Is AI good at content yet?
Experiential learnings from our team to yours
Hello all, and welcome back to this month’s Optimized newsletter - AI update edition.
6 months ago, if you popped onto Twitter (now X) or LinkedIn, AI was being thrown around as the answer to life, the universe, and everything for marketers. We covered the hype and offered our take in a Content That Grows episode back in March, and honestly, it’s aged well.
Early adopters were rabid.
We were all going to lose our jobs.
New products that were essentially ChatGPT skins blew up seemingly overnight.
VC money shuffled its way to everyone who could shoehorn some version of AI into their product (or at least the marketing of it).
It’s been overwhelming to sift through the mountain of opinions about the capabilities of this technology. In fact, almost every single guest on our podcast noted that the most annoying thing on social media has been generative AI.
So, what do content marketing leaders do when something disruptive and unknown appears on the market? Gain first-hand experience.
After a few pilots using AI for SEO content with eager clients (we call it hybrid content) and on our own website, here are our takeaways for getting the most out of this technology’s current limitations.
***Note: Most of our experimentation has been with ChatGPT as we continue to test which platforms are worth investing in long-term. We understand some of these have more robust feature sets.
Works great for ToFu content
You know those top-of-funnel topics that you struggle to justify the prioritization of because they’re miles away from your software product, but you know you’re an authority in that space, and you wouldn’t mind the visibility that comes with those keywords?
These are your “what is customer service” posts or “why customer success is important” posts.
Generative AI now allows you to produce those VERY efficiently.
The argument against AI is often that the content is generic. And that’s largely true. But it’s also the case that somewhere between 70-85% of a ToFu blog will be generic because it’s meant to be high-level, introductory coverage of a broad topic.
Aka, generic content is not always fluff; sometimes, it’s the necessary descriptors.
For example → This definition is just fine. It concisely says what it is and how to understand it relative to other models.
Generated with GPT-4:
Where AI still struggles, although a bit of training with prompts or, if your tools of choice allow, on your own website URLs can help, is still in the MoFu and BoFu arena of content topics. These are also often your product-led content topics.
These posts benefit from current and specific data, and in our experience, unique opinions and positioning from your company’s perspective play an outsized role in the impact of these pieces’ ability to activate a prospect.
Works best for teams who prioritize publish-first, perfect later
For a distribution channel like SEO, teams that consistently hit publish are often rewarded (as long as they still hit the minimum quality thresholds).
AI is great for helping you get that initial draft done, published, and indexed on Google. Then, once you have your SME notes, design images, podcast clips, or those product videos back from editing, you can seamlessly bulk it up to be that fully differentiated, better-than-the-rest post.
This is key to flexing your content/project management skills vs. relying on your individual contributor writing skills.
Recommended podcast episode with Jakub Rudnik of ActiveCampaign. He mentions using AI to create simpler content such as building out a glossary.
Quirks to watch for (fluff, lists, and repetition)
Depending on the types of content you’re working on, you might notice that AI seemingly has a few quirks.
You can consistently remove the first and last sentences of paragraphs to make the copy more reader-friendly. Otherwise, it loves overusing transitions and recapping its points.
Watch how you prompt topics, or you’ll notice you generate responses formatted as lists every time. Have it explain specific things to you rather than prompting it with your overarching section theme or H2s to avoid non-stop listicles.
AI listens well, but sometimes too literally, to your instructions when you tell it to emphasize, for example, a particular CTA or the goal of the overall post. You’ll notice that, in this case, it’ll often repeat, recap, and provide way too much overlap across sections of a blog post. To avoid this, each section of an article should have its own prompt. This gives you more dynamicism and helps differentiate your posts.
Can you rank AI-generated content?
You sure can.
Will you get hit with a penalty?
We haven’t yet across 2 core Google updates, but we didn’t try abusing it, either. We still create a content brief and do the competitor research needed to create a manual outline.
Then, we prompt each section using AI, perform minor edits for correct positioning, insert our CTAs, and do proper internal linking.
After that, we have a copy editor go through and align everything to our clients’ needs.
Initial results for a client beta test are positive. We’re helping with ~ 6-10 posts per month of AI content and have been doing so since March (~80 posts). The result has been an average increase of 2400 organic clicks per month.
(/blog URLs)
Don’t worry; there is a case study coming soon.
Here’s what we’re betting you shouldn’t do and what those getting hit hardest by the algorithm are doing with AI.
Creating a high volume of very thin content (short, uninformed, no value-add included). If Google thinks you only do thin, unengaging content for the majority of your posts, you’ll likely get dinged (especially if you’re getting into the hundreds of posts per month levels).
Creating content so far outside your realm of authority and expertise that the content is never even indexed. Yes, Taylor Swift gets a lot of monthly searches, no, your software has nothing to do with her.
Pro Tip: Authority will likely continue to play a more significant role in your ability to rank if the playing field for creating content continues to level. In addition to clustering internal links, you’ll want to start prioritizing a concerted effort to garner backlinks.
Start building a marketing resource bank ASAP
If you’re active on social media, have a podcast, create thought leadership posts, have a handful of case studies, and guest on other people’s webinars and newsletters, you should have a bank of amazing resources.
Podcast clips
Infographics
First-party data and charts
Explainer graphics and illustrations
Video thought leadership and opinions
Guest opinions and quotes
If you have an integrated strategy approach to content marketing, you’re likely circling around related topics consistently in all your asset types.
These resources make creating the highest quality AI content much easier and faster because it becomes a construction process of prompting, reusing, and remixing rather than creating something new for every part of every single piece of content you’re working on.
We’d like to hear some AI recommendations from you
We’ve primarily used ChatGPT for copy while testing more robust options and dabbled with Midjourney for images.
What AI tools do you like? What tips or tricks have you learned to this point?
We’ll shout out the best ones we see on LinkedIn.