My thoughts on AI as a copywriter
Honestly, I don’t hate AI. I’ve worked with enough tech companies and SaaS startups to know that technology progresses at a terrifyingly fast rate. The rise of AI (which, by the way, has been around since the 1950s) will only accelerate this pace of change.
AI is, as far as I’m concerned, here to stay. So, we may as well embrace it.
With all that said, I understand that AI isn’t without its flaws. Language Learning Models (LLMs) have bias baked into them. LLMs are taught on human intelligence. Their constructs are informed by human behaviour and act. Humans are flawed. So, it’s only fair to expect AI and LLMs to be flawed too.
Those flaws also mean you can’t take AI at face value. I thought I was a people pleaser but my people pleasing tendencies have nothing on AI. It is the biggest people pleaser going. It exists to prove you right. And, when left unchecked, that can be dangerous. Pair that people pleasing with AI’s propensity to hallucinate and you’ll find yourself on shaky ground. For that reason, I don’t believe in using AI without human oversight. And I also don’t believe in using AI to deliver the final output of any work I produce.
And yet, AI is a tool. It has incredible capabilities that we can draw upon to improve efficiencies, gain greater depth of data, and expand the work we’re already doing.
When used correctly, AI can be a powerful tool to have in your back pocket as a copywriter. But, in order to do that, you need to understand how to strategically use AI. You need to understand how to prompt it, how to hold feedback loops, how to develop its capabilities, and where to draw the line.
How I use AI as a copywriter
I’ve made it clear that I’m not anti-AI. I recognise its potential in the world of copywriting. But I know not everyone is comfortable with AI so I wanted to transparently discuss how I use AI in my business.
Research
I use AI in the research phase of copywriting projects. Primarily, I use my proprietary Audience Voice Architect Custom GPT to codify large sets of qualitative data from user research I conduct. This is an AI agent that I have trained on my exact process for user research. Hosted in Chat-GPT, any conversations I have with this custom GPT are secure, meaning they cannot be crawled, read, or used by ChatGPT. I also delete any files uploaded to ensure security.
For some projects, I use Perplexity and Chat GPT to research topics and try find resources to help me better understand or explore a topic I am writing about. I will then personally research anything I uncover through these AI tools. If they reference any specific resources, I will personally track down the original source and read it to ensure the tool isn’t hallucinating or presenting inaccurate or unnecessary information.
Writing
AI can be a helpful tool in the writing process. While I won’t ever use AI to curate finalised content, I do sometimes use it as a writing partner to challenge my own writing. This involves feeding my copy into Chat GPT, along with additional context about the writing, and asking it to provide feedback or to give alternative suggestions.
One of my favourite exercises to do at the start of a new copywriting project is to come up with 100+ headlines. This used to be an exercise I’d solely run by myself. These days, I’ll write the first 20-50 headlines. Then, I will feed those headlines and my prized collection of headline formulas into Chat GPT and ask it to come up with more variations. Sometimes, AI misses the mark completely. Sometimes, the suggested headlines make me cringe beyond belief. But sometimes, it produces the thread of an idea and gives me a new angle to explore. I then put my own pen back to paper (okay fine, fingertips to keyboard) and continue writing new headlines and exploring other copywriting angles for the project.
This co-creation with AI is helpful during early “spit draft” stages of writing where I’m exploring various ideas and writing styles before committing to a project. It’s essentially like having someone to bounce ideas off.
Business
I’m a solopreneur wearing 83 hats. I spin every single plate in my business. And I’m also a high-achieving people-pleaser. It’s a combination that means I operate on the cusp of burnout far more than I’d like to admit.
It also means operating in silos and trying to figure out business challenges on my own. I will occasionally use Chat GPT to co-solve a challenge I’m facing. It can be great for reviewing my business systems and discovering efficiencies and improvements that can be made. It can also be great for gaining an unbiased perspective on a situation where I worry I may be operating from a place of emotion, not business.
I will occasionally use AI to review my business offerings and processes to explore ways to improve my services.
AI tools I use as a copywriter
I like to keep my tool stack lean. There are two AI tools that I use within my business: ChatGPT plus and Perplexity.
I used to use Grammarly to proofread my content but I fell out of love with it a few years ago.
Right now, I don’t have plans to add any more AI tools to my tool stack. It isn’t something I actively search for. I use ChatGPT and Perplexity as they offer enough flexibility to let me lead the process. Basically, I use these tools because I know they can support my research, writing, and business operations without detracting from the humanness of what I do.
On the business side, I also like to limit my AI use mainly from a cybersecurity perspective (more on that later). This means I don’t use any AI-led tools to manage my emails or calendar… although, admittedly, the promise of a fully-managed diary does feel tempting. The cybersecurity risk, for me, outweighs the temptation to use such tools.
I also do not use any automatic AI-led meeting recorders. I only record meetings if I have permission to do so from the other meeting attendees. In this case, I will use TL;DV to record, transcribe, and summarise meetings. As Google Meet improves it’s note-taking abilities, I’ll likely transition to this tool instead.
A note on AI and cybersecurity
Whenever you use a LLM model like Chat GPT, Claude, or Gemini, it keeps hold of all the information and data you feed it. Then, it uses that information to improve future answers it gives you.
But these platforms can also use data you’ve inputted to further train their models. Everything you tell the LLM can, ultimately, be fed back into the system and used for answers it provides other users.
This use of data for training poses a severe cybersecurity risk. Couple that with the idea that someone could hack into your account, read all your conversations, and extract any information, files, or documents you’ve shared with the LLM and you’ve got a serious problem on your hands.
For that reason, I’m extremely careful about the type of information I feed into the AI tools that I use.
I run by the rule of asking myself “If this AI interaction got into the wrong hands, how would I feel?”. It’s a simple way of sense checking what I’m sharing before I share anything.
I do not input any sensitive information into any AI tool that I use. I also make sure that the AI tool’s training feature is turned off so that it cannot use my inputted data to train its model for other users.
In Chat GPT, I will delete conversations as soon as they are no longer active or required. On top of that, I also routinely check the saved memories to make sure it hasn’t stored any information I don’t want it to.
I also do not connect AI tools with other tools to ensure data can not be shared between programmes or apps.
I write enough cybersecurity content to know that the cyber threat of AI is not one to be sniffed at. So, I take the security implications seriously and do my best to keep any risk at a minimum to both my business and my clients businesses.
My public thoughts on AI
Like every other copywriter out there, I’ve done my fair share of AI musing out in public.
In 2023, I write an article for Moz on How to make AI your writing sidekick for content marketing. This article was updated in March 2024. This blog article explores some simple ways you can weave AI into your content writing. It’s a helpful content piece for anyone who is an early-stage content marketer or a generalist marketer wearing many hats.
That article was featured by John Mueller in his Google Search News YouTube series in October 2023.