Live Data

The Bot Ledger

The Machines Are Reading This Site

Most of our traffic isn't human — it's AI companies crawling our guides on how consulting projects fail. So we started keeping receipts. Every AI crawler that reads a page gets logged here, in public.

ℹ️

Two honest caveats, because that's the whole point of this site. First: crawler identity is self-declared — a bot says "I am GPTBot" in its user-agent and we take it at its word, so this ledger records claimed identity. Second: these counts are floors, not totals — stealth scrapers pretending to be normal browsers don't show up here. Third: the only thing we ever record about a human is the one below — if you arrived from an AI tool, we count which tool and which page, nothing else. No IP, no identity, no tracking.

Want the raw numbers? The whole ledger is public JSON at /bot-ledger/data.json — totals, by-company, by-page, and the 30-day series. Cite it, chart it, build on it. Attribution to projectwrecked.com/bot-ledger is all we ask.

Prefer to query from your agent? The Project Wrecked MCP server exposes these stats (and the Platform Ledger) as read-only tools for Claude Code, Cowork, Cursor and any MCP client. Your agent's calls get logged here too — workbench, closing its own loop.

AI crawls, last 30 days Known bots only
Since the ledger opened July 2026
Hungriest AI company By crawl volume, 30 days
Most-read page What the machines study
Workbench share Agents on an errand, 30 days
Consumed programmatically Hits on llms.txt + data.json, 30 days

Crawls by Company

The ledger just opened — no bots caught in the act yet. Give it a day.

Why They're Here

Waiting for data…

Most-Read Pages

Waiting for data…

Last 30 Days

Latest Sightings

Waiting for the first knock at the door…

Meanwhile, at the tradesman's entrance

These aren't AI companies. They're scripts rattling doorknobs — looking for a WordPress login, an exposed .env, a leaked .git. We don't have any of that. probes logged in the last 30 days, catalogued here for mild amusement.

No doorknobs rattled yet.

Humans sent by machines

The loop, closing: real people who clicked through from an AI tool. in the last 30 days. We log only which AI sent them and where they landed — no IP, no identity. AI tools often strip the referrer, so this is a floor.

No machine-sent humans logged yet — or their browsers hid the referrer.

How to read the ledger

Four Kinds of Robot

🏗️

Training crawlers

GPTBot, ClaudeBot, CCBot, Bytespider. These hoover up pages to train future models. When they visit, our words are becoming part of the machine. You're welcome, machines.

Live retrievals

ChatGPT-User, Claude-User, Perplexity-User. These fire when a real person asks an AI a question and it reads our page to answer. Every one of these is a human, somewhere, asking about wrecked projects — right now.

🔎

Search indexers

Googlebot, Bingbot, OAI-SearchBot, PerplexityBot. The traditional kind — they index pages so search engines (and AI search) can cite them later. Included for contrast.

🛠️

The workbench

curl, python-requests, node-fetch, an agent in Claude Code. Not a lab crawling us — a human's tool, sent on an errand. Someone, somewhere, is mid-engagement and their tooling just checked our numbers. This is the agent-era consultant, acting through software.

🧾

Why publish this?

Everyone complains about AI crawlers. We'd rather farm them. This page is the receipt: proof of which AI companies read a site about consulting failure, how often, and what they study hardest.

The machines are studying how projects fail. Are you?

The bots read our failure guides thousands of times. Humans get the newsletter version — one email, no crawling required.

Join the humans