Generative AI is here to stay. But most marketers are using it wrong. Tools like GPT-4 aren’t order-taking geniuses that you can feed complex requests to with the certainty that they’ll produce something useful. But, if you truly understand their possibilities and limitations, they can make your workflow more efficient.
And, in this post, we’ll share everything you need to know to elevate your workflow through AI.
- What’s generative AI
- When to use AI for marketing
- What makes a good prompt
- How to write a good prompt
Without further ado, let’s get started!
What Is Generative AI (Actually)?
To really take advantage of generative AI, you need to understand how it actually works. In short, generative AI is a type of artificial intelligence that can create original content or generate new data based on existing patterns or examples. This technology uses algorithms and machine learning techniques to learn from large datasets and then create new content that is similar to the original data.
For B2B marketers, generative AI can be a powerful tool for creating personalized content and automating certain aspects of the content creation process. This technology can be used to generate product descriptions, marketing copy, and even entire articles or reports. By using generative AI, marketers can save time and resources, while still producing high-quality content that resonates with their target audience.
However, it's important to note that generative AI is still a relatively new technology and requires a high level of human oversight to ensure that the content it generates is accurate and relevant to the target audience. Implementing generative AI in content marketing usually requires:
- Feeding the AI with custom, human-curated information
- Careful prompting
- Human, goal-oriented revision & editing
Aside from blogging, some of the most common applications of generative AI in marketing include:
- Creating personalized email marketing campaigns
- Automating the creation of social media posts
- Generating product recommendations based on customer data
When to Use AI for Marketing
In a recent New Yorker article, Jaron Lanier pointed out that:
“[T]he most accurate way to understand what we are building today is as an innovative form of social collaboration. A program like OpenAI’s GPT-4, which can write sentences to order, is something like a version of Wikipedia that includes much more data, mashed together using statistics. Programs that create images to order are something like a version of online image search, but with a system for combining the pictures. In both cases, it’s people who have written the text and furnished the images. The new programs mash up work done by human minds.”
In short, generative AI isn’t creating anything from scratch. And it’s not a conscious entity with true understanding of your prompts. Instead, it’s a tool that remixes existing information to create seemingly new material.
With that in mind, we don’t recommend using AI to create long-form, conceptually complex content.
Instead, you can use AI to:
- Rewrite content
- Repurpose content
- Create first drafts
- Produce straightforward, entry-level definitions
With these recommendations in mind, you may be tempted to use AI to create programmatic content. But that’s definitely a double-edged sword.
Let’s briefly explore why.
Using AI for Programmatic SEO
Programmatic SEO can be a great strategy to establish your company’s reputation as a useful and value-generating brand. And this technique’s scalable nature may tempt you to try and automate the entire content production process.
In some cases, programmatic SEO content can be fully delegated to your users. In a sense, you become an aggregator and curator. But, if you’re trying to start a programmatic content collection from scratch, automating it can be detrimental to your success.
Regardless of what niche you’re in, all your content should be highly valuable - even your programmatic pieces. And, if your plan consists of creating a collection of entry-level concept definitions, it’s very likely to underperform.
Regardless of how you create your programmatic content, make sure that every page you create offers unique value. If you’re not offering interactive content or unique assets on every programmatic page, your value should come from the content you create. So AI can be a time-saving tool, but it shouldn’t be the end-all-be-all of your content creation process.
What Makes a Good Prompt?
Good prompts are:
- Direct - they have a singular, well-defined goal in mind and enable the AI to give a concrete answer.
- Contextual - they include key information that helps the tool to deliver a relevant and valuable output.
- Actionable - they can be answered satisfactorily within the tool’s limitations.
When writing your prompt, you’ll need to be extremely precise and to-the-point, while giving the tool any information it may need to fulfill the task to the best of its ability. In some cases, being concise and clear is enough. On other occasions, you may need to add some extra information.
In fact, the more information you can provide about your target audience, competitors, or industry trends, the better the AI will be able to understand your prompt and generate relevant content. Additionally, providing examples of the type of content you're looking for can help the AI model better understand your prompt and generate more relevant results.
For example, let’s say you want to generate a definition for the concept of generative AI. You can get a pretty solid response just by asking the model to “define what’s generative AI”. But you can experience a huge leap in quality if you take it a step further and add:
- Tone & format requirements
- Audience specifications
For instance, “Writing in proficient, conversational American English, define what’s generative AI. Your audience are highly tech-savvy startup developers. Keep the definition under 100 words long.”
How to Write a Good ChatGPT Prompt
Now you know what generative AI actually is and when you should and shouldn’t use it. So, let’s move on to how to write a good ChatGPT prompt, with the tool’s limitations in mind.
We recommend you:
- Work to understand these limitations
- Be specific
- Start small
- Use custom data for high-quality results
Let’s take a closer look.
Understand AI tools’ limitations
Remember that ChatGPT and similar tools are unable to:
- Come up with 100% original ideas
- Browse the internet by themselves
Of course, you can create a tool that, for instance, combines Python-powered content scrapers with the OpenAI API, allowing you to feed data from the internet to the AI in real-time, and thus get outputs that are informed by fresh, custom data.
But if you go on ChatGPT’s website or visit OpenAI’s playground and try to get the model to browse the internet, it will fail. For instance, if you ask it to analyze a specific URL and provide technical SEO suggestions, you’ll only get generic answers, not site-specific recommendations.
So if you're not using the model as the foundation for a custom AI-powered tool, you can't expect it to enter a website and give you an answer that's informed by its content. Having this and other limitations in mind can help you adjust your expectations and fine-tune your prompts to get actually applicable results.
Avoid using complex sentences or multiple ideas in one prompt. Instead, break your prompt down into smaller, more specific tasks that the AI can easily understand and execute.
If you want to make multiple requests or ask the AI to create complex long-form outputs, break down your desired output into multiple specific prompts.
For example, if you want GPT-4 to create a long-form blog post draft, don't start by asking it to give you 2,000 words of content. instead begin with some contextual prompting, ask it to give you an outline, ask it to write an introduction, and slowly build up the draft while adjusting each output as you go.
Here are some example prompts to get you started:
- You're a plumber and plumbing expert that writes accessible guides for everyday people. You write in proficient, conversational, and accessible American English. You keep your sentences under 15 words long and your paragraphs are very short. You include lists to make your content scannable. If you understand, answer "yes".
- Write an outline for a guide that will help new homeowners-to-be to assess whether a home's plumbing is reliable and safe when visiting open houses.
Use your own data
You can use your own data either by building a custom AI-powered tool, or even by pasting any relevant information on the “System” panel of the OpenAI playground.
Using your own data can be extremely useful for creating content about very specific topics and synthesizing long-form material.
In this post, we covered the basics of using AI for content marketing & SEO. Additionally, we shared some tips that will help you streamline your content production process with AI. But AI can only go so far. Especially in hyper-specific niches.
Without the help of an expert content team, you won’t have the strategic proficiency or the user-centric human touch that your brand needs to succeed.
Some of the world’s most promising startups choose us as their content partner. Book a free consultation and discover why.