How long does it take to get cited by AI?

Longer than a month. After thirty daily entries this blog had 1,358 machine reads in fifteen days, zero referral clicks from any AI assistant, and zero Bing impressions. Training crawlers arrive first and quoting comes later, which is why GPTBot outnumbered the live fetcher here by more than four to one. Being read is the floor, not the result.

Which crawler reads a new blog first?

On this one it was bingbot, by a distance, at 158 fetches against GPTBot's 106 over fifteen days. That matters because Bing's index feeds ChatGPT's live search, so the crawler that looks least relevant to AI visibility is often the one doing the most for it. Amazonbot came third, ahead of Googlebot.

Is zero clicks from AI a sign something is broken?

Not by itself. A machine fetch and a referral click are different events, and most AI answers never produce a click at all. Check the server log rather than analytics before concluding anything, because analytics runs in a browser and crawlers do not run it. Zero clicks alongside heavy crawling means you are being read and not yet being quoted.

CITED, One Month In: What Getting Quoted Actually Took

¶ FIELD NOTETHE SCOREBOARD

01Thirty entries, and the numbers that look bad

I set out a month ago to write a blog that machines would quote rather than one people would visit, and I said at the start that I would publish the results either way. Here they are, and the honest summary is that half of it worked and the half everybody cares about did not.

I did not chase readers. I chased being the source. One month in, the readers have not arrived and the machines have not left.

What the search engines show

Google, over three months: twenty nine impressions, zero clicks, average position 17.5, and exactly one query with an impression against it. Bing, over the same window: zero impressions and zero clicks, a flat line for four straight weeks. Those are not rounded and they are not softened. If you came here for a growth story you can stop reading now, because on the surface that is a month of writing into a wall.

The single query is worth naming. Somebody searched get cited, this blog appeared once, at a position nobody scrolls to, and they did not click. That is the entire organic search story of thirty entries. I am including it because leaving it out would make the rest of this page dishonest, and because a blog about being quoted accurately does not get to round its own results up.

What the server shows instead

The log tells a completely different story, and it is the reason I am not stopping. In fifteen days, which is all my host retains, these pages were requested 1,358 times by machines. Not people. The search dashboards recorded almost nothing. The server log recorded 1,358 visits nobody was counting. The chart below is who came, and the shape of it surprised me more than the zeros did.

¶ FIELD NOTETHE CRAWL

02Who actually read it

Eight agents did the work. I expected this list to be led by an OpenAI crawler, because the whole thesis of this blog is about being quoted in chat answers. It is not.

Three things in this chart I did not predict

Bing crawled hardest, by a distance. That is the thesis of an earlier entry here confirmed by accident, on this blog, against my own expectations. The engine I wrote about as the hidden layer turned out to be the one reading me most. Second, GPTBot outnumbers ChatGPT-User by more than four to one, and those two crawlers do different jobs. Training arrives first. Quoting comes later, if it comes. Third, the crawl is not steady, it is lumpy: eighty one fetches on one day, two on another. It answers the doorbell rather than keeping a schedule.

// THE MONTH IN CRAWLS~~15 DAYS · CITED PAGES ONLY~~

bingbot158

GPTBot106

Amazonbot104

Googlebot45

OAI-SearchBot39

PerplexityBot32

ChatGPT-User24

Applebot14

1,358 requests in fifteen days. Referral clicks back from ChatGPT, Perplexity, Claude, Copilot and Gemini over the same period: zero.

EXTRACTED — THE SENTENCE THIS ENTRY EXISTS FOR

You get learned before you get quoted. Nobody tells you how long the gap is.

¶ FIELD NOTETHE VERDICT

03What I would keep and what I got wrong

Three things earned their place and two did not, and I would rather write that down now while the numbers are inconvenient than wait until they flatter me.

What I would do again

Publishing every day, because the oldest entries are the most fetched. Nothing on this blog has ever been read on the day it was published. Announcing every change instead of waiting to be found, since the crawl spikes line up with the days I rang the doorbell. And writing numbers into sentences rather than charts, which is the rule from the floor entry and the only reason this recap was possible to write at all.

What I got wrong

I expected referral clicks. Not many, but not zero, and zero is what four weeks produced. I also expected the newest entries to draw the most attention, and the opposite happened: the most fetched pages are among the oldest, which means this compounds rather than spikes. A month is long enough to be read and nowhere near long enough to be quoted. The one thing that genuinely encouraged me is small: an OpenAI agent fetched the llms.txt file four times, and on other sites I run no AI company has ever touched it.

So the scoreboard reads zero clicks, zero Bing impressions, and 1,358 machine reads. If you only count the first two, this failed. I have already argued that counting only those two is the mistake. Tomorrow the calendar carries on, because the thing I am building takes longer than thirty days and I would rather say that plainly than dress up a flat line.