< Back to Insights

9 AI prompting techniques: how to get better results every time

From zero shot to conversational prompting, our AI expert shares the different AI prompting techniques you can use for better outputs.

It’s tempting to use AI like a Google search – type a few words and hope for the best. But AI isn’t a search engine. It’s a collaborator that needs context, examples, and clear instructions to do its best work.

That means the quality of your AI output relies on the quality of your prompt. It doesn’t matter if you’re using the best AI models in the world – if your prompts suck, so will the results. Vague prompts get vague answers and, by the same token, specific, well-structured prompts get specific, useful results.

Or, in other words, good in, good out.

But in these early days of working with AI tools, I’ve found there are two Ps that are more important than anything else: perseverance and prompting. It’s these two that will help you discover the real benefits of AI.

You’ll need to bring the first yourself. But on the second, we’ve got a few pointers.

Prompting techniques: an explainer

Prompting simply means the messages you give the AI to steer what it produces. It’s how you describe the outcome you want, whether that’s a single sentence request or a 20,000‑word set of instructions complete with tone of voice guidance and examples. (We love those, obviously.)

The fun part: you can try different techniques, get creative, mix and match. Whatever achieves the best outputs.

Ready to write brilliant prompts?

If you’re just getting started

Start simple: try zero‑shot prompting.

This is the simplest, quickest way to interact with AI. Ask what you want without using any examples to guide the output, and away you go.

Example:

“Write me a job ad.”

### The impact

Zero‑shot prompting is a good way to get familiar with how AI thinks and responds. It’s quick, low effort, and great for fast answers. It’s always a good way to start a conversation with the AI.

But when you need something more specific – like a detailed job ad – it’ll only get you part of the way there. The results can feel generic or vague because you haven’t told the AI what “good” looks like yet.

To really level up your outputs, here are 10 prompting techniques that’ll transform how you use AI – whether you’re writing copy, generating images, or analysing data.

1. Give it a role – and an audience

Role prompting is when you tell the chatbot to, for instance, ‘Act as a chief marketing officer with [insert specific experience]’. For extra guidance, explain to the AI who the audience is too.

And you can get creative too if you wish, giving it a name, a job title and a bio of sorts (just make sure it’s relevant).

Example:

“Act as a senior recruitment manager called Kelly Jobs, with extensive experience in writing job ads for agencies that attract the right candidates.

“You’ve been commissioned by your client, Definition, to write a job ad for a Senior Writer role. This is a very lucrative commission…”

The impact

AI models have been trained on an unimaginable amount of data. By giving it a role, you help direct that vast knowledge toward a specific perspective or purpose, making its responses more focused and context‑appropriate.

Role prompting can improve the performance of AI on tasks. Adding one to your prompt may not improve the factual accuracy of the AI, but it will likely impact the style, tone and depth of the answer, and that can make a massive difference in how useful the output actually is.

2. Share examples

Don’t just say what you want, give the AI examples to follow. It’s called few-shot or multi-shot prompting, and it will help you improve the accuracy of your output, especially if you want content written in a particular style or tone.

Example:

“Write me a job ad.

“The job is for a Senior Writer position at Definition.

“Here are two examples of previous Definition job ads. Please follow these in terms of structure, style and tone of voice….”

The impact

Research shows modern large language models register more than a 50% increase in accuracy when provided with examples in a prompt.

3. Split your prompt into sections

Use XML tags (beginning with a word in chevrons, and ending with the same word in chevrons with a forward slash to indicate it’s the closing tag) to separate sections of your prompt, like keeping the ‘task’ section apart from ‘context’, for example.

You can also use sets of special characters that don’t naturally occur in text to separate parts of your prompt (e.g. three ampersands in a row), but we’d advise using XML tags for two reasons:

1) they were used in the training data the models were exposed to during creation – therefore they naturally recognise them

2) you can describe what each section is when using XML tags e.g. <content>, <examples>, <instructions> etc.

Example:

Summarise the <text> below as a bullet point list of the most important points.

<text>
[text input here]
</text>

The impact:

Using XML tags – in other words, using ‘delimiters’ – clearly separates instructions from content, reducing ambiguity. And that’s crucial for helping the AI to stay organised and follow directions more accurately.

4. Tell the AI what you want it not to do

While your prompt should be full of clear, direct, affirmative instructions, adding what it should avoid doing reduces the chances of confusion, leading to better outputs.

Example:

Write a product description for our B2B software. Do NOT use buzzwords like ‘revolutionary,’ ‘game-changing,’ or ‘cutting-edge.’ Do NOT make unsubstantiated claims. Do NOT use exclamation marks. Focus on specific features and measurable benefits.

### The impact:

By telling the AI what not to do, as well as what it should do, you’re narrowing the parameters and reducing ambiguity – giving it a much better chance of returning exactly what you want. If in doubt though, focus on affirmation.

5. Chain your prompts

Chain prompting is more like having a conversation with the AI, where, through a back and forth, you’ll feed the AI the information it needs to produce an accurate output.

The real upside of this is that you don’t need to squeeze everything in one perfect prompt. Instead, break down complex goals into separate prompts and chain them together.

Example:

Step 1: “Give me 10 ideas for a blog post”
Step 2: Pick the best idea and prompt: “Create a detailed outline for [chosen idea]”
Step 3: “Write the introduction section based on this outline”
Step 4: “Now write the first main section…”

You get the picture…

The impact:

By breaking down your instructions into smaller, easier steps, you’ll give the AI a better chance of keeping its reasoning functions more organised.

6. Get the AI to write your prompt

For a prompt that really lands, tell the AI what you’re after, then get it to create a killer prompt. This is known as ‘meta prompting’ in the industry.

The impact:

No one knows how better to communicate with AI than AI itself. So using meta prompting, as it’s known, lets the model refine its own instructions, leading to clearer reasoning and higher‑quality results.

This tactic can be extended if the prompt generated by the AI isn’t doing what you want it to do. Provide the AI with the prompt and any undesirable outputs it generated, before asking it to rewrite the prompt to fix those behaviours.

7. Demand a second opinion

Challenge the outputs you get. Ask, for instance, “Are you sure?” This pushes the AI to re-evaluate its response, check for errors and consider different angles. You’ll often get a better answer the second time around.

When this is most valuable

Factual claims (i.e. dates or statistics)
Data analysis
Nuanced interpretations or advice
Anything where accuracy is critical.

8. Summarise your examples

Want the AI to write in your tone of voice, but the tone guidelines are 10,000 words long? Ask the AI to summarise your guide, then paste the shorter version into your prompt.

The impact:

On tone content, yay!

9. Bring in the experts

Writers prompt the best writing. Designers prompt the best images. Directors prompt the best videos. Good work comes from good prompts, which come from talented humans.

With Definition AI, you’ll get all that: the best AI models, great prompts and human experts.

Nick Banks Screen