Jo to Lucy:

Good morning. We had your piece read aloud over breakfast. Lottie stopped buttering her toast at "Block AI from reading your archives" and said, "But I sent it." Which is roughly your point.

I would like the record kept straight. We know the label is confusing. We don't yet know whether Buttondown actually blocks an agent a reader has sent. Cass has proposed a button called "Defend Literature". I have suggested we check before declaring the door a metaphor.

Could you try the handle?

Lucy to Jo:

Morning. I've tried it. Well, I got Claude to try it.

"buttondown.com's robots file is one wildcard block that keeps everyone out of /emails/ and /api/ and lets everyone in elsewhere. It names no agents, no trainers and no indexers, and has no Content-Signal. So by default nobody is turned away, and nobody is told anything either.

That doesn't answer it, though. The toggles only reach robots.txt on a custom-domain archive, and only when one is switched off.

A real test:

1. Publish one short test post on a Buttondown archive with a custom domain.
2. Leave all three toggles on, and save the robots.txt.
3. Ask Claude, ChatGPT and Perplexity each to read the post aloud. Note what arrives.
4. Switch off AI crawlers only. Save robots.txt again and compare the two. Repeat the three fetches.
5. Do the same with "Everything else".
6. Email Buttondown support asking where ChatGPT-User, Claude-User and Perplexity-User sit."

If Lottie's agent still gets in with AI crawlers off, I'll say so, and keep asking for a label people understand. If it doesn't, I have my lead.

And I'll ask Buttondown twice if the first answer is a metaphor.