Write Like an AI Magnet: The Fact Density Formula

Research from the Princeton NLP lab is unambiguous: pages with at least 12 concrete facts per 500 words are 3.4× more likely to be cited by AI systems. But most content reads like marketing copy — vague, qualitative, and impossible for an AI to extract anything useful from.

What Counts as a "Fact"?

In the context of AI extraction, a fact is any statement that is:

• Verifiable — Can be confirmed or denied with evidence
• Specific — Contains named entities, numbers, or dates
• Discrete — Stands alone without needing surrounding context

Examples of extractable facts:

• "PostgreSQL 16 supports HNSW vector indexes via the pgvector extension" ✅
• "The CAX11 ARM instance costs €3.79/month and provides 2 vCPU, 4GB RAM" ✅
• "Cloudflare reported an 80% token reduction when serving Markdown vs. HTML (Feb 2026)" ✅
• "WordPress powers 43% of all websites according to W3Techs" ✅

Examples that are NOT extractable facts:

• "Our solution is industry-leading" ❌ (unverifiable)
• "Many companies have seen great results" ❌ (vague)
• "The market is growing rapidly" ❌ (no numbers)
• "We're passionate about quality" ❌ (opinion)

The Rewrite Method

Take any paragraph from your site and apply this three-step rewrite:

Step 1: Identify Vague Claims

"Our platform significantly improves website performance for many businesses, leading to better outcomes across various metrics."

Count the facts: zero. No numbers, no names, no verifiable claims.

Step 2: Add Specifics

Ask yourself for each claim:

• How much? (Replace "significantly" with a number)
• For whom? (Replace "many businesses" with specific segments or names)
• Which metrics? (Replace "various metrics" with named KPIs)

Step 3: Rewrite with Density

"The platform reduces page load time by 340ms on average (measured across 1,200 WordPress sites), improves Core Web Vitals LCP scores from 3.8s to 1.9s, and is deployed by e-commerce stores generating over €50M in annual revenue, including three Shopify Plus merchants in the DACH region."

Count the facts: 6 in one sentence. Load time reduction (340ms), sample size (1,200 sites), LCP improvement (3.8s → 1.9s), revenue scale (€50M), count (3 merchants), region (DACH).

Fact Types That AI Systems Favor

Density by Content Type

Different content types have different natural fact density targets:

Common Density Killers

These patterns destroy fact density. Find them and rewrite:

The filler opening:

"In today's rapidly evolving digital landscape, businesses need to stay ahead of the curve."

Zero facts. Delete it entirely or replace with a data point.

The vague comparison:

"Much better than the competition."

Replace with: "47% faster than Tool X's median response time (benchmarked Q1 2026)."

The emotion play:

"We're passionate about delivering exceptional results that exceed expectations."

AI can't extract passion. Replace with evidence: "Average client sees 2.3× increase in AI citations within 6 weeks of implementation."

The adjective pile-up:

"Our innovative, cutting-edge, industry-leading platform."

Replace all adjectives with one fact: "Processes 50,000 requests/day with 99.97% uptime."

The 12-Fact Audit

Before publishing any page, count the concrete facts in the first 500 words. If you're below 12:

1. Replace every adjective with a measurement
2. Add named entities (companies, tools, people, places)
3. Include dates for any temporal claim
4. Cite sources for any statistic
5. Add a data table if comparing anything

This single habit — counting facts before publishing — is the highest-leverage GEO improvement most sites can make.

Measuring Fact Density

The Zitably Score includes a Fact Density axis (/15) that uses LLM analysis to count:

• Named entities per 500 words
• Numbers and statistics per 500 words
• Verifiable claims vs. vague assertions

Run your pages through the analyzer to see where they stand and which sections need work.

Get your fact density score. Try the Zitably GEO analyzer →