Jamie Brownlee.
Jamie Brownlee, co-founder and CEO at Brainwaves
Ask ChatGPT for strategic marketing advice and it will give you the same answer it gives everyone else, regardless of context.
New research published in Harvard Business Review confirms that Large Language Models (LLMs) are world-class wafflers serving up strategy slop. The research tested seven leading LLMs across seven core strategic trade-offs. The results: near-identical recommendations across every model, regardless of context, company type or how carefully the prompt was crafted. More alarmingly, feeding the models a richer business context barely shifted their responses.
The researchers called it 'trendslop'. LLMs tend to gravitate not toward what's strategically sound, but toward the buzzwords that dominate the discourse they were trained on. Confident, polished, and completely interchangeable.
The Slop Diet
To understand why, it helps to look at where AI does perform well. Coding agents now write an increasing share of real, working software (ours included). The difference is the diet.
Code either works or it doesn't. Broken code gets fixed or binned; it doesn't get published, shared, or built on. So LLM training data is naturally filtered toward code that actually works.
On the other hand, the marketing content LLMs have been trained on is dominated by performance-driven strategy: post-rationalised case studies written to win awards, thought-leadership written to generate likes, and blog posts written to drive clicks.
Whatever peer-reviewed research, marketing science textbooks, and rigorous effectiveness studies that do exist are a rounding error against the volume of survivorship-biased, boardroom-optimised content that surrounds them.
What the model learns to predict is the consensus that performed best online: strategy as it was presented, not as it was practiced.
So, it’s no surprise LLMs are serving up strategy waffle with zero competitive nutrition.
Ordering Off-Menu
Trendslop is what happens when you ask AI to decide for you. Order off-menu and AI becomes an amplifier of strategic intuition, expanding the field of possibility beyond what any one mind can reach alone. When you stop asking for an answer and start using AI to map what’s possible, you gain the capacity to:
- Synthesize vast unstructured data to identify previously hidden opportunities.
- Map existing patterns within a category to identify emerging white space.
- Generate volumes of thought-starters that force your thinking past the obvious.
- Simulate how a strategy lands across different cultural contexts and audiences.
- Identify the structural holes and gaps in your thinking before you launch.
None of this is AI making the strategic choice. It is AI making you capable of strategy that is broader, deeper, and harder to reach than the consensus.
To get there, you need to feed your AI a diet they were never trained on.
What to put on the plate instead
Better prompts won't fix a slop diet, the researchers confirmed. To get a unique answer, you have to feed the model the unique context that doesn't exist in any training set:
- Your Brand: Your positioning, strategy, and real performance data.
- Your Memory: Past learnings, do's and don'ts, and your best work.
- Your Methods: The playbooks, processes, and frameworks that actually work for your brand and category.
Start by treating it as a collaborator that needs a proper briefing, not a magic box. The more proprietary the input, the less generic the output.
For those looking to truly order off-menu, purpose-built agents take this further. Unlike general assistants, specialist agents are built on domain expertise and best-practice methodologies, trained on how strategy actually works rather than how it gets written up after the fact.
After all, garbage in, garbage out. Feed your AI something worth biting into and the thinking that comes out the other side is yours, not everyone else's.