We have officially entered the Artificial Intelligence (AI) age. While we’ve all been using AI for a long time in products for many years without really knowing it (like Google Maps or Grammarly), it feels like it’s really officially here. The release of AI-powered Large Language Models like ChatGPT late last year has made it accessible for regular folks like you and me to use this new technology much more easily.

These are tools that can be used to do some things faster, help us brainstorm new ideas and be more efficient: but they can’t do everything. They are great tools to use in collaboration with humans: but not to replace them.

Here’s how I’ve found ChatGPT to be a useful tool in my daily life.

But first up: what is ChatGPT?

ChatGPT is a Large Language Model (LLM) by the company OpenAI – it’s one of many that are hitting the market right now. Google’s Bard is another popular one. I’ve spent the most time with ChatGPT, so that’s what I’ll be referring to most in this post.

ChatGPT is a type of AI that can understand and generate text. It’s not a supercomputer or a search engine, even though it sometimes feels like one!

It’s a language prediction model. ChatGPT learns from vast amounts of text to make predictions and generate text based on patterns it has seen before. It works in the same way that if I ask you to “fill in the blanks” for this sentence, “I’d like to order a coff_ _”. You can predict the next two letters in that sentence based on patterns you’ve seen before: that is what ChatGPT is doing. Except that ChatGPT has absorbed what feels like most of the text on the internet. So if you ask it to write an email to your vet to book an appointment, it will have read countless similar messages and can predict what it thinks should go in there. So ChatGPT seems like it “knows things”: but it’s just working with the patterns it’s absorbed.

It’s really useful to understand what Large Language Models are doing so we can give them tasks that it’s capable of helping with.

There’s a great TED talk (below) that discusses some of the flaws of AI where it simply doesn’t understand common sense. “Would I get a flat tyre if I’m riding my bike over a bridge that has broken glass underneath?” An LLM AI says “yes“: because it can’t experience the real world to understand the context that the glass is nowhere near the bicycle.

So it’s instrumental in understanding what LLMs can do – and what they can’t (yet).

How can we use ChatGPT to help us with marketing?

I’ve been experimenting with ChatGPT for a few months, and there are some things I’ve tried that have worked and some things that haven’t.

The things it can do is:

• Finding patterns in data: like if you want it to find the best-performing data in a report.
• Coding: if you need Excel to do something, it can write code for you.
• Summarising content: if you want it to summarise a report.
• An information resource: good for general questions, but it doesn’t have real-time internet access for up-to-date information (but read onwards about its “hallucinations”).
• It can help with generating blog post ideas, creating ad copy, or assisting with social media content (more on this below).

Things it can’t do:

• It can’t read websites and give you information about it (it’s drawing on what it’s been given historically)
• It can’t Google current information for you (although some plugins can change this, I’ve not found it to be all that helpful in its current form).
• It doesn’t always get humour – and let’s face it, a lot of marketing involves humour.

In terms of limitations, we’ve heard one of the big problems of ChatGPT is that it “hallucinates” or makes things up. This is where it can generate plausible-sounding but inaccurate or nonsensical information. This is often because if you’ve asked it a question where it doesn’t have an answer based on what it’s been trained on, it simply does its best to “complete the pattern”. In short, it’s doing what it’s designed to do: which is to predict what it thinks the next piece of information is. So, treat your LLM with a grain of salt if you’re doing research with it and cross-reference the results knowing why it behaves that way.

How I approach working with ChatGPT

I use ChatGPT as a tool to assist me in getting the job done – not to replace my work. The simple fact is that it isn’t capable of doing everything a human can do and doesn’t understand the nuance that goes into human communication in some contexts (and I feel that content marketing is a big area where its skills lack). ChatGPT requires human prompt crafting, human oversight, editing, and proofreading to ensure high-quality content.

I like to think of ChatGPT as an intern who is, perhaps, also an alien. It’s pretty good at writing a first draft, but you really need to spell out the context for it to do a good job – and then – you still need to edit it to make sure it’s correct and makes sense. But, despite this, it still can be a huge time saver.

So, you can ask ChatGPT to write a blog post, but it will just write something pretty generic unless you explain who the audience is, what the tone of voice is, the key points you want it to cover, and the objective of the piece. I also like to paste in examples of similar pieces of content to give it a better understanding of things like structure or tone of voice that I want it to emulate.

Essentially, you need to write a brief.

And, like any alien intern, I would always proofread it and revise any content through the lens of a human before sending it anywhere.

So working with ChatGPT is definitely not an act of “press a button and a perfect blog comes out”. Not. at. all.

It’s also worth flagging that ChatGPT also has a ‘memory’ limit, which means that if you have a really long conversation, it will forget what you said at the start. This means it may simply forget what you said in the initial brief. So just being aware of when it starts to get a bit sloppy with its responses, you may need to re-enter that original context.

Now, LLMs are very much in their infancy, and the things I write about today may be out of date in just a few months as it develops.

So, for now, here’s how I’ve been using ChatGPT to help me with my digital marketing.

