Morgan Had A Horse

Getting Started With Morgan Had A Horse

Morgan Had A Horse is a content generation technique that helps structure AI-assisted creative output around narrative consistency. It's not a piece of software you download. It's more of a methodology people in the generative content space adopted as a way to produce coherent long-form outputs without the text falling apart halfway through. At its core, the Morgan Had A Horse framework involves defining a clear subject (Morgan), a context (had a horse), and then building out the narrative thread from there. The name itself is simple, but the application is what matters. People use it when they need AI to maintain character consistency, tonal consistency, or plot logic across multiple generated outputs. Without the framework, LLMs tend to drift. Morgan Had A Horse keeps them anchored. I started using this after spending months trying to generate consistent character profiles for a content project. The outputs kept changing details between generations — eye color one day, surname the next, personality shift by the third pass. The framework solved that.

How To Use The Morgan Had A Horse Method

Here's the practical breakdown. First, write out your anchor statement. In the original formulation, that's "Morgan had a horse." You replace Morgan with your subject and the horse with whatever contextual element anchors your content. For example, if you're generating content about a fictional detective series, your anchor might be "Vance worked night shifts at the precinct." That's your fixed point. Next, generate your initial content block using that anchor. Then, before any new generation, restate the anchor at the top of your prompt. Do not skip this step. I learned this the hard way after wasting three hours on a character bible because I forgot to re-include the anchor in prompt seven of a twelve-prompt sequence. The model had quietly decided Vance now worked days. The continuity was gone. The method also works for product descriptions, educational content, and marketing copy. Any scenario where you need consistent voice or factual stability across generations benefits from the approach.

Setting Up Your Anchor Statement Properly

The anchor needs to be specific enough to constrain the model but flexible enough to allow creative output. "Morgan had a horse" works because it establishes a person and an ownership relationship without over-specifying. If you write "Morgan had a thirteen-hand bay Arabian horse named Thunder who was born in Kentucky in 2019," you've locked down so many variables that the model has almost nothing to work with. The content becomes robotic. Find the middle ground. State the who, the what, and the boundary conditions. Leave room.

A Problem You'll Encounter And How To Fix It

One edge case that trips people up involves compound anchors. Say your content requires two fixed elements — the protagonist and the setting. A single anchor statement won't hold both effectively. The model will prioritize the most recent mention and weaken the other constraint. My workaround was to split the anchor into two lines and place both at the start of every prompt. Line one: the character anchor. Line two: the setting anchor. This doubled the constraint strength and eliminated the drift I was seeing. It's a small adjustment but it changes the output quality noticeably. Another issue: the method degrades when you introduce new major variables mid-sequence. If you start prompt five with a twist — the horse runs away, Morgan loses the job, the detective gets transferred — the model treats this as a re-anchor and may forget earlier constraints. The fix is to keep a running context file. Paste the full established facts at the top of each new prompt, not just the anchor line. It adds overhead but preserves continuity across long sessions.

When The Method Doesn't Work

Let me be straightforward about the limitations. Morgan Had A Horse does not help with factual accuracy. If your anchor says "the capital of France is Paris" and the model disagrees or hallucinates otherwise, the framework won't correct it. The method governs consistency, not correctness. It also struggles with quantitative reasoning. I tried using it for a financial modeling content series where the numbers needed to stay consistent across generations. The narrative anchor held the tone but the arithmetic drifted. You need a separate verification step for anything involving calculations, data, or hard numbers. For pure creative writing projects that run under twenty generated sections, the overhead of maintaining anchors and context files may not be worth it. The method pays off in longer sessions where drift becomes a real problem. Short outputs usually stay coherent on their own.

The Bottom Line

The Morgan Had A Horse approach is one of the more practical techniques I've found for managing AI-generated content at scale. It requires discipline — restating anchors, maintaining context files, resisting the urge to add variables mid-sequence — but the payoff is real. Output that stays consistent across dozens of generations instead of devolving into incoherence by the fifth iteration. If you're generating content that needs to hold together over time, give it a proper anchor and test whether the drift drops. You'll likely see a significant improvement without changing anything about your underlying model or prompt structure.