Retiring Trackbacks After Spam Escalates, Housekeeping WordPress with AI Pt. 2

In Part 1 of the series, I said I intended to keep trackbacks open. I had to close them. Bot traffic previously only sniffed around pingbacks, but during the last month since XML-RPC was retired, the threats escalated. Currently, WordPress accepts no trackbacks, and Cloudflare rejects the standard trackback routes before they get that far.

On November 15, I noticed new comments in the moderation queue on a site where comments are off. They were three trackbacks, sent on November 10 and 13. Part 1 had said none had ever arrived. I asked the AI how they were getting in after the XML-RPC closure. Trackbacks use their own endpoint, it said. Keeping trackbacks open was what let them in.

Two of the three were keyword titles about iPhones pointing at one domain’s homepage. The third praised the retirement of XML-RPC, which is what Part 1 was about, but its link was unrelated. None of the links had anything to do with the topic at hand, so I’m calling all three spam.

That third one is the smoking gun. It was written as if answering my post. A reply like that wouldn’t fool an agent, but it’s enough that WordPress doesn’t bin it as spam on its own. All three made it into the queue.

I told the AI to close trackbacks everywhere and to take the old records with them. I’d said earlier pingbacks would stay published. The three pingbacks went too, along with the three trackbacks. We can still cross-link each other without them. Not ideal, but I suspect that kind of conversation has moved to social platforms anyway.

I also wanted Cloudflare to refuse them, because the endpoint was likely to be abused and probed, and asked the AI to check that this was the conventional method before we built it. Its answer was to close trackbacks in WordPress and optionally block the endpoints before PHP runs. Thirty test requests ended at Cloudflare with a 403, and none reached the server. I expect the bots to keep adapting. The point is that I can make these decisions on the statistics and guidelines the AI reads, and double-check the guidelines myself, instead of bearing the burden of being SEO and traffic analyst at the same time.