In short
Getting the most from AI generated content means finding a balance between software application, human intervention, and your processes for improvement and approval. The wrong mix will consume resources and stop you from scaling – so it’s vital to customise the solution for your specific use case. This article tells you how.
Key takeaways
- AI content generators differ in philosophy and purpose
- Not one is “fire-and-forget”; human curation is essential
- Output quality is a function of input quality (like the prompt)
- Systems for quality control and risk management are essential
- The best AI results come from a mix of software, people, and processes
Introduction
The trouble with Artificial Intelligence (AI) is that it’s, well, artificial. AI content generation is an incredible technology, with vast potential. But it’s not human and isn’t embodied in our skin. It doesn’t have our talents for emotional attachment or empathy towards others. And that’s fine.
Because AI isn’t a replacement for humans, but an assistant for them. And like all assistants, it works best when its talents are used as part of a team: knowing what everyone’s tasks are and what its own role is. Getting it right isn’t easy. This means that, when it comes to AI generated content, the purpose most marketers are interested in applying it to, companies are experiencing a wide range of success levels.
Some have treated AI content creation as the new killer app, but ended up with bland blurbs that say nothing special. Others have trained themselves to brief these tools precisely, but discovered it takes longer to write the prompt than it takes a human writer to complete the article. And there’s a vast middle group of organisations making genuine efforts to leverage AI in their blogs and articles, but finding the SEO reach isn’t what they expected.
There’s an answer to this but it’s not so much a technology as it is a strategy. It involves seeing AI generated content as part of a system, not a standalone application. In this system, every moving part has a distinct goal, success metrics are known, and human professionalism, machine intelligence, and approval pathways are designed to work together.
We don’t have all the answers at Contentoo but we’ve certainly discovered a few things that work. So, let’s take a deep dive into how AI content generation works, and the roles you need to define in order to create great content in this exciting new world.
What is AI generated content?
Content produced by machines, called “generative” AI, isn’t the same as the creative process of humans. Rather, these “Large Language Models” break down your initial instructions (called a “prompt”) into small fragments, analyse the patterns they fall into, and return text that’s statistically likely to be a proper response.
It’s the same principle as AutoComplete on your mobile phone, guessing you’re going to the “shops” when you type “I’m off to the sh-”. But unlike the few megabytes your phone uses, an LLM (like ChatGPT) generates its response based on hundreds of billions of examples, making the statistical probabilities of a good answer very precise.
Statistical inferences, however, have some strange real-world outcomes. It’s common for AI content generators to simply make things up, since they fit the expected statistical pattern. And when you’ve learned from the internet, your output will only be as good as the internet. In other words, the data used to train an AI contains all sorts of assumptions, biases, and plain old mistakes.
So, that’s our first learning: these incredible technologies aren’t perfect. They’re not human but they are fallible.
Benefits of using AI generated content
An LLM-based generative AI can produce AI content practically instantly: it’ll start generating as soon as you hit Enter on your prompt. Which means you can produce masses of content at scale, and fast. This is great if you’ve got 10,000 product descriptions to write, or want to test a dozen different CTAs on your marketing emails. As well, this works if you’re simply short on inspiration and want ideas.
And since it’s trained on countless examples of language in use, you don’t have to accept an AI article generator’s standard bland voice; tell it to sound jaunty, or funny, or like an obscure 16th-century playwright, and the results can be surprisingly good. So, it is possible for an AI to use your Tone of Voice, too.
Best of all, of course, an AI is a machine. It doesn’t get bored, or need breaks, or a salary. (Yet.) So, it can burble away 24/7 generating writing for your web pages and blogs, increasing your web footprint for visitors. There are, however, a few downsides.
Potential risks of AI generated content
Even Sam Altman, CEO of OpenAI (the company behind ChatGPT) freely admits it: “[AI] still has serious weaknesses and makes very silly mistakes.”
While it can sound just like a human, remember your AI assistant isn’t human, and lacks the tacit “high-touch” knowledge that lets us interact with other people on a basis of shared understanding.
This means the answers it gives aren’t informed by things like moral judgement or ethical behaviour. (Or even accuracy, if accurate data wasn’t part of its training model.) It can treat causation as correlation, thinking two events are linked even if it’s just coincidence. It can draw insights and inferences that it can support with data, but which simply don’t make sense. Perhaps, it’ll see a connection between your umbrella sales and weather, and advise your marketers they should be making it rain, or attribute a quote to the wrong person simply because it fits the pattern. And so on.
All this means that whatever AI tool you use, it needs to be supervised and its content checked. The consequences of getting this wrong are considerable.
Google's stance on AI generated content
Some marketers believe AI generated content writing won’t get the Google Juices flowing (i.e. it’ll be bad for SEO.) This isn’t true. Google doesn’t care about who or what wrote the content – it just cares that the content is useful to people searching the web. And a well-prompted AI-written article can be just as useful as a human-written one.