How to use ChatGPT for social media marketing

Data analysis

One of the things ChatGPT is excellent at is finding patterns. I can paste in several months of social media data and ask it to highlight the posts with the highest engagement and any trends it sees.

It might find that there’s a pattern in posts about money: but it would take some human insight to connect the dots around this being due to, say, it being a cost-of-living crisis.

You can use ChatGPT to flag the most engaging content, content with the highest reach, and the best time of day to post. Essentially, it’s doing what you would manually do in Excel, just a bit faster.

This could also be used for your Google Analytics content, and insights from there could also give your content ideas for your social media.

Brainstorm buddy

I love to use ChatGPT as a brainstorming buddy. I can paste in information about a product, brand or project and then ask for ten different topics I could use as social media content.  I can choose the ones that make the most sense to create content about.

I can also ask it to iterate even further: of any one of the suggestions, to create more angles off that topic.

I also find it really useful if I’m writing pun-tastic social media content. While it’s not so great at coming up with funny jokes on its own, if you ask for hundred puns about stationary, it will generate a hundred responses, and a few of them will be good. So you can pick and choose the best ones and incorporate them on your own. Again – just a bit faster than a Google Search.

Summarise content

While ChatGPT claims it can create social media content from scratch, I’ve found it to be less than compelling for any of the work that I need it for.

ChatGPT doesn’t really write in the way that a human does: it’s missing a bit of heart, so I struggle to use it for a lot of my social media posts which are about human connection or are designed for engagement.

Further to this, if asked ChatGPT to read a blog post and create a social media post about it, it could do so, but it would be lacklustre because it can technically summarise the post – but it’s not capable of finding the hook an audience would find most compelling. This may be possible with more prompting and context, but it’s not something I’ve been able to master yet.

But if I’m doing a social media post where I’m summarising the key features of a product to sell, I could paste in the information from a sales page and ask it to create Instagram copy about that product. It’s excellent for this specific task because it’s great at summarising content.

In this way, it can also be useful for ad copy or metadata descriptions if the goal is to succinctly summarise what you’re selling.

So – it can create social media content, but I find it only does a good job if that content is essentially the summary of another piece of content. Otherwise, it’s something that a human is still a bit stronger for.

A hole-spotter

Because ChatGPT has so many patterns at its fingertips, it can help you spot gaps because it can see what’s missing from the pattern. For this reason, it’s great to help you spot flaws in a project.

Let’s say you’ve written a blog about a topic, and you want to see if there’s any information you’ve missed to make the topic clearer. You can ask ChatGPT to read the posts and, based on x audience understanding y, what is missing? It will give you a list of what you can improve.

You could even use this for market research. For example, if you’ve created a sales page, you could then ask it to “imagine it is a specific audience” and ask it what information is missing that might encourage people to buy.

It’s brilliant for helping highlight things your tired brain may have missed.

Challenges with AI and LLMs

There’s a lot of hot debate about AI usage at the moment. But I’m doing my best to be a techno-optimist. I see this as a tool, similar to the printing press or the internet: it is a technology that will create a seismic change in our world, and it’s not going away. Regulation isn’t here yet, and the companies developing it are run by humans who are, like all of us, flawed. So things aren’t perfect. But humanity has experienced this many, many times before: a thing is invented, and then regulation follows.

There are problems with some companies not gaining consent to train some LLMs, especially for image-based tools, meaning AI is essentially copying work at superhuman speed. But we’re also seeing more companies emerge who are using image sources that are using a library of images where consent has been obtained. As consumers, we can choose to work with companies doing their best to make ethical decisions.

We are likely to see jobs shift and change. But I think it’s too early to tell how big this change will be. For example, the automation of reconciliation in accounting software has saved a lot of manual labour for small business owners. But accountants aren’t out of a job and still exist to help advise us at tax time. I would like to hope that these tools help us automate laborious tasks so we can spend our energy in better, more fulfilling ways.

There’s also a discussion on how – and if – we credit the use of AI in content creation. For this blog post, I wrote it myself, but I did ask ChatGPT to help me explain LLMs, review this article for accuracy on how I explained ChatGPT, and then used Grammarly to check my spelling. We may find that attribution around AI becomes the norm – but for now, we’re inventing the culture and the future use of these tools. So let’s do it consciously and wisely.

Conclusion

We’re in the early days of LLMs and tools like ChatGPT can do some amazing things – but it is also limited in a lot of ways, too. In the end, it’s about us learning how this technology works, where it can be useful for us and the ethical implementation of that technology. It’s also useful to see where out humanness is more important and more powerful than AI.

For me, as a small business owner, having ChatGPT some days it’s a lifesaver. Other days, it’s a source of great frustration. It’s not perfect. It’s not a magic wand. It’s complex tool, and I’m doing my best to learn how to use it with the hope that it’s a tool that can help shape the world for the better.

---

Want to work with Rachel?

Rachel Beaney is a writer and social media content specialist, helping businesses connect with their audiences.

She’s worked with local, national and global companies, in addition to not-for-profits and government bodies. She loves helping businesses tell their stories with creative and data-driven solutions.

She is based in Sydney, Australia.

Want to work together? Rachel would love to hear from you. Get in touch today.