We’re at a turning point.
University marketers are beginning to take generative AI writing tools seriously. But what happens next?
The quality of HE content could suffer until every institution sounds the same. Or AI could bring new efficiencies that free up time for marketers to be more creative than ever.
I think we’d all prefer option two.
Here’s how we see it happening.
“In the rapidly evolving landscape of higher education, universities are constantly seeking innovative ways to engage with prospective students, faculty, and stakeholders. One such innovation is the use of generative AI for copywriting. This technology is transforming how universities create content, making the process more efficient, personalized, and impactful.” Copilot AI's mediocre attempt at an intro for this blog - not very timely, not at all provocative, not particularly easy to read. Which version do you prefer?
Step 1: policies and planning
Universities are waking up to the fact that generative AI is here to stay. They’re developing responsible use policies. Consulting with staff. Trialling different platforms. It’s still early in the journey, but gen AI is becoming a standardised tool in the HE marketer’s toolkit.
This is the time to act before it’s too late.
Once staff start depending on ChatGPT, they’ll use it for everything. Marketing plans, course descriptions, even sick notes to HR.
Soon, people will lose the ability to think for themselves. Even if they change a few words here and there, content will be overly formulaic and logical. It will lose its human touch.
The earlier you can define some rules, the better. Here are a few suggestions:
Think and plan what you need before you turn to AI – otherwise you’ll accept its suggestions without understanding what your goals are
Be as specific as possible with your prompts – AI can do a lot but it interprets requests a bit too literally
Only use AI for tasks you are already familiar with – if you don’t know what makes a good fundraising letter, for example, how will you judge the quality if you ask ChatGPT to make one for you?
Always fact-check any claim made by AI – yes, AI can lie
Be aware of in-built bias and try to remove it – it’s easy to miss if you’re not actively looking for it
Be aware of wider ethical and environmental issues – AI uses a huge amount of power, for example, so use it wisely
Step 2: training and investment
A policy is a great starting point. But turning that into a reality will be the real challenge. Easier said than done and all that.
People are lazy. You have to assume that even the most conscientious worker in your team will choose what seems like the simplest route to achieving a goal. And these days, the simplest route often seems like getting AI to do it for you.
Prompt writing
The prompt is everything. If you asked a human writer to write something in the style of Shakespeare, they’d know what you mean. But AI is more likely to just use some Elizabethan-sounding words like “thee” and give you something sounding like an amateur at a reenactment. You need to give it more detail – like write in “iambic pentameter” or “in the style of his tragedies” or “using authentic Elizabethan English”.
It won’t come close to the Bard himself but at least you’ve given it a chance.
Pros and cons of various tools
Some people are technology nerds and want to learn everything they can about what’s out there. But most people aren’t.
If you want the best results from gen AI, you need to keep on top of developments and make sure your whole team knows what tools they should be using or avoiding. And be specific – explain the nuances and the suitability of different tools for different tasks.
How to evaluate outputs
Anyone can get AI to write a thousand-word blog post in a few seconds. There’s no skill involved. And it’s easy to be impressed by the technology and gloss over the quality.
But just because you can do something, doesn’t mean you should. Your content creators shouldn’t be using AI to write anything if they don’t have the skills to evaluate what they’ve produced.
Is it engaging? Accessible? Does it use a persuasive structure? Or any established rhetorical devices? There’s a lot more to good quality writing than just whether it makes sense or not.
Step 3: moderating and refining
So, you’ve created some policies and upskilled your staff. Will your AI-produced content be good enough to use in your marketing?
Absolutely not.
The output on the screen should be thought of as source material or inspiration – not something you use right away.
It’s lazy at best, stealing at worst
Let’s say you want a blog post about the importance of trees on campus. You found something you like on another university’s website. Should you just copy and paste and put it on your own blog? Of course not.
You ask AI to write it for you instead. But little did you know, the model was trained on that original blog post and just regurgitated something similar.
Facts aren’t what they used to be
Generative AI is known to make things up. Sometimes because it’s trained on unreliable content to begin with, and sometimes there’s no reason at all.
Don’t just put AI content out there without checking it thoroughly first.
It’s got some way to go still
A lot of content on the internet is rubbish. So that means a lot of what generative AI writes is rubbish too – it’s just doing what it knows.
If you’re aware of its bad writing habits, you can undo them and turn mediocre content into something worth using.
Repetition
Repetition can be a powerful device if it’s used carefully, but AI content is often unnecessarily repetitive.
Little variation in sentence length
Human speech is varied and dynamic. We don’t speak in perfectly average sentence lengths. We mix things up. AI doesn’t seem capable of understanding that just yet.
Predictable structure
In school, you might have learnt the Point Evidence Explain structure. You probably wrote every essay in the same format. AI is a bit like that – it knows a basic way of organising information and sticks to it.
Gerund phrases at the end of a sentence
Gerunds are words ending with -ing. Content generated by AI tends to overuse these, ensuring that writing sounds like a high-school essay.
Telltale words
Maybe all gen AI models were simply trained on Forbes articles. Who knows. Watch out for those AI-sounding words like “transforming”, “unlock”, “landscape”, “significant”, “ensuring”.
AI-generated example
Prompt
write a short introduction to a blog post about using generative AI within university marketing
Result with Claude
Generative AI is revolutionizing higher education marketing, offering universities unprecedented opportunities to personalize, streamline, and amplify their outreach strategies. From crafting targeted content to generating dynamic student communications, AI tools are transforming how institutions connect with prospective students, alumni, and stakeholders. This blog post explores practical, ethical approaches to integrating generative AI into university marketing workflows, revealing how innovative institutions are leveraging technology to enhance engagement and storytelling.
Our analysis
Three long sentences
The same point repeated each time (new ways to engage people)
Lots of AI buzzwords (revolutionizing, unprecedented, transforming, innovating, leveraging, enhance)
Gerund phrase
Boring
Ways to improve it
Work to the tool’s strengths – Claude allows you to upload content to adjust the tone and create a new custom style
Refine the prompt – think of the points you’d like to make and include those instead of asking for something generic
Rewrite in your own words – if you’re broadly happy with the content, start again but writing like you would if you’d written it from scratch
Step 4: keep your job
Generative AI isn’t good enough to start replacing skilled human writers yet. If all you need is low-quality bulk content, then go ahead and use it as much as you like. But if you still value human engagement and want your institution to represent quality, then it’s time to start thinking ahead to how you will use (and more importantly, limit) AI within your marketing team.
Comments