March 7, 2023
How will ChatGPT and generative AI affect the future of marketing?
As AI technology continues to advance, ChatGPT and other language models have the potential to revolutionize the way businesses approach marketing, providing new and powerful tools for customer engagement and insights.
– An opener taken straight from the lion’s mouth. A bit vague and lacking in personality perhaps, but undeniably impressive; ChatGPT might not be able to compete with professional writers just yet, but as you can see, it’s not far away.
ChatGPT is the frontrunner of the AI craze sweeping through the world – but as the chatbot would undoubtedly describe it, it’s just the tip of the iceberg.
From revolutionising Bing to enabling 30-minute conversations with a bot about your dog, AI is set to change the digital landscape, and marketing, perhaps more than any other sector, stands to be most affected.
Human vs robot: what can we do better?
Why you shouldn’t put all your eggs into one virtual basket
The future is now?
Artificial intelligence has made huge strides in an incredibly short space of time, but if it’s the future, it’s been here for a while. You’ve been using AI for much longer than the last six months, whether you’ve realised it or not. Tools like Grammarly, predictive text, facial recognition software and personalised social feeds all use AI technology.
We may be familiar with AI as a concept, but that hasn’t stopped ChatGPT from coming in and absolutely flooring everyone. Generative AI, which takes human prompts and generates something entirely new, is creating impressive pieces of content and art – it’s so good, in fact, that it can almost be mistaken for human.
Everyone’s talking about it, and it’s easy to see the appeal. At a time where budgets are tightening, AI could be a solution to drive down costs, cutting labour time and performing tasks that once took hours in seconds.
In marketing, the tech has the potential to completely change the industry.
We take a look at the ways ChatGPT and other AI can be used in marketing – how we can use it, its strengths and weaknesses, and why humans are still necessary.
What is ChatGPT?
Before we dive into the intricacies of AI technology and its impact on marketing, first thing’s first: let’s clear up what it is. ChatGPT is currently the most popular AI tool, so it’s the one we’ve chosen to place under the microscope.
ChatGPT is a chatbot developed by OpenAI that uses natural language processing technology to give human-like responses to prompts. To clear up yet another acronym entering the marketing world, GPT stands for Generative Pre-Trained Transformer, which is essentially just an elaborate way of saying that it has learned to respond to our inputs. You can engage in full conversations with ChatGPT, as it’s capable of remembering what a user has said earlier in the chat.
You can ask it to expand on questions you’ve already asked, write an entire blog on whichever subject you desire, or even request a Shakespearean sonnet.
As you can see, it doesn’t always get it right.
The AI was trained on copy written by humans – it’s used a vast amount of training data to learn how to have a conversation and generate text responses that mimic patterns of speech. ChatGPT has also been trained through reinforcement learning to decline inappropriate requests, and won’t be baited into saying hateful language.
End of the search engine?
ChatGPT has triggered the tech giants’ answer to the space race, with the likes of Microsoft and Google scrambling to catch up. While fears that advanced chatbots could replace search engines altogether are likely unfounded – some people will always be looking for a variety of responses or an article with verifiable sources – it’s enough of a competitor to worry the big guns.
Many queries once put to a search engine may be put to ChatGPT instead. Its ability to give users bespoke responses to their queries in a conversational manner may be preferable to some, in many cases making the search engine an unnecessary middle-man.
How we can work with AI
To use a very AI-esque cliché: work smarter, not harder.
Resistance to technological advancement is nothing new: it’s happened throughout human history. From the printing press to the internet, new tech is often met with fear and hesitance – but time and time again, it’s the people who fail to adapt that are left behind.
Revolutionary technology can change society, and forward-thinking individuals won’t treat AI as something to fear, but instead see it as an opportunity. Technology makes us more efficient, speeding up processes into seconds that would otherwise take days. Marketing teams can and should be using AI to their own benefit.
Sticking your head in the sand and ignoring new technology is never the right answer. We’ve put our heads together across the agency to investigate the ways we could use AI, analysing its potential for various departments.
With the most advanced AI currently the writing tools, it’s unsurprising that content marketers would be paying close attention.
Investigating both ChatGPT and Jasper AI has given the content team a good view on the technology. It’s impressive, but the general consensus seems to be that these tools require a great deal of human intervention to truly be usable at a professional level.
