Here's a number that should change how you think about content: 42% of B2B buyers now start their research in AI search tools — ChatGPT, Claude, Perplexity — before ever visiting a vendor website.
That's up from 11% just twelve months ago. This isn't a trend line. It's a structural shift.
For three decades, content strategy has meant one thing: write for humans, optimize for search engines. The audience was always, ultimately, a person. That's no longer true.
This week: how to write for two audiences — humans who read and AI systems that parse. Plus: an explainer on MCP, the protocol that will reshape how AI finds your product.
—Chris
The industry is starting to call this Generative Engine Optimization (GEO) — the practice of structuring content so that AI systems can accurately understand and represent it.
What works for GEO is different from what worked for SEO:
| SEO (Optimizing for Search) | GEO (Optimizing for AI) |
|---|---|
| Keywords and phrases | Structured data and schemas |
| Backlinks and domain authority | Accuracy and verifiability |
| Click-through optimization | Question-answer clarity |
| Page rank position | Inclusion in AI responses |
| Human readability first | Machine parseability first |
The uncomfortable truth: a lot of content that ranks well in Google is nearly useless to AI systems. Long-form blog posts with keywords sprinkled throughout don't give AI what it needs — clear, structured, verifiable answers to specific questions.
When an AI assistant tries to answer a question about your product, it's looking for:
Not "Our enterprise-grade platform delivers ROI" but actual specifications: what the product does, what it integrates with, what it costs, what's included at each tier.
Related: Evertune helps brands monitor and optimize how they appear in AI-generated search results.
The actual questions buyers ask, with direct answers. Not buried in a 2,000-word blog post — explicitly structured as Q&A.
Related: See how Evertune tracks AI visibility across ChatGPT, Claude, Gemini, and Perplexity.
AI systems are increasingly checking claims against multiple sources. Unsubstantiated superlatives ("industry-leading," "best-in-class") get filtered out or flagged as unreliable.
JSON-LD schemas, structured data markup, clean APIs. The stuff that feels like "developer work" is becoming PMM-relevant.
There's a new acronym making the rounds in AI infrastructure circles: MCP, or Model Context Protocol. Think of it as USB-C for AI — a universal standard that lets any AI system connect to any data source.
MCP is how AI agents will query your product specs, check your pricing, compare your features to competitors, and make recommendations.
If your content isn't structured in a way that MCP servers can expose, you become invisible to a growing share of purchase decisions.
MCP servers expose three types of things:
When an AI agent encounters a question it can't answer from training data — "What's the pricing for Enterprise tier?" — it can query an MCP server to get real-time, authoritative information directly from you.
Here's a practical framework for creating content that serves both humans and AI:
Narrative storytelling, brand voice, emotional resonance, visual design. The content that makes humans want to engage, trust, and buy.
Schema.org markup, structured FAQs, clean JSON for specifications, consistent taxonomy. Data that AI can parse underneath the human-readable layer.
For companies ready to go further: endpoints that AI agents can query for real-time pricing, inventory, compatibility. The stuff that lets agents transact, not just research.
The key insight: You're not replacing human-focused content with machine-focused content. You're adding a layer underneath it. The blog post still exists for the human reader; the structured data exists for the AI agent that needs to pull a specific fact from it.
From 35+ stories this week, we stack-ranked the five with the biggest PMM impact.
The killer stat. Nearly half of buyers form opinions in ChatGPT/Claude/Perplexity before visiting your site. This changes everything about content strategy.
Not a rebrand — a re-architecture. Marketing becomes dialogue, not broadcast. The "do not reply" era is dead. Category-defining move.
Why AI crushes code but struggles in marketing: your logic lives in Slack threads, not systems. McKinsey's $400-660B opportunity requires codified knowledge.
The B2C companion to #1. Combined, these stats show the buyer's journey has fundamentally rewired — across both B2B and B2C.
Built "in a weekend." The shift from AI-as-advisor to AI-as-operator is here. Which parts of your job require human judgment?
by Eli Schwartz · SEO Strategist & Author of "Product-Led SEO"
Eli has been at the forefront of the SEO-to-GEO shift we cover in this issue. His recent writing on how AI is reshaping organic discovery — and what marketers need to do about it — is essential reading for anyone thinking about visibility in an AI-first world.
Deep expertise, practical frameworks, and a clear-eyed view of what's actually changing vs. what's hype.
Follow Eli Schwartz →The content strategy that worked for the last decade — write for humans, optimize for Google — is becoming incomplete. The new job is writing for two audiences simultaneously: humans who need to be persuaded and AI systems that need to be informed. The PMMs who figure this out early will have an advantage. The ones who wait will find themselves invisible to a growing share of the buyer's journey.
Everything PMMs need to know about Model Context Protocol — the standard that will reshape how AI finds your product.
Read the Explainer →A newsletter about AI, product marketing, and what comes next.
You're receiving this because you subscribed at thefullstack.chrisohara.com