Few emerging technologies are as polarizing as AI. ChatGPT3 kicked off a hype wave that swept us all in its wake – with froth about its world-shaking potential matched only by panic over the robots taking our jobs. Of course, as the noise quietens, it has become clear that we’re certainly not there yet (and very likely never will be). The robots just aren’t that smart. But that doesn’t mean they aren’t useful.
The first lesson learned was that large language models, such as ChatGPT, certainly aren’t what we think of as intelligent. They are more like highly advanced predictive text: very good at stringing words together in a way that sounds plausible, but without any real understanding of the subject matter. So the obvious use case is, well, spam. It doesn’t have to be meaningful, just to fill pages and sell ads. But what if you want to fill your pages with content that actually has real value? A raft of AI tools developed specifically for copywriting have been launched, promising to “reduce cognitive mental labor” while generating useful content. That may be possible, if the writer is smart about how they use them. Because the pitfalls are many.
What could possibly go wrong?
The most obvious problem, of course, is that AI-generated content is highly unreliable. LLM notoriously suffer from hallucinations: they don’t actually “know” anything, but they are very good at making things up. Unfortunately, the better the quality of the copy, the harder it is to fillet out these fictions. The same goes for sources: chatbots have been known to confidently cite non-existent works from real authors and publications. It can be very difficult to determine whether such sources are reliable; trusting the bot’s own assessment of fact certainly isn’t a solution, as at least one lawyer has already found out.
On top of that, it’s just not that good yet. For the most part, bot-generated content reads like bot-generated content: clunky and repetitive, with no insight or originality. This kind of mass-produced verbiage may serve to fill up content farms that aren’t intended to be read by humans, but we assume your goals are set a little higher. If you want your content to attract real readers and build your brand, you’ll need to add some real human insight and analysis.
AI’s lack of understanding is another problem – as evidenced by these attempts to get language models to describe a photo of the world’s only unspotted giraffe. The bots simply aren’t equipped to know what matters. They won’t know what is important about your product unless you specifically tell them, they won’t understand the problem your offering addresses, and they certainly won’t know what distinguishes you from competitors. If you’ve been through THE RELEVANCE HOUSE brand development process, you know how complex this is: a strong brand is rooted in psychological and market analysis as much as specific product benefits, and that is well beyond the scope of what a bot can absorb.
All this shows how careless use of AI could expose your company to risks ranging from the legal – the company may be held liable for made-up product claims, or for plagiarism – to reputational. With content marketing, the goal is to build credibility, but bot-generated copy could actually lower trust. Clearly, if you use AI to write an article, you should be prepared to spend a chunk of time fact-checking, rewriting and adding nuance. So where’s the promised productivity boost?
Help for the hard parts (and the boring bits)
The best ways to use AI come at the earliest and latest stages of content generation – first, getting past the blank page, and finally, optimizing and leveraging what you’ve written. Here are some practical ways to turn that bot into your new best friend.
- Roll the dice: With only a broad topic in mind, you can generate a list of possible article ideas; then once you’ve picked your topic, ask AI to propose an outline. It may well not be a good outline, but it will give you something to start with: a list of points to cover and a workable structure.
If you’ve ever tried to make a decision by tossing a coin, on the famous principle that when it’s in the air you’ll know which way you want it to land, you will see the value of having an outline you don’t like. It’s easier to shuffle the furniture around than to choose all the elements from scratch.
- Proceed with caution: Remember, even at the outline stage, you can’t rely on AI for research. Put your own thought into whether every angle has been covered. The bot may well come up with points you hadn’t considered, but miss some essentials. If you’re including any auto-generated material, be sure to check carefully for dodgy facts (any statistics, in particular, should be traced to a reputable source) and confident fictions.
- Dress it up: It takes only seconds to get a raft of headline suggestions from a chatbot. Don’t expect brilliance (despite the hallucinations, creativity is not AI’s strong suit), but there’s a good chance you’ll find at least one good idea that you can work with. You can try the same trick to find hashtags, email subject lines, calls to action and more.
- Take a second look: Where AI really shines is in generating copy fast – so leverage that to get alternative versions of the text you’ve created. This is very helpful in experimenting with tone of voice, for instance, or repurposing copy for a different audience. You can even use it to help you segment your audience and personalize the copy accordingly. Of course it is also ideal for A/B testing.
- Bring out your dead: Along the same lines, try using AI to create new content based on old copy. That might mean updating an old blog post, or using it to seed a series of posts to fill your social media calendar.
- Optimize, optimize, optimize: SEO is governed by algorithms, so it stands to reason that you should use algorithms to work it! Chatbots can help with everything from finding the best keywords for your topic, to working them into the text – plus more of that A/B testing. Contrary to fears, Google won’t penalize AI content, but the EEAT rules apply: you need to demonstrate expertise, experience, authoritativeness and trustworthiness. As we’ve seen, pure bot-generated content is unlikely to pass muster, but bot-assisted writing certainly can.
At THE RELEVANCE HOUSE, we’re not scared of AI. New technologies are our playground! But we know how much depends on how you use it. To get the most out of chatbots, you need to be very specific – feed in your exact requirements for factual details, audience, context and tone of voice – and very cautious. Don’t believe anything it says, and treat the results as an unseasoned dish, to be spiced up with a human voice. (In case you’ve been wondering: yes, AI had a part in writing this article. We bet you can’t tell where or how.)
The promised transformation may turn out to be subtle, but powerful: writers will become editors. taking on the job of guiding AI to deliver suitable content, reviewing and polishing the results. Your chatbot buddy can help you scale up your output, and do more with it. But human readers will always respond better to a human voice. If you need help with copywriting to support your marketing strategy, it might be worth talking to the experts. We’re right here.