Zero-shot
In one line: Asking the AI to do something without giving it any examples. The opposite of few-shot prompting.
What is Zero-shot?
Zero-shot prompting means asking an AI to perform a task without providing any worked examples - Just the instruction. The model relies entirely on patterns from its training data. Modern LLMs are remarkably capable zero-shot because they have seen vast quantities of similar tasks during pre-training. Understanding when zero-shot is enough - And when you need to invest extra tokens in examples - Is one of the core skills of prompt engineering.
A concrete pair makes the difference obvious. Zero-shot: "Classify this review as positive or negative: [review text]". The few-shot version of the same task prepends two or three labelled reviews so the model can copy the pattern before seeing yours. Same model, same task; the only difference is whether you spend tokens demonstrating what you want.
Zero-shot vs few-shot vs fine-tuning
| Approach | Examples needed | Cost | When to use |
|---|---|---|---|
| Zero-shot | None | Lowest token cost | Common, well-defined tasks where model clearly understands the goal |
| One-shot | 1 | Slightly higher | When you need to demonstrate output format without many examples |
| Few-shot | 2–10 | Moderate (tokens per call) | Unusual formats, domain-specific style, subjective tasks |
| Fine-tuning | Hundreds to thousands | High upfront, low per-call | High-volume production use where consistency and style must be locked in |
When zero-shot is sufficient
Zero-shot works well for tasks that are common, well-defined, and where the model's training data almost certainly contained many examples:
- Translation between major languages
- Sentiment analysis and classification
- Summarisation of standard prose
- Spelling and grammar correction
- Simple code generation in popular languages
- Extracting named entities (dates, names, organisations) from text
For these tasks, adding examples rarely improves output quality enough to justify the extra tokens. Start zero-shot and only escalate if results are inconsistent.
When few-shot is worth the extra tokens
- Unusual output formats - Custom JSON schemas, bespoke table structures, or proprietary report templates the model has not seen before.
- Domain-specific terminology or style - Medical note format, legal brief structure, or a house style that differs from standard usage.
- Subjective standards - When the 'right' answer depends on your preferences and you need to demonstrate those preferences through examples.
- High-volume production tasks - When you need thousands of consistent outputs and variance between calls is unacceptable.
See also chain-of-thought prompting, which works in both zero-shot and few-shot settings by asking the model to show its reasoning steps. The prompt library includes ready-made zero-shot and few-shot templates for common tasks so you don't have to write them from scratch, and the FAQ on asking better questions covers the practical basics that make zero-shot prompts land on the first try.
Zero-shot example
If you are using AskAI.free, a practical way to understand zero-shot is to ask a model to explain it, then ask for a concrete example in your own workflow. For example: "Explain zero-shot for someone using AI to write, code, research, or create images."
This turns the term from a dictionary definition into a decision-making tool: you can see when it affects prompt quality, model choice, output reliability, privacy, cost, or how much context the AI can use.
Why Zero-shot matters
Zero-shot matters because it changes how you choose, prompt, compare or trust AI systems. If you understand this term, you can ask better questions, spot weak answers faster and choose the right model or tool for the job.
A common mistake is treating zero-shot as isolated jargon. It usually connects to nearby ideas like Tool use (function calling) and AI agent, so check those next if you want the full picture.
Common mistake with Zero-shot
The most common mistake is using the term as a label without changing behavior. When zero-shot comes up, ask what action should change: the prompt, the model, the input length, the evidence you request, or the way you verify the answer.
See it in action - Ask any AI about zero-shot on AskAI.free.
Try it free →