It’s best to think of ChatGPT or Jasper, or whichever AI tool you choose, as your very own virtual assistant. All those hours spent doing desk research, or time spent condensing information into fewer words, can be significantly reduced. While the thought behind the content is still in the realm of the human being, research and a good chunk of writing (heavily edited, of course), could be completed through AI.
These are just a few of the ways the content team felt they could use AI:
- Research – desk research, particularly for less technical or product-focused topics.
- Overcoming writer’s block
- With heavy editing, potentially writing the bulk of the prose for straightforward topics and simple content structures; allowing time to be used elsewhere
- Adjust existing copy to target different audiences, which could help with trends towards hyper-personalisation
- To simplify text. Sometimes you just need to get to the point, and AI can help us cut through the waffle to shorten our paragraphs and sentences
- For simplistic language, translation. While it would need to be checked by a translator, AI is capable of writing in other languages and translating many colloquialisms
- Use AI art generators to create images for articles.
If it’s a robot’s world, it wouldn’t be nothing without an editor and a good content strategy – James Brownbot, 2023.
Our digital marketing team also investigated the potential uses of AI writing tools. The team felt that, though a great tool for kickstarting ideas and research, the limitations of AI were immediately clear. Its inability to provide the human element behind understanding the nuances of SEO or engaging a target audience makes anything it produces very surface level.
Again, it would need heavy intervention from someone in the team to reach the level of quality required. In some cases, reworking AI’s creations could take longer than starting from scratch.
Like the content department, digital marketing viewed AI as a potential assistant, not a replacement. Many uses were similar to that of content i.e. research and creating text variations, but there were a few additional points:
- While it’s too top-level for social content generation (it can’t consider the audience in its messaging), it can paraphrase captions to meet character constraints e.g. 280 characters on Twitter
- Good for editing social content and for hashtag generation
- AI can assist in writing metadata – though, it’s worth highlighting, not necessarily optimised metadata
- Could provide a useful starting point for competitor research
- ChatGPT can generate HTML to add structured data to your website
- Could assist in creating social calendars; particularly for finding relevant awareness days.
Despite using a completely different AI tool, the creative team came to similar conclusions to content and digital marketing: using an AI image generator like DALL-E-2 is currently best suited to research.
These were a few potential uses for the creative team:
- To research art styles, themes and artists
- For storyboarding and getting ideas across to others
- Test out ideas quickly
- Discover older art styles
- Use it to compare artist techniques
Human vs robot: what can we do better?
At the moment? Really it’s everything except speed. Professional writers and creators are a level above tools like ChatGPT, and there’s still a slight uncanny valley gloss over AI-produced work – like a 50-year-old trying to pass themselves off as Gen-Z – that means you can tell it’s not quite human.
But we all know generative AI is a baby, and it’s going to grow up. No-one is really sure of the heights it will reach, but we are aware of its innate limitations.
Looking longer-term, these are areas humans will triumph over machine – and where marketing teams can really show their worth.
Strategy and expertise
A coherent strategy that is tailored to an audience, with extensive knowledge of a brand and their goals in mind, isn’t an option for generative AI.
Expertise and experience are where marketing teams can really add value to a brand. AI might be able to fill in some of the gaps, but it can’t create an entire marketing strategy or know a brand inside and out.
People are capable of innovation and creativity, and can draw on their knowledge of the industry to add something new to the conversation. Generative AI might be able to add a brick to a wall, but marketing teams will still be the architect.
Being a market leader
Brands want to lead the way – with AI, you can only ever be following the pack. You’re never going to add anything new to the conversation with AI, which makes vital marketing tools like thought leadership and case studies beyond its reach.
AI assimilates a vast quantity of data, but it’s out of date. ChatGPT, for instance, was trained on a dataset that is limited to 2021. It won’t be scouring the internet for updated information on your selected topic, but drawing from a database with set parameters – making it unaware of any more recent news or updates.
Generative AI is incapable of being truly original, and will always go with the safe, standard option: it may be able to produce some decent content, but nothing groundbreaking. Market leading brands will expect more from their marketing teams than AI is capable of providing.
Originality – it will push us creatively
There’s been a lot of talk about the harm AI will do to creative industries. But how about the ways it will push us further? By its very nature, AI is incapable of original thought: it can rehash an endless mine of data – and may, at some stage, write as well as the best of us – but content strategy? Fresh insights? A tailored brand voice? Those are beyond the capabilities of a computer.
