Prompt engineering: how to make AI work for you
Alex Olsson
SEO Specialist
What is prompt engineering?
Prompt engineering is all about crafting clear, specific instructions that guide AI to deliver relevant and accurate responses. It’s a mix of critical thinking, experimentation, and a bit of trial and error to nail down the right phrasing, level of detail, and format that yield the best results.
Key considerations for creating a good prompt
1. Clarity
Simple language goes a long way in helping AI understand the task. Avoid vague phrasing and aim for straightforward wording.
2. Specificity
The more details you add, the more likely you’ll get a tailored response. Mention what you’re looking for, any specific context, and the desired outcome.
3. Conciseness
Clarity and detail are key, but avoid swamping AI with too much information. Aim for balance: enough context to set it up, but not so much it gets overwhelming.4.
4. Context
Adding background information helps (a lot), especially with specialised topics or industries. This can make the difference between a generic response and one that feels spot-on.
5. Iterative refinement
The first prompt might not always hit the mark — it’s all about trial and error to find the best way to ask what you’re looking for. Tweak and refine based on initial outputs to guide the AI closer to the ideal response. Soon enough, you’ll have created the right prompt for your task.
How to write prompts that work
When it comes to crafting prompts, knowing the type of answer you need is half the battle. Start with scoping out your request. If you simply need help drafting an email, this might not be as lengthy a process — but say you’re crafting an outline for a client’s landing page; this might require a bit more thought.
Here are a few guidelines to get you started:
1. Direct questions for factual information
Want factual answers? Stick with direct questions. For instance, “What are the main components of machine learning?” is clearer than “Tell me about machine learning.”
2. Step-by-step instructions for complex tasks
For multi-step tasks, break them down so AI can follow more accurately. If you’re coding, for example, try: “Write Python code that calculates the factorial of a number using recursion. Explain each line.”
3. Role-playing for creative responses
Set up scenarios if you want creative responses in your content. Something like, “Imagine you’re a 19th-century detective. How would you investigate a missing person’s case?” can produce some fun, detailed answers.
4. Examples and formats
If you need specific formatting, say so. For example, “Summarise the following article in bullet points,” or, “Provide a JSON output with name, age, and occupation.”
5. Iteration-based queries
Refining a response? Request changes based on what you’ve already been given. This iterative approach lets you get closer to your target with each tweak.
6. Use the scale method
Give AI an example that sounds/looks just how you want it to. Then, ask it to compare your draft with that example and rate it from 1 to 10 on things like how relatable, clear, and natural it sounds. From there, you can ask for specific edits based on the score.
For instance, say the output you received is a 6/10 in tone compared to your example. You could prompt: “Rewrite this to get closer to the example tone — make it a 10/10 by making it more conversational and relatable.
Examples of prompts across fields
Our team uses AI to scale content production processes for many of our clients. For short-form content like category texts, we use prompting to craft brand-aligned content at scale, to increase visibility and provide ranking opportunities for our clients. Take the e-commerce industry, for example:
E-commerce industry
When crafting content for an e-commerce client’s website, it’s important to ensure keyword visibility and alignment with the brand identity.
Example: Your task is to create category texts for Client X, an online e-commerce marketplace that sells toys. I will provide you with all the relevant keywords you should implement throughout the category text. The texts should be between 300 and 500 words, and they should be informative, helpful, interesting, to-the-point. Incorporate the brandʼs tone of voice: professional, playful, fun, witty, informative, helpful, smart, and knowledgeable. The category texts should be in Swedish and provided in HTML format.
When long-form content is the task, we instead use prompting to craft outlines and scale our process that way. This is often used when we write blog articles, for example, where we can both scale the outline & part of the research process. Take a look at this example prompt for a client within the finance industry:
The finance industry
As this is a YMYL industry, creating helpful, factual, trustworthy content is key. Prompting AI to create an outline based on real research is imperative, and should more information be required, asking for sources to be cited is a great way to ensure real, credible information is included.
Example: Your task is to write an outline for an article for Client Y, a bank specifically for small business owners who have just started/are considering starting a company. The article is entitled: Starting a Company. The goal of the article is to inform readers about tips on starting a company, local resources, and FAQs and should be between 1,000 and 1,500 words in total. Attached is another article for the same company I want you to analyse and use as inspiration for structure, tone of voice, and layout for your outline. Below are the main and secondary keywords you should optimise the outline for in various headings, as well as the research I’ve conducted on the topic to base your outline around. If you conduct any additional research on the topic, ensure you cite your sources in your outline.
Disclaimer: It’s always important to look through AI-generated content and fact-check output. As the actual (human) experts, it’s our job to ensure accuracy and quality in our work. Think of AI output as a base to work from.
Enhance AI prompts with visual context
Using screenshots of a category or product page together with a well-developed prompt when creating category or product descriptions is an effective way to give AI a deeper understanding of the context and target audience.
By providing a visual representation of the actual category assortment, you help AI analyse the types of products included, their features, and the potential needs of the customer. This approach enables the creation of more specific and customer-tailored texts that capture the right tone while highlighting relevant benefits.
In this way, the content you’re creating is not only informative but also engaging and optimised for the user.
Prompting in action
Within the SEO team at Precis, we’ve employed our prompting techniques at scale across a few of our clients – and seen impressive results. The combination of AI-based content production (for efficiency) with human-backed intervention (for accuracy and brand relevance) enables us to produce valuable content for our clients, bringing concrete results at scale.
