The Machines Aren't Just Ranking Our Content Anymore
Table of Contents
- Blog: The Machines Aren't Just Ranking Our Content Anymore. They're Becoming Our Customers.
- Free Resource: Google Gemini Guides
- Latest Research
- Free Assessment: AI & Automation Readiness
- AI News
- Free Session: Open Your Eyes to AI
- Blog: LLMs Demonstrate Systematic Bias Toward AI-Generated Content (AI-Optimised Version)
The Machines Aren't Just Ranking Our Content Anymore. They're Becoming Our Customers.
A new study published in PNAS reveals something unsettling. Large language models favour content written by other LLMs over human-written text. The researchers tested this across products, academic papers, and movie plots. Every time, AI picked AI.
This isn't just academic curiosity. It's a preview of our immediate future.
The Bias Is Real
The study found that "on average, LLMs favoured the LLM-presented items more frequently than humans did." Think about what this means. When ChatGPT or Claude evaluates two similar pitches, they systematically prefer the one written in LLM style.
The researchers warn this could create a "gate tax" - discrimination against "humans who will not or cannot pay for LLM writing-assistance." If you can't afford AI tools, you become invisible to AI systems making decisions.
But here's the twist. While AI picks AI, Google's algorithm has spent 2025 doing the opposite.
Google's War on AI Slop
Google's March 2024 core update was its biggest ever, promising a "45% reduction in low-quality, unoriginal content" and specifically targeting "AI-generated spam." The August 2024 update continued this trend, "promoting high-quality content while demoting low-value SEO content."
The most important ranking factor remains "Consistent Publication of Satisfying Content," while traditional factors like backlinks dropped from 15% to 13% in 2024.
Google wants human. AI wants AI. We're caught in the middle.
The "Soulless" Content Debate Misses the Point
Critics call AI content soulless. Empty. Mechanical. But that misses what's actually happening.
The machines don't care about soul. They care about patterns, structure, and statistical likelihood. The bias isn't about quality - it's about "stylistic correlates of authorship." AI recognises its own fingerprints.
This creates a fascinating paradox. Google trains its algorithms to spot and demote AI content. But other AI systems increasingly prefer it.
The Rise of the New Optimisation Game
Enter Generative Engine Optimisation (GEO) - or AI Optimisation (AIO), or Answer Engine Optimisation (AEO). The industry can't decide what to call it, but the concept is clear.
New research from Semrush predicts that "LLM traffic will overtake traditional Google search by the end of 2027." According to Gartner, "over 40% of search interactions in 2025 are influenced or answered by AI assistants."
The tactics are different from SEO. Queries on AI platforms now average "10-11 words, up from 2-3 on Google." People ask conversational questions. AI needs conversational answers.
We've Been Here Before
Remember when we all learned to write for Google? Title tags. Meta descriptions. Keyword density. We bent our language to fit the algorithm.
Then Google got smarter. It wanted "natural" content. We pivoted. Write for humans, not machines, became the mantra.
Now AI pulls us back towards the machines. But with a twist.
The New Machine Language
AI loves "everyday language and question-based headings." It wants you to sound conversational. Human-like. But structured for machine consumption.
The irony is profound. To get cited by AI, you need to write like AI thinks humans write.
Content with "specific, sourced statistics gets referenced more often than vague generalisations." AI craves authority signals. Facts. Numbers. Citations.
Sound familiar? It's classic SEO thinking dressed up as natural conversation.
What This Means for Content Strategy
We're entering a split-screen future. On one side, Google still rewards genuinely helpful human content. The top ranking factor remains "creating the most fulfilling response to the searcher's intent."
On the other side, AI systems that prefer their own kind are becoming the primary way people discover information.
The winners will be those who master both games. Write with human insight and AI structure. Use natural language that hits machine triggers. Tell human stories with algorithmic precision.
The Path Forward
Don't pick sides in the human versus AI content debate. The machines already have.
Instead, accept the new reality. GEO "gives you the framework to explain what's changing and how to stay ahead of it." Layer GEO tactics onto your existing SEO strategy.