That said, a lot of content is appearing that uses a lowest-common-denominator approach: the AI’s default voice, generic prompts, and bland, boring articles. This is bad for SEO, given that few people will be interested in it! So, Google’s stance is best described as “steady as she goes.”
Content marketing strategies using AI
Fortunately, these new tools aren’t limited to writing articles. When properly briefed, they can suggest content calendars, topic ideas, email campaign cadences, and a range of structures for marketing activities – nurturing pathways, sales funnels, customer milestones and so on, all of which is higher-level stuff previously performed by a manager.
AI can also be used to create media. There are tools out there that’ll produce a PowerPoint, animate a talking head reading a script, and fill your video backgrounds with relevant eye candy like stock images. The results can sometimes look a bit generic – but the time savings can be awesome.
And with data driving every business these days, AI can help you make sense of it. Many AIs can extract key ideas and principles from unstructured data (like customer service inbound emails) and show you what’s driving your biggest customer pain point, giving you subjects to use in your content marketing campaigns.
So when using these tools, don’t limit yourself to content for customers’ eyes – they can help out internally too.
How to use artificial intelligence in content creation
Here’s a general approach to making AI work harder for you. On a large sheet of paper (or a spreadsheet!), list some tasks you think it could help with. These could be:
- Web content, blogs and articles
- Campaign content, such as CRM marketing
- Service content, such as How-Tos and instructables
- Media content, such as YouTube videos
Next, list all the people who would benefit from these capabilities across the business, and who would be responsible for what the AI creates. This gives you your approval map: the quality control checks that keep your AI-originated content safe.
After that, produce some sample content to show these people, and work up some usable “prompts” they can use as models. This will let them use AI in a consistent fashion, with agreed processes for giving feedback, updating prompt principles, and adding new tasks to the list.
Doing this takes your AI content generation from software to system, which effectively turns this into a business process with checks and balances from the right human input. It won’t limit your AI’s usefulness but it will prevent bad prompts or inaccurate output affecting your customers.
Why is it important to humanise AI generated content?
So, it’s now obvious humans need to be in the process, as a chain of command that can blunt AI’s wild side. But the actual content generated by AI needs a human touch, too.
Being trained as an “average” of billions of pages of text means the content AI produces often sounds middle-of-the-road, like the easy-listening Muzak you hear in a lift. It’s boring and can be a turnoff. A human writer will add some zing.
But there’s a benefit here: your human writer, assisted by AI, can be far more productive than before. It’s possible to humanise many thousands of words a day, meaning you can create content at scale – as much as your marketing opportunity needs, whether that’s 10,000 product descriptions or 100 intro paragraphs for your email newsletter.
This mix of human talent and AI tech gives you the best of both: creativity with a natural human voice, in volumes no human copywriter could produce, answering your need for content without busting your budget.
And that’s where Contentoo can help.
What services will Contentoo AI provide for your company?
Contentoo started life as a content marketing platform, linking the top 10% of freelance talent worldwide with the companies who need them. That’s still our purpose. But for a year-plus we’ve been adding AI into our offer, finding the use cases where generative artificial intelligence can add value to the expert professionals we work with.
For our customers, that means greater volume, faster outputs, and variants so you can A/B test, plus the data and tools to make critical decisions about your marketing strategy. Everything you’ve seen above – blending people, processes, and technology – is what we do here at Contentoo. And there’s still time to join us at the start, with our Beta programme.
For additional insights on AI content tools that we recommend, feel free to read this blog post which features five AI tools for content creation.
How will Contentoo differentiate from other AI content tools?
Many AI content generators are good, but they often lack the human touch that makes content truly great.
Contentoo is a people business: our mission is to find talent all over the world, and connect it to the companies who need it. That talent gets even more effective when you swing in AI as part of the process.
So, that’s what makes us different to a content generation tool. We’re actively connecting people and processes to technology, with the goal of better, more effective content creation at scale for our customers.
How to choose the right AI content generator for you
So, it isn’t about finding an app you like. It’s about how well you merge its capabilities with your business processes, so you can get the most out of it for the perfect integration of human skill with machine intelligence. Test and train any AI tool you want, see whether its focus and approach answer your pain points, and gather data about how it performs.
And if you’d like us to take on that role for you, remember what Contentoo can add.
Make the most of AI with Contentoo
We’re a fast-growing team of marketers, writers, translators and strategists, working with many of Europe’s biggest companies on their content needs. We’re seeing new uses for AI every single day – and we’d like to share those findings with you.
AI is going to change your business. To make sure that happens in a positive way, why not join our Beta programme for AI content writing today?
