The machines came first.
Then the customers.

Thirteen AI crawlers from ten companies have read our small business website 5,840 times in three months. Two weeks after our first answer pages went live, a real sales lead arrived. Zero dollars spent. Never submitted anywhere. Zero outreach of any kind. Here are the raw numbers, from our own server logs.

Our own server logs $0 marketing spend No submissions, no ads, no outreach Three months of data
5,840
AI engine reads in three months — and counting
13
AI crawlers to date, from 10 companies
$0
spent on marketing or outreach
14
days from first pages to first sales lead

The cleanest experiment we never meant to run.

This site went live the quiet way. Pages published, structured so search engines and AI engines could read them clearly — and then we did absolutely nothing else.

We never submitted the site to Google. Never pinged a single AI company. No ads. No influencer deals. No PR. No outreach to the public, to directories, to anybody. Not a small budget — a zero budget. The door was unlocked and the lights were on, but nobody knocked on anyone's door to say we existed.

Which means the crawlers came because they found us on their own. The lead arrived because the pages did their job. That's what makes these numbers worth trusting: there's exactly one explanation here, and it's the farming method — plant pages that answer real questions, tend them, and let the machines discover you.

Nobody asked ChatGPT to stop by.

So far, thirteen AI crawlers from ten companies have knocked — 5,840 visits in three months, every one of them found on their own. OpenAI's GPTBot leads the list at 2,422. Anthropic's ClaudeBot is right behind at 1,685 — mostly the same 22 computers, returning again and again. Then the librarian, OAI-SearchBot at 650. Meta at 431, Bing at 326, Apple at 205.

And the cast keeps changing. In month two, one new arrival mattered most: ChatGPT-User, the bot that fetches your page live when a real person asks ChatGPT a question. That's the recommendation machinery, caught in the act. It only fetches when a real person is asking, so its count starts small — but every hit is a live question, not a scan. In month three, Gemini and Brave showed up for the first time. The list of engines that want to read a small business website is still growing.

Here's the whole cast in plain English. GPTBot is a student reading the library, so the machine knows things later. OAI-SearchBot is the librarian building the card catalog, so your pages can be found and quoted later. And ChatGPT-User is a customer at the desk, asking a question right now — and the librarian walking over to pull your book off the shelf. The first two are the machines getting ready. The third is the machine answering.

Not long ago, "AI visibility" was a theory people argued about. Now it's in our server logs: 5,840 visits across three months, from the engines that decide who gets recommended when somebody asks an AI for help.

AI crawlerWho runs itReads to date
GPTBotOpenAI — teaching the machine2,422
ClaudeBotAnthropic1,685
OAI-SearchBotOpenAI — the card catalog650
Meta AIMeta431
Bing AIMicrosoft326
ApplebotApple — Siri / Apple Intelligence205
GeminiGoogle — Gemini75
AmazonbotAmazon17
BytespiderByteDance (TikTok)15
Google-ExtendedGoogle — Gemini7
PerplexityBotPerplexity4
ChatGPT-UserOpenAI — answering someone right now2
BravebotBrave Search1
Total13 crawlers, 10 companies5,840

Data: our own server logs, August–October 2026. July, for reference: about 230 AI crawler visits.

Those visits were auditions.

When someone asks ChatGPT "who's a good plumber near me," it doesn't guess. It quotes businesses whose websites answer real questions clearly — pages it has read, trusted, and remembered.

Most small business sites never get read at all. There's nothing for a crawler to quote: no clear answers, no fresh pages, no proof. So when a customer asks an AI for a recommendation, the AI recommends somebody else — quietly, every day, and the owner never knows it happened.

Here's the part that surprised us, though probably shouldn't have: the crawlers didn't read everywhere equally. The most-read pages were the ones that each answer one real customer question — our FAQ library led all three months: 428 views, then 431, then 318. More than 1,100 reads in total. Case studies and service pages got a fraction of that.

That's the whole trick, and it's the opposite of what most agencies sell. Not more pages. Not more jargon. Better answers to the questions your customers actually ask.

Then we checked whether it replicates.

One site could be luck. So we ran the same method on a second site we own — an AI answer hub for business owners — and watched the logs the same way.

Same pattern. AI engines read that site more than 6,300 times in September and October alone — ten different engines each month, with new ones (YouBot, Common Crawl) still showing up. And just like here, the thing they read most wasn't the homepage or the sales pages. It was the glossary: plain-English answers to real questions, the most-read pages on the site two months running.

Two different sites. Two different topics. Same result: the pages that answer real questions are the pages the machines quote. That's not an anecdote anymore. That's the method.

Then the field produced early.

The industry timelines say a new page takes weeks to even get noticed and months to produce anything. We planned for that.

The field had other ideas. Two weeks after our first answer pages went live — before the timelines said the noticing would even be done — a visitor found one and reached out. A real lead, with a real project — an online services company asking us to redo their website.

A proud business owner holding a woven basket of her first picked vegetables at a farm stand

The first harvest: the first basket from your own field.

We won't pretend every page produces in fourteen days. Some fields take a full season, and honest timelines still matter. But that first lead set the floor — a floor, not a promise — and it's already holding: a second lead just arrived — another online services company, this one looking for help growing their business. The quiet months are quieter on paper than in practice. The scouting never stops. The answers start working as soon as they're found.

What this means for your business.

1The machines are already scouting your category.

AI engines have sent crawlers to our two sites more than 12,100 times this fall — reading, judging, deciding who to recommend. They're doing the same in your market right now. The only question is whether they find answers worth quoting when they reach your site.

2What they read is exactly what you'd hope.

Clear answers to real customer questions. That's plantable — page by page, month by month. For a plumber, it's the twenty questions every customer asks before calling, each with its own page. For a dentist, the same in their trade. No tricks, no ad spend. And everything we plant lives on your website and profiles, so it stays yours whether you stay one month or five years.

3It can start producing sooner than they tell you.

Not guaranteed-in-14-days sooner — anyone who promises that is selling. But the first harvest arrived before the industry's own timelines said it could, and it's repeating: a second lead is already in. The crawlers keep coming too — 13 of them now, from 10 companies, with new engines still showing up. Answers planted once keep getting read, month after month.

This is Growth Farming — the same method we used here, applied field by field to your business: search engines, AI answer engines, maps, reviews, and the platforms your customers actually use. Month-to-month. We earn the next month by delivering this one.

Ready to start growing?

No pitch, no jargon. We'll learn your business, tell you which fields are worth farming, and give you an honest read on what's possible — before you spend anything.