Across the board in marketing, from design to copywriting, AI will push our creative limits. In order to stand out from the standardised blueprint of AI content and art, marketing teams will have to go a level above.
AI presents a problem for lazy content and creators. Specialists in unoriginal, uninspired regurgitations of the top search results on Google should be concerned. Journalist Sydney G. Harris stated this decades ago, but it’s particularly apt now: “The real danger is not that computers will begin to think like men, but that men will begin to think like computers.” Letting AI do the thinking for us, or sticking to the formulaic creations they produce, could lead to marketing teams becoming obsolete.
It’s not a harbinger, but for some it may be a wake-up call.
Humour and brand voice
Humour is so tricky that even humans, who have access to a wealth of experience and the ability to feel empathy, can be painfully unfunny. But we know what makes us laugh, and though humour is subjective, we’re generally aware when a joke falls flat.
AI isn’t capable of feeling emotion or drawing on personal experience, so has all the hilarity of a Christmas cracker. From a style perspective, it’s one of AI’s big weaknesses. And from our admittedly limited experience with prompting ChatGPT to be funny, it seems to draw heavily on references to Elon Musk…
How to get the best out of AI
Getting the best out of AI is a skill in itself. It’s all about asking the right questions, as the quality of the answers depends on it.
Prompts need to be clear and concise, detailing the format you want (i.e. a blog, a Facebook post etc.), the tone of voice or style of the content, and the subject matter. It may take a few follow-up questions to get the AI to produce results you’re happy with. To get the best out of AI, you will also have to act as editor. Cleaning up the copy, fact-checking and adding personality are all jobs you’ll still need to perform; especially if you want to avoid being penalised by Google for using AI content!
AI can be a huge time saver, taking away some of the mundane, monotonous tasks from your work day. These are a few AI tools that might, if they aren’t already, soon be creeping into your daily use:
- Content writing – Jasper
- Research/AI assistant – ChatGPT
- Note taking and transcription – Fireflies.ai
- Text to art image generator – DALL-E-2
- Grammar and spell-checker – Grammarly
- Create content for Twitter – Tweet Hunter
- Paraphrasing – Quillbot
- Translation – DeepL
- Transcription – Descript
Why you shouldn’t put all your eggs into one virtual basket
It may be tempting to throw all caution to the wind, jump on the ChatGPT bandwagon that’s sweeping through the digital world, and ride off into an AI-generated sunset. After all, Buzzfeed’s stock rose 92% after committing to AI – surely all businesses producing content will follow suit?
ChatGPT and generative AI’s advancement is an exciting development, but should be met with some caution. Stocks can fall as quickly as they rise (hello, Google), and AI as we know it in 2023 is in its early stages. There are some kinks still to work out, and some of these obstacles, as we’ll discuss, may be insurmountable.
It’s not quite there yet
There’s no denying that ChatGPT is impressive – the leap in technology is larger than anyone anticipated, and it’s got everyone excited.
But anyone planning on relying completely and utterly on an AI writing tool and publishing content without so much as a glance over will find themselves in trouble.
With an editor on hand, ChatGPT, Jasper or any similar AI tool can compete with an average writer. You can ask follow up questions to a ChatGPT response and get a logical answer, which is, without doubt, mightily impressive. But a heavy reliance on conjunctions and clichés distinctly mark a piece of writing as AI produced.
Creative tools like DALL-E 2 are even further behind, and will often require several input attempts to come close to what you’re looking for. Hands seem to be tricky, and what may seem obvious to humans – like how to hold a pen, for instance – may not be so simple to AI.
TL;DR: it’s good, but not that good.
It’s not a fact-checker
This one is huge. ‘Fake news’ and misinformation is nothing new, but for technology that wants to replace search engines, AI chatbots can get it dangerously wrong.
ChatGPT, Jasper and any other generative AI writing tool has been taught how to write persuasive, human-like responses in a conversational manner. It writes plausible sounding answers to your questions – whether they’re true or not, well, the AI doesn’t really care.
When they get it wrong, these chatbots are the definition of confidently incorrect. When you’re pulling from unverified sources all over the internet, there’s bound to be a few misses! If the content it’s using as a source had a bias or contained misinformation, ChatGPT’s response will too.
