AI in action

– 8 min read

What marketers need to know about AI-generated content

Katherine Duh

Katherine Duh

What marketers need to know about AI-generated content

Summarized by Writer

  • The hype around AI-generated content is real, but the vast majority of content generation tools on the market produce generic content.
  • In order to use this technology most effectively, marketers need to fine-tune their models for their specific use cases and train the models on their own (expertly human-written) content.
  • The key to high-quality AI-generated content is bringing in good writers who understand how content ladders up to relevant strategic goals, and what makes content high-quality, factual, and captivating.
  • The increasing usage of AI will change marketers’ jobs for the better, by replacing the routine and repetitive with more fulfilling strategic and creative work.
  • AI-generated content will, instead of causing layoffs, actually create more content marketing jobs.

Recent advancements in machine learning have meant that AI has gotten better at writing content that sounds like a human wrote it. How will marketers be able to best leverage this new technology, especially as they’re increasingly being asked to do more with less?

In a recent webinar with Insight Partners, Writer co-founder and CEO May Habib interviewed three marketing and brand leaders about how marketers can use AI to scale content teams, align their brands, and create high-performing content:

  • Jamie Barnett, Advisor, investor, and board member
  • Ty Magnin, VP of marketing at Emotive
  • Ryan Law, VP of content at Animalz

Here are three major takeaways from their discussion. You can watch the full webinar recording at the end of the article.

Takeaway 1: The hype is real, if you train your own models or use AI for specific parts of your creative process

May Habib started the webinar by giving an overview of the current state of natural language generation (NLG) technology. The vast majority of content generation tools on the market produce generic content, and many marketers who have experimented with the tools have come away unimpressed. The key, said Habib, is that in order to use this technology most effectively, marketers need to fine-tune their models for their specific use cases and train the models on their own (expertly human-written) content.

“Fine-tuning is basically taking a large foundational transformer model or large language model and narrowing it more precisely for a specific set of tasks,” said Habib. By fine-tuning a model, you can train it to more effectively accomplish the specific types of writing that you want — like composing blog posts, summarizing a piece of text, or writing press releases. “When the fine-tuning is done on a company’s own content, the results are actually pretty good,” said Habib.

Marketing not only requires writing across the entire user journey, from the top to the bottom of the funnel, it also requires many different types of writing. Ad copy, landing pages, email newsletters, help articles: with so many touch points requiring written content in so many formats, there’s a broad range of potential opportunities for fine-tuned AI to help expedite the writing process.

The writing we do in marketing: A screenshot from the webinar demonstrating all of the different parts of the user journey that require marketers to create written content: from ads in the awareness phase to support articles in the retention phase.
A screenshot from the webinar demonstrating all of the different parts of the user journey that require marketers to create written content: from ads in the awareness phase to support articles in the retention phase.

Not all of marketers’ writing needs are created equal. “There are parts of our job that aren’t just hard; they’re actually, in some cases, unskilled or unfun and unsatisfying,” said Ryan Law. A lot of marketers’ time is spent on routine and predictable writing tasks like rephrasing and summarizing content for promotion on different channels, creating utilitarian SEO content, or iterating on dozens of similar ad copy variants — and it’s these kinds of tasks that are great use cases for AI-generated content tools.

“What are the discrete process improvements we can unlock using AI?” asked Law. He then went on to cite a specific example from his experience at Animalz: generating blog post titles. He asks his writers to provide 20 different titles for every blog post they create because the process of iteration is “where the magic happens.”

However, he recognizes that this kind of iteration can be very difficult for humans: “We pick one title that we like and we get it lodged in our brain…and we then create 19 minor variations of that.” AI isn’t susceptible to this particular human cognitive bias, which makes it well-suited for coming up with dozens of different blog titles in a fraction of the time it would take a human.

Ryan Law discusses how marketers can unlock discrete process improvements with AI.

Takeaway 2: The bar is still great content, which means you’ll still need good writers

When it comes to the quality of AI-generated content, Jamie Barnett said she’s still “very skeptical” of the efficacy of most AI-generated content  and hypothesized that “the biggest problem is that a lot of it is created by non-writers.” Barnett then went on to outline five requirements marketers can use to ensure high-quality writing when using AI: 

  1. Bringing in good writers: “People who care about good writing and the fundamentals” need to be in the driver’s seat, and guiding the AI in the right ways.
  2. Training the model on your own voice: To make sure that the generated content sounds like your company, you need to fine-tune the model with your best content.
  3. Incorporating your style guide: In order to create on-brand content at scale, AI-generated content should automatically abide by the style rules that give your brand its unique voice.
  4. Having guardrails for bias: Content should promote healthy and positive communication, use inclusive language, and be the result of models that adjust for underlying bias in the data set.
  5. Validating claims: To avoid publishing inaccurate content, AI should either be able to identify if a fact is true, or be able to flag where a fact needs validation from a human fact-checker.

There’s a lot that AI can do, but ultimately, human writers are still the people who best understand how content ladders up to relevant strategic goals, and what makes content high-quality, factual, and captivating.

Jamie Barnett outlines the five requirements for high-quality AI-generated content.

The role of human writers in the production of AI-generated content continued to be a theme throughout the webinar. The first question from the audience was about how search engines like Google differentiate between SEO content written by an AI and SEO content written by a human, and what they might or might not do about AI-generated SEO content.

“My pet hypothesis is that Google probably doesn’t want AI-generated content, but it’s probably going to happen anyway, and there’s not much they can do about it,” said Law. Because these latest language models are creating net-new, original content by using the rules of language, it will be much harder for Google to detect AI-generated content. Furthermore, it’s not always obvious how to define whether something is “written by AI” or “written by a human.” “If an article is outlined by a human, generated by AI, and edited by a human, should that be penalized?” said Law.

Habib agreed that search engines will probably face difficulty in defining and identifying AI-generated content. She predicted that Google might take the approach of identifying and encouraging writing that emphasizes thought leadership and personal anecdotes. “What is easier to do in the rankings is to privilege that content that is more personal,” she said.

Takeaway 3: AI isn’t going to take any writing jobs, but it will probably change many writing jobs

Individual content marketers can be apprehensive about the idea of AI-driven content generation, especially in a period of deep layoffs and cost cuts in tech. However, the panelists all agreed that the increasing usage of AI would change marketers’ jobs for the better.

“Writing has for me two parts: a smart part and a slog,” said Barnett. “I can understand some trepidation, but for me, as a content generator myself, I love the idea of leverage.” Law concurred, predicting that future content roles would be “more fun” and “higher leverage.”

Ty Magnin had an additional perspective on the question of automation.

“If you take a step back, it’s about automation and it’s about AI. This dialogue’s happening at a very high level, and it’s been going on for years,” he said. “Sometimes, people get displaced, and it sucks. But more often than that — five times, 10 times — you get to see people really focus on higher value work.”

For Magnin, automation represents an opportunity to fundamentally change the nature of work for the better, by replacing the routine and repetitive with more fulfilling strategic and creative work: “For content writers, now you can focus on strategy. What posts are you going to write? Where are you going to distribute it?”

Habib agreed, and predicted that AI-generated content would, instead of causing layoffs, actually create more content marketing jobs: “AI generated content is going to mean more content marketing jobs, because people are going to get leverage faster — which means content marketing will impact marketers meeting goals faster, which means marketers get more budget to do more content marketing.”

Ty Magnin and May Habib discuss how automation can empower marketers to focus on more impactful work.

To hear more insights from our panelists, watch the webinar recording. You’ll get a great overview of the current state of NLG technology and the landscape of existing AI content generation tools, and learn more about how AI content generation will change the way we market content.