-ai, GPTBot, Google-Extended, CCBot or WalkingPointsBot. A rule for * alone is not read as a reservation.
• TDMRep. The file /.well-known/tdmrep.json on your domain, in the standard form (a list of locations, each with tdm-reservation set to 1).
• The tdm-reservation header. A response header tdm-reservation: 1 on the page, or a <meta name="tdm-reservation" content="1"> tag in it.

What happens when a source has reserved its content: In a public series, that source is not cited. In an episode made for a single listener, it is kept out unless the same news item also rests on a source that has not reserved. If a site refuses our check, we cannot see its answer, and the source is not rejected on that account. If you want your site left out of every episode, whatever it says in these files, ask us below.

Our checks of these files identify themselves as:

WalkingPointsBot/1.0 (+https://walkingpoints.fm/publishers)

The pages an episode cites are read again for the fact-check and the copying check, as:

Mozilla/5.0 (compatible; WalkingPointsBot/1.0; +https://walkingpoints.fm/publishers)

Both carry the name WalkingPointsBot, which is what a robots.txt rule should name. The research that finds sources in the first place uses the search and reading tools of the AI service we write with. Those send that service's own agent name, not ours.

Ask us to remove an episode, or to leave your site out

Use the report form and choose "This uses my work (I represent the publisher or author)". Tell us which of your work we use and, if you can, give a link, so we know which domain you mean. You can also write to hello@walkingpoints.fm.

• We answer within five working days. A person reads every message.
• Nothing happens automatically. Anyone can say they represent a publisher, so a person decides, after reading what you wrote.
• Remove an episode. The episode is taken out of our feed and its audio is removed from our public storage. A podcast app that has already downloaded it keeps its copy, and we cannot recall that.
• Leave your site out. We add your domain to a list our source check reads, and an episode that cites it is rejected before the audio is made. It applies to every listener and to every future episode. It does not change episodes already published, so ask for those to be removed too if you want that.

A different kind of message, such as a factual mistake in an episode, goes through the same form, with its own answer time (48 hours).