AI has made content production dramatically faster and cheaper. It has also created a new problem: the internet is filling with content that looks useful but often adds little original information.

In this episode of The Deeep Dive, we examine what happens when AI-generated content starts feeding other AI systems, why repeated information can create false consensus, and why traditional SEO approaches are changing as search engines and AI answer engines become more sophisticated.

The discussion starts with a simple question: How do you find trustworthy information when the web itself is increasingly difficult to verify?

When AI starts citing AI

One of the biggest risks discussed in the episode is the AI hallucination loop.

AI systems increasingly retrieve information from websites that may themselves contain AI-generated material. If the same unsupported claim appears across multiple websites, an AI system can interpret repetition as evidence of credibility.

That creates a dangerous cycle:

AI-generated content → AI retrieval → AI-generated summaries → more AI content → repeated claims → perceived consensus

The episode explores how this can turn an incorrect claim into something that appears authoritative simply because it has been repeated across enough sources.

The discussion also examines the importance of checking primary sources rather than trusting AI-generated citations, summaries or apparently authoritative references.

AI content is not automatically bad SEO

Another important distinction is between using AI and using AI to mass-produce low-value content.

The episode discusses Google’s approach to scaled content and why the important question is not simply whether AI was involved in creating a page.

The more useful questions are:

• Does the content provide original information?
• Does it demonstrate real expertise or experience?
• Does it answer the user’s actual question?
• Does it contain information that competitors do not have?
• Is the content useful beyond what an AI-generated summary could provide?

This leads to one of the central concepts of the episode: information gain.

Information gain is becoming a major content advantage

Information gain is the original value you add to a topic.

If 100 websites repeat the same basic explanation, another generic article adds very little. A page containing original research, proprietary data, expert interviews, customer insights, first-hand experience or a genuinely different perspective can contribute something new.

The episode breaks this down through practical examples:

Interviewing subject matter experts
Publishing original research
Using first-party customer data
Mining insights from customer support and sales teams
Publishing proprietary statistics
Adding first-hand experiences
Using real examples that competitors cannot reproduce

AI can help identify the existing consensus. Humans still need to supply the information that sits outside that consensus.

Why EEAT matters in AI search

The discussion connects information gain with Google’s EEAT framework: Experience, Expertise, Authoritativeness and Trustworthiness.

AI can summarize existing information extremely well. It does not automatically possess your company’s customer experience, proprietary data or first-hand knowledge.

That creates an opportunity for businesses.

Instead of using AI simply to produce more articles, organizations can use AI to process the information they already own and help turn that information into useful content.

Your customer complaints, campaign data, product usage patterns, research, interviews and internal expertise can become valuable sources of information that generic AI content cannot reproduce.

From prompts to reusable AI skills

The episode then moves beyond conventional prompting.

A basic prompt is usually a one-time instruction. An AI skill can be designed as a reusable workflow containing instructions, inputs, reasoning steps and output requirements.

This changes how businesses can use AI.

Instead of asking:

“Write me an article.”

An organization can create a repeatable process that tells an AI system:

• Which data to access
• Which sources to use
• What checks to perform
• Which reasoning steps to follow
• What information to exclude
• How to structure the output
• How to validate the result

This creates a more consistent approach to AI-assisted research, content creation and analysis.

What MCP changes

The episode also explores Model Context Protocol (MCP) and why connecting AI models to live business data can change how organizations use AI.

Rather than manually copying information into an AI chat, MCP can provide a standardized way for AI systems to interact with connected tools and data sources, subject to the permissions and implementation of the system.

The potential applications discussed include:

• Analyzing advertising performance
• Diagnosing CPM anomalies
• Finding patterns in successful campaigns
• Searching large content and asset libraries
• Identifying gaps in existing creative assets
• Working with proprietary business data
• Building data-grounded marketing workflows

The important shift is from AI that generates generic output to AI that reasons over business-specific information.

AI should handle the repetitive work. Humans should add the original thinking.

The episode also raises an important limitation.

If an AI system analyzes historical campaign winners and simply reproduces the same patterns, it can create another form of sameness.

AI can identify patterns in existing data. Humans still need to decide when to challenge those patterns, introduce new ideas and create experiences that have no historical precedent.

That distinction is central to the discussion:

AI can process the past. Human judgment can decide what to do differently next.

Optimizing content for humans and AI systems

Modern content also has to work across multiple discovery environments.

People need content that is readable, credible and useful.

AI search systems need content with clear entities, factual grounding, structured information, attributable claims and easily interpretable relationships.

This means SEO increasingly overlaps with LLM optimization, Generative Engine Optimization (GEO) and AI search optimization.

The goal is no longer simply to rank a page in traditional search results.

Businesses increasingly need their information to be:

• Found
• Understood
• Trusted
• Retrieved
• Quoted
• Correctly attributed

by both people and AI systems.

The question beyond AI optimization

The episode ends with a larger question.

If every company eventually has access to sophisticated AI agents, proprietary data connections, automated workflows and highly optimized content, those capabilities could become the baseline rather than the differentiator.

That raises an interesting possibility:

What happens when digital optimization becomes normal?

The answer may lie in the things AI cannot easily reproduce: first-hand experiences, original research, physical interactions, human relationships, real-world experimentation and ideas that have not yet been documented.

The future of AI-powered SEO may therefore involve more technology while placing greater value on genuinely human sources of information.

Listen to the Episode

In this episode, we explore:

• Why AI-generated content is creating information quality problems
• How AI hallucination loops develop
• Why AI systems can repeat unsupported claims
• The importance of primary-source verification
• Google’s approach to scaled content abuse
• Why AI-generated content itself is not the core issue
• EEAT and the role of first-hand experience
• Information gain and original content
• Using SMEs and first-party data
• Moving from prompts to reusable AI skills
• Model Context Protocol (MCP)
• AI agents and data-grounded workflows
• AI search, ChatGPT Search and Perplexity
• LLM optimization and content structure
• Why human creativity remains important in an agentic web