By Christina D’oro Hooley
Long-form content has always been the golden child of SEO. More words mean more keywords, more backlinks, and more authority or so the algorithm tells us. But now, with generative AI as both consumer and creator of digital text, we may be pushing things too far.
Welcome to the age of AI logorrhoea a sickness of excess words. And we’re all to blame.
Why AI Is Getting “Sick”
Generative AI models like ChatGPT, Gemini, and Claude are trained on oceans of human-written content. They rely on patterns across billions of words to predict what comes next in a sentence. But here’s the twist: these models now learn from the content we write for them. And we’ve trained ourselves to write like machines.
Take SEO blog culture. It’s bloated. Articles that could say it in 300 words stretch to 2,000, packed with keyword fluff, listicles within listicles, and redundant summaries. That’s not writing that’s bait.
The result? Generative engines mimic us. They spew long-winded prose because that’s what we feed them. In trying to win the algorithm, we’ve created a new one and it’s repeating us, verbatim and bloated.
From SEO to GEO: A New Frontier
This isn’t just a content problem; it’s an ecosystem shift.
A recent paper by Gao et al. (2023) introduced the concept of Generative Engine Optimization (GEO) a next-gen version of SEO designed not for human readers, but for AI readers. In this world, formatting, semantic clarity, and strategic structuring beat brute length.
Because here’s the kicker: AI doesn’t reward verbosity. It rewards digestibility.
Science Agrees: Logorrhoea Hurts
Linguists and NLP researchers are raising the red flag. A study by Passali et al. (2023) shows how long sentences degrade model comprehension. Another by Cabanac et al. (2021) introduces “tortured phrases” awkward AI-generated rewrites that sound like bad translations of human thought.
We’re not just making AI sick. We’re eroding trust in content itself.
What Writers & Strategists Need to Do Now
- Stop padding. Value density over word count.
- Write for clarity, not just clicks. Generative engines are getting smarter; they detect fluff.
- Embrace structure. Use metadata, headings, and bullet points not as decoration, but as scaffolding for both human and AI cognition.
- Balance human warmth with machine legibility. If it sounds like a robot wrote it, it’s probably a robot who will read it and misread it.
Final Thought: The Paradox We Built
In our race to please the old gods of Google, we’ve birthed a new pantheon models that echo our own worst habits. But here’s the opportunity: we can still teach them better. The future belongs to those who can write with precision, personality, and purpose.
Because when everyone is vomiting content, the cure is craft.
Sources:
- Gao et al. (2023). Generative Engine Optimization (GEO). arXiv.
- Passali et al. (2023). From Lengthy to Lucid: NLP Techniques for Taming Long Sentences.
- Cabanac, Labbé & Magazinov (2021). Tortured phrases: A dubious writing style emerging in science.
- Mailchimp: The Rise of Generative Engine Optimization in SEO
Academic & Research Sources
1. Gao et al. (2023) “GEO: Generative Engine Optimization”
This foundational arXiv paper defines and dissects Generative Engine Optimization (GEO) a strategy tailored to influence generative AI systems (e.g. ChatGPT, Gemini) to cite and surface your content. GEO extends SEO into AI-driven landscapes using metadata, structured cues, and AI-specific formats to ensure long‐form text is favored in AI summaries
2. Dong et al. (2023) “A Survey on Long Text Modeling with Transformers”
This survey explores how NLP models particularly Transformer-based architectures process long documents. It highlights both architectural innovations and limitations in handling very long text, underscoring the computational challenges in generating and managing verbose content.
3. Passali et al. (2023) “From Lengthy to Lucid: … NLP Techniques for Taming Long Sentences”
Using PRISMA methodology, this work reviews strategies (like compression and sentence splitting) to improve clarity and coherence in extended texts. It directly addresses how to manage “logorrhoea” induced by generative models
https://arxiv.org/abs/2312.05172
4. Cabanac, Labbé & Magazinov (2021) “Tortured phrases: A dubious writing style emerging in science”
Identifies anomalous phrasing (“tortured phrases”) in AI-generated scientific manuscripts, linking them to AI-driven verbosity and poor semantic coherence, clear evidence of generative “logorrhoea” negatively affecting readability and trustworthiness
Leave a Reply