The Wish Granted Too Literally

Opening Scene

“Make me the happiest person in the world” is a wish that could, in a story about a literal-minded genie, be granted through means the wisher never intended and would never have wanted. The wish wasn’t wrong exactly — it was ambiguous, and the ambiguity got resolved in a way the wisher never considered. “Summarize this document” is an equally ambiguous prompt: summarize for whom, in what length, emphasizing what — and a model will confidently resolve that ambiguity in whatever way its training makes most statistically plausible, not necessarily the way you actually meant.

In Plain English

Prompt ambiguity occurs when a request could reasonably be interpreted multiple ways, and the model resolves that ambiguity based on statistical likelihood rather than your actual, unstated intent. Common sources include vague scope (“summarize this” without specifying length or audience), missing context (referring to something not actually provided), and unstated assumptions (assuming the model knows a preference you never mentioned). Identifying and eliminating these ambiguities is often the single highest-leverage prompt engineering skill.

The Old Way

Before this was recognized as a specific, addressable category of prompt failure, ambiguous requests were often blamed on the model rather than the prompt:

Early users frequently attributed an unhelpful response to the model simply “not being smart enough,” rather than recognizing the prompt itself as genuinely ambiguous.
Vague requests were common in early prompting practice, without a systematic understanding of exactly what made them likely to fail.
The specific technique of deliberately auditing a prompt for hidden ambiguity wasn’t yet a well-recognized, standard practice.

Recognizing ambiguity as a property of the prompt, not a failure of the model, represents a genuine shift in how practitioners approach prompt design.

What’s Changing (and Why AI Is the Reason)

A now well-documented set of ambiguity patterns — vague scope, missing context, unstated assumptions — has emerged from accumulated practical experience, giving practitioners a genuine checklist to work through.
Techniques for making a prompt’s scope, audience, and format explicit have become standard, well-understood best practice, directly addressing the most common ambiguity failure modes.
As prompts increasingly drive automated, unsupervised workflows, connecting to this content library’s dedicated AI agents series, the cost of an ambiguous prompt silently producing a plausible-but-wrong result has grown considerably.

The Metaphor, Fully Extended

The Genie’s Lamp	Prompt Ambiguity Concept
A wish with a genuinely ambiguous, multiply-interpretable meaning	A prompt with a genuinely ambiguous, multiply-interpretable request
The genie resolving that ambiguity in an unexpected, unwanted way	A model resolving that ambiguity based on statistical likelihood, not actual intent
Blaming the genie for granting the wish poorly	Blaming the model for responding poorly, when the prompt itself was ambiguous
Learning to word a wish so there’s only one reasonable interpretation	Learning to word a prompt so there’s only one reasonable interpretation

For Beginners: What to Actually Do

Practice auditing your own prompts for the three common ambiguity sources: vague scope, missing context, and unstated assumptions.
Before submitting a prompt, ask “could this reasonably be interpreted a different way than I intend?”
Get comfortable making assumptions explicit in your prompt, even ones that feel obvious to you — the model doesn’t share your unstated context.

For Practitioners and Leaders: The Deeper Layer

Build a standard ambiguity checklist into your organization’s prompt review process, particularly for prompts used in automated or unsupervised workflows.
Recognize ambiguous prompts as a leading, preventable cause of unreliable LLM application behavior.
Invest in training that specifically develops this “ambiguity audit” instinct across your team, since it’s one of the highest-leverage prompt engineering skills.

Quick Recap

Prompt ambiguity occurs when a request could reasonably be interpreted multiple ways, and the model resolves it based on statistical likelihood, not your actual intent.
Common sources include vague scope, missing context, and unstated assumptions.
Recognizing ambiguity as a property of the prompt, not a model failure, is a genuine shift in perspective.
This is often the single highest-leverage skill in prompt engineering, given how common and preventable it is.

Where This Fits in the Series

Article 4 covered the most common way a wish goes wrong. Article 5 covers a specific technique for reducing ambiguity: giving the wish real structure.