To see this in action, look no further than the rather unfortunate unveiling of Google’s answer to ChatGPT, Bard. It incorrectly claimed that the James Webb Space Telescope took the first pictures of exoplanets in an ad plastered all over Twitter.
It struggles with technical information
Keep in mind that the more technical or niche the question, the more likely the answer will be wrong, as the dataset is smaller.
Anything that requires recent information, like a newly released album or news story, will also be beyond its capabilities. Unfortunately, the mistakes are often subtle, and few people will do their due diligence when asking ChatGPT questions, risking the spread of misinformation; it’s a great reminder to always check your sources!
AI content is surface level and boring
As we touched on earlier, marketing teams will be pushed – if they weren’t already doing so – to produce more original, higher quality work to stand out. AI plays it safe and follows a clear formula, and excels where there is a lot of information already online.
To avoid getting things wrong, it tends to stay in the zone of surface level information, unless prompted otherwise. ChatGPT is often deliberately vague, and struggles with very technical information or the nitty-gritty of product information.
It typically struggles with tone of voice and writing with empathy or humour, and the structure is predictable, which makes it feel safe and, if we’re honest, boring. Excessive use of AI could lead to the homogenisation of content, causing readers to lose interest.
Uncertainty over legal ramifications
AI seems to work in a legal grey area as far as copyright is concerned. Writers and artists in particular are understandably up in arms about the role of original work in training AI. Every piece of art and content used to train AI was done so without permission, and isn’t credited or compensated.
It’s a big unknown hanging over AI. While there is nothing illegal about the way AI works at present, the law certainly hasn’t caught up to the technology.
Sam Altman, CEO of OpenAI, alluded to these uncertainties in a tweet; “we also need enough time for our institutions to figure out what to do. regulation will be critical and will take time to figure out”. It may be that copyright law is updated to protect original creators and restrict AI – we just don’t know yet.
Beyond that, there is a small risk of plagiarism when using AI. CNET was recently caught out for using AI-generated content – it was found to be a heavily plagiarised copy of competitors’ articles, where the AI had swapped in a few synonyms to disguise the almost identical text.
Well, more or less – but it’s the risk that really counts. Google and other search engines are constantly trying to detect AI content, and if it’s suspected to be AI-generated, your article will be pushed down the rankings.
Google uses an algorithm to decide if content is useful or not. Low quality work won’t be at the top of search results, and Google is likely to decide that an unedited piece of AI content isn’t worthy of ranking well.
The Ethics of AI
The ethics of AI has been as hotly discussed as its potential. People are excited and appalled in equal measure about the technology; but all can agree that its origins deserve further scrutiny.
OpenAI’s treatment of Kenyan workers
It may have only been on the scene a handful of months, but ChatGPT and OpenAI is already subject to huge controversy.
Reportedly only paying Kenyan workers less than $2 dollars an hour to train ChatGPT, OpenAI has come under serious criticism for its less-than-stellar work practices. Their Kenyan staff were tasked with sorting through a huge amount of text pulled from the darkest corners of the internet, with the end goal of safeguarding ChatGPT against hateful speech.
The work they were asked to do was described as “torture”, as they were subjected to explicit stories and harmful content, which many reported as damaging to their mental health.
While brands may be open to the new and fresh idea of an advanced AI chatbot being introduced to handle specific tasks, they may be less enthralled with being associated with the unethical practices of OpenAI. With such serious ethical concerns on the data science behind the language model, this is one new idea that might be deemed too controversial for businesses to use.
Backlash from creatives
For reasons mentioned already (namely copyright controversy), generative AI has been receiving a huge amount of backlash from the creative industry. Singer-songwriter Nick Cave described AI as “a grotesque mockery of what it is to be human”.
This is just a tiny sample of the more PG options available.
AI and marketing: the future?
We’ve looked at the strengths and weaknesses of AI, and how it could be used in marketing. While there are still a lot of uncertainties surrounding AI, it seems only likely to grow in influence, not disappear.
Generative AI is a tool that all marketing teams should be keeping close tabs on. Though it has its weaknesses, used in the right way, it could be a great time saver. But at this stage at least, it’s clear that AI is a tool, not a marketing solution.
Written by Jo Luffman, with contributions from the creative, digital marketing and social teams.