The benefits of generative AI in copywriting

Copywriting is central to what we do as a growth and digital marketing agency. We use it for new websites, refreshing websites, landing pages, social media, blogs, articles, rich content, PR, brochures, advertising and just about everything else we do. And for us, AI in copywriting is vital.

Generative AI, or generative artificial intelligence, is a type of AI that can create new content like text, images, videos, and music. Our writers all use generative AI regularly, but many people we deal with have never had it explained to them – they have just learned ‘on the job’. So this article may be particularly relevant to understanding AI vs human copywriting, showing how to use AI in marketing as well as demonstrating the benefits of AI in content writing: what it can bring and what it could do for you, not forgetting the AI writing limitations.

We all use AI daily

In a recent CIM (Chartered Institute of Marketing) poll 30% of marketing professionals use AI daily and 27% at least three times a week, so it is clearly a massive factor in marketing.

The best known and used generative AI tools are ChatGPT, Claude, Gemini, CoPilot, Deepseek, Mistral and Poe. They all work slightly differently and will provide alternative answers, but essentially operate in the same way: they use generative models that learn patterns in data and then create new data with similar characteristics. Or in other words, they look for patterns in the data, then evaluate the patterns and make a prediction. Simple really.

How does generative AI work?

Generative AI is set up to work a bit like an app – a computer programme that’s always at your side, which you can instruct to produce the answers you are looking for. For most people it uses contextual prompts within a single conversation – though it goes further than this for advanced users who can train it so it gets to know you and your preferences and can automatically respond to you in the way you want. Of course this can also become a problem if you need variety, so you cannot be complacent and let it rule you!

AI can also be used non-generatively, for instance for search – Perplexity is one of the market leaders in this (if you discount the traditional search engines such as google, bing, yahoo, etc, all of which use AI to a growing extent). It is sometimes helpful for search because you only get one response, and you can then ask it follow up questions to deepen your knowledge – whereas a search engine will give you many options to trawl through and you have to continue to refine your search until you get the information you want.

How to use ‘prompts’ in generative AI

Key to the successful use of generative AI are the prompts you give it, and there are several structured methods, or models, which are largely similar and enable users to ensure that their prompts are producing the right answers. One is ‘GCSE’.

G Goal                  Clearly state what you want the model to generate.

C Context            Provide context and set the scene. Who are the target audience? What is the nature of the project? Etc.

S Sources             Give the model examples of what you would like to see.

E Expectations    Set your expectations so the model knows what to provide, eg bullet points, a table, a short paragraph, tone of voice, etc.

Another structured method is TACO:

T Task                   What do you want the AI to do?

Example: “Write a blog post about AI in copywriting.”

A Audience         Who is the content for?

Example: “Write for marketing professionals who are new to AI.”

C Context            Provide background details to guide the AI.

Example: “Mention tools like ChatGPT and Gemini, and focus on their benefits in content creation.”

O Output             Specify the format, tone, and structure.

Example: “Write in a professional but friendly tone. Use bullet points for key takeaways.”

These are easy to remember and quick to implement; employ them in your generative AI search and you should get what you want, first time.

Hints and tips on using AI successfully

Another thing you can do is instruct the model – before the output – to ask me questions to further understand any gaps. Its works really well!

It’s also important to ensure you let the generative AI know if you are looking for short or longform content. It’ll do whatever you want, provided you are specific enough for it to know what to do.

AI can also be helpful in sparking ideas, brainstorming and research. We have all tried setting up AI, only to then use the generated content as a springboard to more creative thinking, because what it produces can be a bit flat, or bland. Although in reality this has more to do with the prompt and the user than the AI engine. With a good prompt, AI can turn around ’boring content’ into creative pieces. Equally it can take you down routes that you might not have otherwise considered, so it can also be useful in this way when researching.

It is also helpful for summarising an item, so you can decide whether to read something in full. If you have to read 10 lengthy research documents or white papers, which will take you more hours than you have available, ask AI for summaries, then you can decide which ones to read in full.

‘NotebookLM’ is helpful in this also. This is an AI-powered collaborator that ‘helps you do your best thinking’. After uploading your documents, NotebookLM becomes an instant expert in those sources so you can read, take notes, and collaborate with it to refine and organize your ideas. It’s latest function is an automatically generated podcast that is based on the selected content. It’s game-changing for learning about a topic!

Where generative AI goes wrong

But always beware of AI ‘hallucinating’ – which is where it incorrectly interprets your input data and produces very strange and unexpected results – so you have to be on your guard from being completely mislead!

As you may have heard on the news or have had experience of, AI can cause problems, either by hallucinating or by not reflecting your brief, or just being bland. Therefore we at Marcom ensure that we partly rely on our own input, both before and after use. Our own ‘human intelligence’ is essential.

We use what is sometimes known as the ‘AI sandwich’, which puts AI in between two slices of our own native human intelligence:

Human intelligence          Objectives, strategy, creative, research, GCSE/TACO prompts

AI                                         Generation of content

Human Intelligence          Crafting, sense and fact-checking (for hallucinations etc).

Points to remember

The final point to consider is that you can ask for improvements. Don’t feel tied to what you are given. AI is a machine which will respond to whatever you tell it. We tend to get the best results when we ask it for further ideas, or different styles.

Equally, do not be overwhelmed by the WOW factor. Many users are so impressed and excited by what they get back, that they do not give it sufficiently objective and critical assessment. Remember, in business people will judge you on the content, not the AI.

As an example, we produced a Marcom Christmas card several years ago when AI was brand new. We asked for funny and wacky and that’s what we got – though frankly the jokes were pretty lame. Several attempts later, and after a good deal of brainstorming jokes amongst our colleagues, we had the funniest and cleverest Christmas card ever. In this case, generative AI was more like a colleague than a machine.

I hope this has been helpful. It’s just the briefest overview, but AI will never go away and will become more and more ubiquitous so a clear understanding of how it works, and how it should be used, is essential.

My thanks to Kerry Harrison, copywriter and CIM Course Director, for some of the learning in this article.