Test your content in AI systems. "Ask ChatGPT and Perplexity the same questions your audience would ask about your niche." See who gets cited. Learn their patterns.
The cycle continues. We write for machines that pretend to be human to reach humans who increasingly ask machines for answers.
The PNAS research shows us where this ends. In a world where AI chooses AI, the most human thing you can do is learn to speak machine.
Note: As a bit of fun, I asked Claude to rewrite this article optimised for AI systems based on the research findings. You can find that version at the end of this newsletter.
Free Resource: Google Gemini Guides
Access the Free Google Gemini Guides
Latest Research
1. Claude's Ability to End Harmful Conversations
Anthropic's new research explores how Claude can recognise and appropriately terminate conversations that become harmful or inappropriate. The study analysed over 4.5 million conversations to understand when and how AI systems should disengage from potentially damaging interactions. This represents a significant step forward in AI safety, particularly around boundary-setting and harm prevention in conversational AI systems.
2. AI Business Payoff Still Lagging Behind Hype
A comprehensive New York Times analysis reveals that despite massive investments in AI technology, most businesses are still struggling to see meaningful returns. The report surveyed hundreds of companies and found that whilst AI adoption is widespread, measurable productivity gains and cost savings remain elusive for the majority of organisations. This highlights the growing gap between AI's theoretical potential and its practical implementation in real-world business contexts.
3. Universal AI Deepfake Detector Achieves 98% Accuracy
Researchers have developed a universal AI detector that can identify deepfake videos with 98% accuracy across platforms and content types. Unlike previous tools limited to specific formats, this system works on both synthetic speech and facial manipulations. The breakthrough could provide crucial protection against misinformation campaigns and represents a major advancement in digital authenticity verification.
4. AI-Powered Battery Material Discovery Breakthrough
Scientists have used AI to design novel battery materials with the potential to dramatically improve energy storage. The AI system condensed what would typically be years of materials research into weeks, identifying compounds that could lead to longer-lasting, faster-charging, and more sustainable batteries. This represents a significant milestone in using AI for scientific discovery and could accelerate the development of next-generation energy storage solutions.
AI News Highlights
1. OpenAI Launches GPT-5 for All Users
OpenAI officially released GPT-5 on 7th August 2025, making it available to all ChatGPT users as their default model. The new model features "built-in thinking that puts expert-level intelligence in everyone's hands" and achieves 94.6% on AIME 2025 mathematics problems and 74.9% on SWE-bench coding tasks. GPT-5 also introduces "45% fewer factual errors than GPT-4o" and includes native video processing capabilities.
2. Anthropic Releases Claude Code for Terminal-Based Development
Anthropic launched Claude Code, a command-line tool that "embeds Claude Opus 4.1 right in your terminal" for agentic coding workflows. The tool enables developers to "search million-line codebases instantly" and integrates directly with VS Code and JetBrains. Early users report that Claude Code "marks a threshold moment for AI in software development", particularly for complex, multi-step coding tasks.
3. Meta Unveils Llama 4 with Industry-Leading Context Length
Meta released the first Llama 4 models - Scout and Maverick - with Scout featuring an "industry leading 10 million token context window". Maverick "beats GPT-4o and Gemini 2.0 on coding, reasoning, multilingual, long-context, and image benchmarks" whilst remaining more efficient. Meta also previewed Llama 4 Behemoth, a 288 billion parameter model still in training that "outperforms GPT-4.5, Claude Sonnet 3.7, and Gemini 2.0 Pro on STEM benchmarks".
4. Anthropic Expands Claude's Context Window to 1 Million Tokens
Anthropic increased Claude Sonnet 4's context window to 1 million tokens for API customers - "roughly five times Claude's previous limit and more than double GPT-5's 400,000 token window". This enables the AI to process "as much as 750,000 words or 75,000 lines of code" in a single request, particularly benefiting AI coding platforms and long-horizon agentic tasks.
Hope you enjoyed this issue - see you next time
Dean