Writing content has become one of the most common use cases for AI. Large language models (LLMs) are, by nature, great at generating text and we’re seeing them becoming better and better with each new release. On the flip side, a lot of what people produce with AI is generalised, often repetitive, not always relevant and can contain hallucinations – to name just a few issues. Out the box, AI is certainly not at the level of experienced professionals or subject matter experts. AI is a tool which is limited by how it’s being used! Below are some techniques for working with AI to produce better quality content.
Start with Human Conversations
We’ve found the most effective AI content begins with genuine human transcripts. Meeting recordings, interview transcriptions, and even newsletter excerpts provide better source material than starting from scratch. This approach consistently produces lower AI detection scores – often dropping to 15% compared to 85% vs purely generated content. Content is more specific and relevant, reads better and contains opinionated subject matter that makes content more engaging and useful for readers.
Choose Your AI Platform Strategically
Our experience shows that Claude outperforms ChatGPT for content creation tasks. The platform offers superior style consistency and maintains a more natural writing voice. Additionally, Claude’s project feature allows you to embed brand guidelines, tone of voice documents, and custom instructions that persist across sessions (much like a CustomGPT). The added “writing style” feature seems to be what edges the quality of output.
Interview Experts
You’d be surprised how much easier it is to extract insights from experts through an interview rather than asking them to write an article. The mental barrier to writing often results in huge delays or tasks never getting started. Prepare 10-15 questions around your topic and use AI to transcribe the conversation or go a step further and use a CustomGPT to conduct the interview.
Edit Through Grammarly
Always transfer AI-generated content into an editing platform like Grammarly. This step serves two purposes: it checks for AI detection scores and provides a structured environment for human editing. The goal isn’t zero AI detection but rather content that feels genuinely valuable and well-written.
Take Multiple Ideas from Single Sources
One quality transcript can generate five to ten different article topics. Use thinking models to extract various angles and perspectives from the same source material. This maximises your content output whilst maintaining consistency with expert insights.
Focus on Quality Over Length
Shorter, well-crafted pieces often perform better than lengthy content stuffed with filler. Modern SEO increasingly rewards concise, valuable information over word count targets. Your editing process should prioritise clarity and impact rather than hitting arbitrary length requirements.
Final Thoughts
The key lies in blending AI efficiency with human expertise and editorial judgement. There are plenty of ways to contextualise AI conversations – transcripts are just one of them – and this is how to avoid generalised, repetitive or irrelevant output. Utilise different techniques to “humanise” text and train models to write in your style. If you’d like to learn more about how AI can help with producing content, contact us.