We released the latest white paper in our SEO series this week: AI in Content Creation. This email is a summarised overview.
In just over two years since ChatGPT launched, the web has seen a surge in AI-written content. Large Language Models (LLMs) like ChatGPT, Claude and Gemini can now write coherent articles, which are often indistinguishable from human writing. But what does this mean for SEO?
We’re entering a phase of ‘content ubiquity’ - a world where vast amounts of content are created instantly and cheaply. That’s already raising big questions about originality, quality and how Google sorts that web’s information.
Can you rank on Google with AI generated content?
Yes, with caveats. Google has made it clear: AI content is fine as long as it’s helpful. The March 2024 core update cracked down on low-quality, unoriginal content - slashing the visibility of ‘thin’ pages. Google doesn’t penalise content just for being AI generated, but if it doesn’t provide ‘information gain’ on what’s already out there, it’s unlikely to get much traction.
The key is combining AI-generated drafts with human oversight. A Semrush study showed that 73% of marketers using AI combine it with editorial review - and many saw traffic increases.
Semrush Content Shake: Can AI Content Rank on Google?
Should AI content be disclosed?
Technically, it doesn’t have to be. But ethically - and practically - yes. Google encourages transparency: users should be able to tell how and why content was created. The majority of Internet users would also like to see disclosure when content was created by AI.
This isn’t just theoretical. CNET and Sports Illustrated both got called out for publishing AI-written content without disclosure. Both faced audience backlash. Adding a byline from a real expert or reviewer builds trust - and trust builds authority.
How can you create good AI generated content?
Good prompting is key to getting useful chatbot responses. We use the ‘RACE’ framework:
Role – Assign a persona (e.g. fashion editor)
Action – Define the task (e.g. write an article)
Constraints – Set tone/length/style
Examples – Provide structure
A similar structure was provided by OpenAI President Greg Brockman:
Upon receiving a chatbot’s draft, human editors can Verify, Embellish, and add Insight (VEI). AI can give you speed. Humans give the real content depth.
Tools like Make or n8n can integrate with LLMs to automate publishing workflows. But doing this well still needs oversight. Otherwise, you’re serving the same bland content as everyone else. Think of AI text like tap water—cheap, accessible, but not very exciting unless you do something more with it.
What do humans need to add?
The Google guidelines have been clear for a while now. Content needs to display Experience, Expertise, Authoritativeness and Trustworthiness (EEAT). LLMs are predictive text engines rather than thinking machines, and have no real world experience. What they produce is essentially a rehash of what they have been trained on.
Thus if you have an article about places to go in London, it’s essential the writer has actually been there - else there is very unlikely to be any information gain. Same if you’re listing or testing products - has the writer actually used them? What was their experience? These are things AI cannot provide.
Google has not done a great job historically of prioritizing authors in its index. With AI commoditizing content production, this may change.
There is definitely turbulence in the jobs market for content roles. In many cases these are being trimmed back as companies see an opportunity to cut costs in content production. But we see this as problematic because authors will probably become a more important ranking factor. Creating lots of anonymous articles won’t cut it.
If you’d like to go deeper into this topic, read our latest white paper.
