Introduction to Brandolini's Law

The sentence that captures our era

January 2013, XP Days conference in Milan. Alberto Brandolini, an Italian software craftsmanship consultant exasperated by a TV debate between a journalist and a politician, tweets a sentence that will become one of the most cited observations of the digital era:

"The amount of energy needed to refute bullshit is an order of magnitude bigger than that needed to produce it."

Brandolini didn't invent the idea — he made it operational. Before him:

• Jonathan Swift (1710) wrote: "Falsehood flies, and the truth comes limping after it."
• David Hume (1748) formalized: "An extraordinary claim requires extraordinary evidence."
• Carl Sagan (1980) revives Hume in Cosmos under the name Sagan standard — ECREE: Extraordinary Claims Require Extraordinary Evidence.

Brandolini adds the economic and industrial dimension: for the first time, we talk about a cost ratio between production and refutation. This quantification — an order of magnitude, i.e. ×10 — makes the law operational for businesses, sales teams, and AI builders.

Why this law explodes in 2024-2026

Three forces converge to transform Brandolini's Law from philosophical observation into operational risk #1:

1. Cost of producing bullshit → 0. An LLM generates 1,000 credible words in 3 seconds for $0.002. A "plausible but false" SEO article on your product: 30 seconds, $0.02.
2. Cost of refutation remains high. Refuting a claim requires research, sources, writing, legal validation, distribution — typically 4 to 40 hours of qualified human work.
3. Distribution asymmetry. A fake review on Trustpilot appears immediately to 10,000 monthly visitors; your rebuttal, if published, will be buried under 50 new reviews.

The cost-refutation / cost-production ratio was estimated at ×10 by Brandolini; in 2025, on LLM-generated content, we empirically observe ratios of ×100 to ×1000.

Five real-world enterprise scenarios

• 💼 B2B Sales: a prospect arrives at a demo saying "I read on Reddit that you had a security issue in 2023" — it's false, but the SDR must spend 20 minutes of the demo reassuring them instead of pushing value.
• 🛒 E-commerce: a competitor pays a review farm 2★ "the product broke in 2 days" — your real 5★ reviews don't compensate, your rating drops from 4.7 to 4.2, CTR collapses by 30%.
• 🏢 Recruiting: an ex-employee posts on Glassdoor "toxic management" — true or false, you can't reply individually, your candidate pipeline dries up for 6 months.
• 🤖 AI Support: ChatGPT tells a customer "your service includes X" — it's false, the customer screenshots it, your support spends 45 min explaining the difference between an LLM and official documentation.
• 📰 Product communication: a journalist misinterprets your roadmap, writes "startup X drops feature Y" — the article stays on Google page 1 for 18 months, your communication team produces 4 corrective articles that never dethrone it.

In every case, the cost of defense is asymmetrically higher than the cost of attack. That's the operational signature of Brandolini.

The underlying cognitive mechanics

Why does bullshit stick so well to the human brain? Three mechanisms documented in cognitive psychology:

graph LR
    A[Simple, vivid<br/>claim] --> B[System 1 accepts<br/>by default]
    B --> C[Strong episodic<br/>memory encoding]
    D[Complex, nuanced<br/>refutation] --> E[Requires System 2<br/>cognitive effort]
    E --> F[Weak memory<br/>of refutation]
    C --> G[Original claim<br/>dominates memory]
    F --> G
    style B fill:#ffcdd2
    style E fill:#e1f5fe

1. System 1 / System 2 asymmetry (Kahneman, Thinking Fast and Slow, 2011): the brain accepts by default any claim heard (System 1, fast), and only mobilizes critical thinking (System 2, slow, costly) on explicit signal. By default, there is no signal.
2. Continued influence effect (Lewandowsky et al., Psychological Science, 2012): even after explicit retraction, the initial information continues to influence judgment. The retraction merely reactivates the initial memory.
3. Illusory truth effect (Hasher, Goldstein & Toppino, 1977; Fazio et al., 2015): mere repetition of a claim increases its perceived truthfulness, regardless of its truth value. A bullshit repeated 5 times is judged truer than a truth said once.

These three mechanisms reinforce each other: a simple, vivid, repeated bullshit imprints; a complex, nuanced, rare refutation fades.

Three legendary Brandolini cases in business

1. The Audi accelerator scandal (1986)

A 60 Minutes segment claims Audi 5000 cars "accelerate on their own" and kill drivers. The NHTSA investigation, concluded 5 years later, demonstrates drivers were confusing accelerator and brake pedals. Audi US sales drop from 74,000 to 12,000 units. Audi takes 15 years to recover. Cost of bullshit (a 13-minute report) vs. cost of refutation ($6 billion and 15 years).

2. The P&G satanism rumor (1980-1995)

A rumor claims the P&G logo (a man in the moon with 13 stars) is a satanic symbol and that part of profits go to the Church of Satan. P&G spends over $10 million in lawsuits and communication campaigns over 15 years. The rumor only dies with the logo change in 1991 — proof that sometimes the only viable defense is removing the disputed object.

3. The "OpenAI sued for stealing X" cycle (2024)

In every AI news cycle, dozens of "OpenAI stole data from X" articles emerge — some true, many speculative. OpenAI's legal team handles ~30 articles/week. Cost of producing a speculative article: 5 minutes by a junior. Cost of refutation (reading, legal analysis, communication, sometimes lawsuit): 20 to 200 hours.

Course outline

By the end of the course, you'll be able to:

1. Recognize the Brandolini asymmetry in your sales, support, and marketing pipelines
2. Quantify the cost-attack / cost-defense ratio for your critical channels
3. Build a pre-bunking and single source of truth system
4. Detect and neutralize LLM hallucinations in your internal and customer-facing tools
5. Choose when to respond, when to ignore, when to remove the object of the rumor

Welcome to one of the most brutally operational laws of the AI era.