Good to Great: How and Why We Humanize AI Content

Is it just me or does it feel like content marketing changed overnight? It’s easier than ever for anyone to instantly generate a blog post using AI. All you need is a basic brief or simple prompt. 

Boom. 2,000 words on whatever you want.

But the results from creating content this way? They often leave a lot to be desired. Even with an intricate AI workflow based on your brand voice, target personas, and guardrails in place, AI content still comes out feeling awfully… generic.

In many ways, the use of LLMs like ChatGPT and Claude is compounding a problem that existed before AI. We were already drowning in a sea of sameness. Now, we have a sea of sameness at scale.

Thankfully, there’s a way to break free. You can take advantage of AI and still publish content that’s a homerun.

Throwing strikes with humanized AI content

Nearly everyone can throw a baseball. But there are only certain pitchers who make it to the majors and a select few who can throw the real 100 MPH heaters.

There’s actually a significant difference between a 92 MPH fastball and a pitch travelling at 98 MPH. A few more miles per hour means a lot more batters swing and miss. That’s because the slightly slower pitches are predictable. Batters get used to them and can time their swings appropriately.

AI-generated content is fast too. But, like an average throw, it can be pretty predictable (hallucinations aside).

Humanizing your AI content is like adding some extra oomph to a pitch. It’s what turns a halfway decent blog post from ChatGPT into something your audience will read, remember, and share.

The upside: What does good AI content look like?

At Blue Ivory, we completely understand how and why AI is transforming the way our clients produce blog posts, landing pages, eBooks, and social media content. Anyone who isn’t taking advantage of AI is missing out.

When you generate a blog post with AI, for example, it can help in several ways:

1. Speed: It’s fast. First drafts used to take hours, but AI content gives you a headstart.
2. Reliability: AI doesn’t call in sick or get writer’s block. It’s always ready to pump out something else.
3. Scalability: Need ten basic blog posts on a topic in a hurry? Your friendly neighborhood LLM is here to help you expand your content library.
4. A reduced workload: When you’ve got a small but busy team, AI can take care of tedious content marketing tasks like SEO updates.

Good AI content may have a solid structure. It’s usually safe and conventional. It reveals surface-level insights that are helpful to some readers. 

Plenty of content from LLMs sounds a lot like a typical blog post because the models were trained on typical blog posts.

But typical isn’t what breaks through the noise. To win at the highest levels of this marketing game, more content faster shouldn’t be your only move. You need better content. That’s where humans come into the picture.

The downside: five ways AI content goes wrong

Sounding generic isn’t the only problem with publishing unchecked AI content. Here are a few of the bigger risks of relying too heavily on AI for content marketing:

1. Overconfidence without expertise

Even with extremely vast training data, AI still makes mistakes. That’s why you always see a disclaimer in tiny print just beneath the chat interface as well as with Google’s AI overviews.

The Dunning-Kruger Effect (you don’t know what you don’t know) seems to impact AI almost as much as people. It could also be making that kind of cognitive bias worse in humans. That’s exactly why you need real human expertise in the mix.

2. SEO and quality concerns

A lot of marketing teams are using AI to help them produce content designed to rank in search engines like Google. The risk is that, if you’re doing this too often without human oversight, AI content may not improve search visibility.

Google pays close attention to content quality. While they’re a leader in AI technology, Google also employs human reviewers to judge the quality of what’s published online. The reviewers use what’s called the E-E-A-T framework, which stands for Experience, Expertise, Authoritativeness, and Trustworthiness.

A study on the impact of mass-produced, AI-generated content in Search Engine Land found that visibility decreased significantly over time. The author believes failure to demonstrate E-E-A-T was a key factor and concluded:

“Early traffic and impressions may look promising, but without a clear SEO strategy and human guidance, those gains will likely fade within a few months.” – Bogdan Babiak, CMO, SE Ranking

3. Overuse of certain words and phrases

The generative AI critics out there often refer to it as “spicy autocomplete.” That’s a little harsh, but they’ve got a point. While LLMs do more than simply generate the next word, they still rely on common patterns and familiar arguments in language.

We all know the em dash got a bad rap thanks to LLMs overusing it. There are plenty of other signals that suggest AI generated the content. That includes the use of antithesis: “It’s not just about X. It’s also Y and Z.”

