Using AI for content generation? Here's how to do it right

15th May 2024

Artificial intelligence has the potential to change content production forever. From graphic design and video animation, through to voiceovers and potentially even acting in the future, AI is going to be used by an increasing number of organisations to create content.

There are very important conversations to be had about the ethical and social impact of AI, particularly when it comes to jobs, but for now, at least, the technology hasn’t yet evolved to be capable of replacing copywriters, graphic designers and actors.

However, that’s not to say that businesses aren’t tentatively using AI to create content already. The output of these AI tools can vary in quality, but there are some best practices you can follow to ensure you’re creating better content.

READ MORE: What can we expect from AI in 2024?

Best practices for AI copywriting

OpenAI’s ChatGPT was one of the first tools to take the world by storm, and v3.5 is free to use with an account (GPT-4, which also includes AI image generator DALL-E, is available for a monthly subscription). Since the launch of ChatGPT in 2022, various competitors have emerged, such as Google’s Gemini (formerly Bard) and Copy.ai, but they all largely work in the same way—by combining machine learning with huge databases of existing content.

More than 18 months on from the launch of ChatGPT and the quality of written output from these tools has improved significantly, but this depends a lot on how you craft your prompts.  Here are some tips:

1. Be clear, concise and direct
2. Provide context
3. Give one instruction at a time
4. Provide examples
5. Define the required content format and tone of voice
6. Check (and factcheck) the work thoroughly

1. Be clear, concise and direct

Despite AI’s best efforts to make it appear like you’re speaking to a human, you’re not. 

Don’t write your prompts like you’re chatting to a colleague on Slack. Avoid colloquialisms, slang and any other conversational language that might not translate well. Note that writing a concise prompt doesn’t mean a vague prompt—make sure you include as much specific detail about what you’re looking for as possible.

2. Provide context

If you’re asking for a blog post from a human copywriter you’d provide a brief that includes lots of background information to help them, and you should be doing the same for AI. Take a look at the example prompt below:

“How can I improve my fitness?”

Without context about your current age, weight and fitness level this is a very hard question to answer. People who are severely unfit might need to begin with going for frequent walks or low resistance weight training, while advice for someone who’s already relatively healthy would be completely different.

When crafting a prompt for an AI writing tool, add background information so it can better ‘understand’ the request. Applying this to the example above:

“I’m a 5’10” tall, 35-year male weighing 160 pounds, and I currently run 5k once a week. How can I improve my fitness?”

3. Give one instruction at a time

Multiple requests in a single prompt can confuse an AI tool, so allow it to focus on doing one thing at a time, otherwise the varying contexts of the different requests can get muddled.

4. Provide examples

Do you want AI to write some copy in the tone of voice of your business? Share an example of that copy with the AI tool and ask it to replicate the tone of voice used in that content.

5. Define the required content format and tone of voice

Consider the below example output for the prompt, ‘Write a 150 word paragraph explaining the benefits of Digital Asset Management’:

“Digital Asset Management (DAM) offers numerous benefits for organizations managing large volumes of digital content. First, DAM enhances efficiency by providing a centralized repository for storing, organizing, and retrieving digital assets. This streamlines workflows, reduces duplication of efforts, and improves collaboration among teams. Second, DAM ensures consistency and brand compliance by enforcing standardized metadata and usage rights across all assets. This helps maintain brand integrity and simplifies content distribution across various platforms. Third, DAM enhances security and control with robust access controls and versioning capabilities, ensuring that sensitive assets are protected and only accessible to authorized users. Additionally, DAM improves productivity by automating repetitive tasks such as file conversions, approvals, and distribution, freeing up time for more strategic activities. Finally, DAM facilitates analytics and reporting, providing insights into asset performance and usage patterns, which can inform future content strategies and optimize ROI. Overall, DAM empowers organizations to manage their digital assets more effectively, ultimately driving business growth and success.”

Compare this to what we get when we add this to the prompt: ‘This copy is intended for people who do not understand Digital Asset Management, so use simple language. The tone of voice is friendly and approachable.’

“Digital Asset Management (DAM) offers valuable benefits for anyone dealing with lots of digital files. First, it helps keep everything organized in one place, making it easy to find what you need quickly. No more digging through folders or searching endlessly! Second, DAM ensures that everyone is using the correct and up-to-date versions of files, which saves time and avoids confusion. Third, it enhances collaboration by allowing multiple people to access and work on files simultaneously, even if they're in different locations. This boosts teamwork and productivity. Fourth, DAM improves security by controlling who can access certain files and tracking changes made to them. Finally, it simplifies sharing files with others, whether it's within your team or with external partners, making the process smooth and efficient. Overall, DAM makes managing digital assets a breeze, helping you work smarter and more effectively with your files.”

The content is explaining the same thing, but the tone is completely different.

6. Check (and factcheck) the work thoroughly

Finally, although AI copy is improving all the time, it’s still a long way from being good-to-go straight away. You need to proofread AI content just like you would from a human writer, while you also need to thoroughly factcheck it. AI makes things up a lot, and there’s even a term for it: hallucinations.

There are a number of high profile examples of this, not least an error included in promotional content for Google’s Bard, so you need to ensure that your AI content doesn’t make the same mistake.

When checking over AI content, if there’s a claim, fact or stat you don’t know yourself to be correct, check it.

Best practices for AI image generation

Just like AI copywriting tools, generative AI for artwork and design has improved over the last 18 months. However, when it comes to imagery specifically, the progress has been remarkable.

Midjourney v1 (Jan 2022) to v5 (May 2023), Source: Reddit.com

However, to get high quality images produced by AI you still need to follow best practices to get your prompts right.

1. Use the best tool for the job
2. Use it as a starting point
3. Define your art style and vision
4. Continually refine and iterate
5. Touch up any mistakes or weird elements

1. Use the best tool for the job

There’s not a huge amount that separates the major AI copywriting tools, but that’s not the case for AI image generators. 

DALL-E 3 is considered by many to be the best, but it can be limited for more advanced use. On the other hand, Midjourney offers extensive customisation and detailed output, but it has a steep learning curve and complicated setup so won’t be suitable for everyone.

Make sure you’re using the best tool for your specific requirement. You can find out more about the various AI image generation tools here.

2. Use it as a starting point

One of the most powerful applications for AI tools (including copywriting) is using it to get you started. Suffering from an artistic block? Enter a prompt into an AI image generator and see what it creates. It might just inspire you to take a particular direction, even if you don’t use what the AI created at all.

3. Define your art style and vision

Just like when using it for copywriting, AI is at its best when given clearly defined prompts.

Start by outlining your vision for the image, including the desired art style. For example, ‘sunset over a city skyline’ will define what you’re trying to create, while ‘sunset over a city skyline in a noir visual style’ will create something much more specific.

4. Continually refine and iterate

An AI image generator will rarely nail your request the first time, but you can continually refine that initial starting point by tweaking the words used in the prompt, adding more context or iterating the colours used in certain elements—there are lots of options.

5. Touch up any mistakes or weird elements

Although you can refine AI images within the tool you create them in, you might still need some human intervention to fix mistakes or weird elements. For example, for a long time AI struggled to render realistic human hands.

You might also need to upscale the image, as they can often be low-res initially. This can be achieved in a third-party app like Photoshop if the AI tool itself doesn’t include that feature.

At ResourceSpace we’ve been making use of AI to improve our DAM system for a number of years. Want to find out more? Book your free demo below.