Podcast episode 2: A new page, zero links: ClaudeBot came in 3 hours. Bingbot never came

Episode 2 of Measured, Not Guessed: we published one article on a new domain with zero links and an IndexNow ping on 29 October 2026. ClaudeBot crawled it after 3.3 hours, Googlebot after 13.4, Bingbot not at all. Listen, with chapters and full transcript.

Last updated

  • 3.3 huntil ClaudeBot first crawled the new articleExperiment CT-001, published 29 Oct 2026, 11:27 UTC
  • 13.4 huntil Googlebot first crawled itSlowest of the nine pages we charted
  • 0Bingbot visits recorded, despite an IndexNow pingBy the last update of the experiment page

Short answerWe published one article on a new domain with zero links and an IndexNow ping on 29 October 2026: ClaudeBot made the first verified crawl 3.3 hours later, Googlebot 13.4 hours later, and Bingbot none by the page’s last update. In this 4-minute episode, Jake and Dylan go through experiment CT-001, including the prediction we got wrong.

Listen on Apple Podcasts, Overcast or RSS.com, or add the RSS feed to any podcast app.

Chapters

  1. 0:00 How fast does AI find a new page?
  2. 0:20 The setup: zero links and an IndexNow ping
  3. 1:33 Crawlers: ClaudeBot 3.3h, Googlebot 13.4h
  4. 2:29 Citations: Gemini first, ChatGPT on day 14
  5. 3:33 What to take from it

Read the full data: Experiment CT-001 on the Experiments page. Lab note 003: what five AI engines cited in the first 14 days. How long a new website takes to get indexed and ranked in Google.

Transcript

Jake: Okay, so here’s the question we get all the time. You publish a brand-new page on a brand-new site. How long until AI actually finds it?

Dylan: And the honest answer was, we didn’t know. So we tested it.

Jake: This is Measured, Not Guessed, from the AI Visibility Census. I’m Jake.

Dylan: And I’m Dylan.

Jake: Today it’s experiment CT-001. One article, a new domain, and basically a stopwatch. So Dylan, set it up for me. What did you publish?

Dylan: An article on keyword research for AI search. It went live on October twenty-ninth, at eleven twenty-seven in the morning, UTC. And we built zero links to it.

Jake: Zero? Not even a social post?

Dylan: Nothing pointing at it. We wanted the raw number.

Jake: Okay, but you didn’t just hit publish and pray.

Dylan: No, we did all the stuff people tell you to do. An IndexNow ping one second after publishing. It’s in the sitemap, it’s in llms.txt, it’s got Article and FAQ structured data.

Jake: Did you submit it to Google by hand?

Dylan: Nope. No manual submission to Google or Bing. And we put a made-up phrase on the page, a test marker that doesn’t exist anywhere else. So if an AI engine ever repeats it, we know exactly where it came from.

Jake: Smart. And you made a prediction before any of this, right?

Dylan: We did, and we wrote it down first. We said Bing would get there before Google, because of IndexNow. And we said Copilot or Perplexity would pick up the marker before ChatGPT or Google.

Jake: How’d that go?

Dylan: Badly.

Jake: Ha. Okay, give me the crawlers. Who showed up first?

Dylan: ClaudeBot. That’s Anthropic’s crawler. Three point three hours after publishing.

Jake: Wait, before Google?

Dylan: Way before Google. Googlebot showed up thirteen point four hours after publishing. That was the slowest first visit of the nine pages we’ve charted.

Jake: And Bing? You literally pinged Bing.

Dylan: Bingbot never showed up. None recorded, at least by the last time we updated the page.

Jake: That’s wild. The one engine you told about the page didn’t come.

Dylan: Yep.

Jake: One thing though. Was Google just slow, or was it waiting on something?

Dylan: Good question. No Googlebot request came before we connected Search Console, and that was at two fifty in the afternoon. So we can’t rule out that Search Console nudged it. Google showed the page as indexed at nine fifty-seven the next morning.

Jake: So call it a day to get indexed.

Dylan: About twenty-two and a half hours, yeah.

Jake: Okay, crawling is one thing. Did anything actually cite it?

Dylan: That’s the part people care about. Day one, nothing. No engine cited it or named us. Most of them said the marker wasn’t a real term.

Jake: Which, fair. It isn’t.

Dylan: Right. Then day three, Gemini cited the article. It was the only source in the answer.

Jake: Gemini first. Not Perplexity.

Dylan: Perplexity we couldn’t test. Every attempt hit a login or sign-up wall. So that’s a gap, not a zero.

Jake: Got it. Then what?

Dylan: Day seven, Gemini stopped citing the article, but it named the AI Visibility Census. Then day fourteen, ChatGPT named us too. But it cited our Experiments page, not the article.

Jake: Hmm. Why does that matter?

Dylan: Because the Experiments page quotes the marker. So ChatGPT found the write-up of the test, not necessarily the article. We’re not counting that as ChatGPT reading the article.

Jake: That’s an honest call.

Dylan: It’s the only call. If we fudge that, the whole thing’s worthless.

Jake: Okay. So somebody’s launching a new site tomorrow. What do they take from this?

Dylan: Three things. One, don’t assume Bing is fast just because you pinged it. We sent IndexNow and Bingbot never came. Two, AI crawlers can beat Google to a new page. ClaudeBot was there in about three hours. And three, getting crawled isn’t getting cited. Gemini cited it on day three and dropped it by day seven.

Jake: And the catch?

Dylan: Big one. This is one article on one site. It tells you what happened here, not what always happens. That’s why we keep running these.

Jake: When’s the next check?

Dylan: Day thirty, November twenty-eighth. And we’ll publish it either way.

Jake: The timings and the full method are linked in the show notes. This was Measured, Not Guessed.

Dylan: AI visibility, measured, not guessed.

Next: one experiment or finding per episode. All episodes, or get new findings by email.

Cite this

An article published with zero links on a new domain on 29 October 2026 was first crawled by ClaudeBot after 3.3 hours and by Googlebot after 13.4 hours; Bingbot had made no recorded visit despite an IndexNow ping (AI Visibility Census experiment CT-001).

Source: AI Visibility Census, “Podcast episode 2: A new page, zero links: ClaudeBot came in 3 hours. Bingbot never came”, https://aivisibilitycensus.com/podcast/new-page-crawl-timing/ (CC BY 4.0)
How we made this episode, and limits

Voices: the hosts, Jake and Dylan, are AI-generated voices made with ElevenLabs. They are show hosts, not real people. The research, the data and the script are by the AI Visibility Census team.

Checks and AI use: the script was drafted with AI assistance from the published study and every figure was checked against that page before recording. After recording, we ran the audio through speech recognition and confirmed every number is read as written. The transcript is our script, timed to the audio.

Limits: one article on one new site, so it shows what happened here, not what always happens. Perplexity could not be tested because every attempt hit a login or sign-up wall. The day 30 check is due 28 November 2026; the full record and caveats are on the Experiments page.

Spotted an error? Email hello@aivisibilitycensus.com and we will correct it.