AI can also overuse certain words, such as “delve,” “leverage,” “landscape,” and “tapestry.” Phrases like “at the intersection of,” “digital era,” and “navigating the complexities …” also pop up a lot.

It isn’t wrong to include em dashes or words and phrases like that. But the LLM’s propensity to use them made their appearance in content a bit of a tell. The danger is that prospects reading your content dismiss it as less valuable or unreliable.

4. Inaccuracies and hallucinations

While ChatGPT, Claude and Gemini are regularly improving and releasing new models, they’re still far from perfect. When it comes to generating content around data and research, facts sometimes get misconstrued, misrepresented, or worse.

Imagine that ChatGPT gave you the perfect statistic for something you’ve been researching. Then you try to check the source and it’s nowhere to be found. Fabrications like that can even trip up the biggest consulting firms in the world.

It happened to Deloitte. The company produced reports for government clients containing facts from academic studies that didn’t exist. One of those reports had a $1.6 million price tag. Deloitte had to admit the mistakes occurred because AI assisted with the writing.

5. Damaged trust and reputation

It all comes down to trust. One of the main reasons you’re publishing content is to earn the trust of the people you want to reach. People won’t buy from you until they know they can trust you.

If your content is too basic, too generic, inaccurate, or totally wrong, the people you’re trying to convert could easily lose trust in your brand.

Ever felt silly when you watched an AI-generated video online and thought it was real at first? Sometimes people feel duped when they realize they’re reading AI content. That’s not what you want when someone is evaluating your product or service.

The people visiting your website are taking time to explore your content. Shouldn’t you take the time to make sure it’s trustworthy, accurate, and valuable?

Why AI content needs a human in the loop

The best way to get the advantages of AI while avoiding the pitfalls is to include humans in the process, both at the start and end of content creation.

Here’s how it should really work:

1. Start with thoughtful human input: Give the AI the information it needs to produce the best results. That includes a strong prompt, a detailed brief, existing content examples, writing style guides, and even a knowledge base to access.
2. Produce AI-generated drafts: Have the AI ask questions to clarify what you need before it starts generating. Then, let it do its thing.
3. Finish with a human-led review: Put the AI draft in front of someone who has the time to do thorough editing and fact-checking. The goal is enhancing, optimizing and humanizing the draft.

When it comes to truly making AI content more human, there are no shortcuts. You’ll find AI-powered tools claiming they can humanize your content. But stop for a second to think about the irony of using AI to humanize AI content.

Using AI to fix AI content doesn’t solve your problems or address all the risks. It simply adds another layer of AI to what you want to publish.

How Blue Ivory humanizes AI content for our clients

AI makes it easier to throw content together. But if you want to throw a fastball right over the plate, you need to bring in the taste, judgement, and perspectives of real people. That’s how you add the extra six MPH to game-winning pitches. 

Many of our clients come from fast-paced B2B SaaS companies that were among the first to start using AI in content marketing. What some are finding is that the effort it takes to thoroughly review and humanize AI content reduces the time-saving benefits that were the initial promise.

You want your team doing meaningful work. You don’t want them babysitting AI. That’s why Blue Ivory started offering AI content editing as a service.

Here’s what our editors are looking for when they review and humanize AI content:

1. Fact checking: Can we verify claims and stats made in the article?
2. Business accuracy: Does the copy accurately describe your products and services?
3. Structural editing: Does the piece flow the way a human would make their points? 
4. AI slop cleanup: Can we remove and rewrite AI filler and recognizable tropes to improve the reading experience?
5. Brand alignment: Does the tone and style adhere to your brand guidelines?
6. Additional expertise: Can we include quotations from your thought leaders or executives? Are there testimonials, case studies, or customer stories that would enhance the piece?
7. Audience specificity: Does the content speak to your readers or to a generic version?
8. Originality: Does the piece have a point of view or does it just cover the topic?

In addition to those steps, Blue Ivory’s team also does traditional proofreading, considers SEO best practices, and looks for opportunities to add internal links to your content.

Send an AI-generated blog post to us for editing and what you get back is something that’s accurate, on-brand, authentic, and has a human touch. That’s content you can be proud of and is ready to publish.

The debate over AI vs human content writers is over. The truth is, you need both.

AI puts your content on the pitcher’s mound. Editing with a human touch is how you throw that 98 MPH fastball.