- 15 mins

AI for Content Creation: Where We Currently Stand at D*S

May 24, 2024

As you can imagine, AI is a hot topic for us these days. It comes from clients in two opposing directions: 

  • Some want reassurance that we’re *not* solely relying on AI for content creation (we’re not), that our writers are real humans (they are!), and that we consider things beyond SEO and keywords (we do!).
  • Others want more AI and reduced content costs. They hesitate about the price of content and wonder if it’s necessary and if AI can just do it for them. 

To be clear, we fully expect our stance to evolve and change. These are still the early days of this new world of AI. Tip of the iceberg and all that… So this isn’t yet another blog post hypothesizing about the possible ramifications of technology most of us don’t fully grasp. It’s a post about where we currently stand and how we’re talking about AI with our clients at Digital Sisco.

What Many Clients and Leaders Believe About AI

Reading LinkedIn headlines can lead you to some strongly held opinions. But not all of them represent today’s reality. Here’s just one that popped into my feed this week.

Altman

What’s buried in the body of this article is the timeline on this, i.e. "five years, give or take, maybe slightly longer." Basically, this is about the future, not today. We all have our AI predictions for the future. (We’re aligned with the version that looks a little like this.) 

But most of our clients are looking to hit numbers this month, this quarter. Where we’ll be in five years is an interesting conversation, for sure. We’re all talking about it. But it’s not where marketing is today.

To take another example, your execs might be reading “SEO is dead” headlines. 

SEO-Headlines

But then there are also posts like this:
<iframe src="https://www.linkedin.com/embed/feed/update/urn:li:share:7172950209944129536" height="470" width="504" frameborder="0" allowfullscreen="" title="Embedded post"></iframe>

And, of course, there was the now-infamous content heist that looked all fine and dandy until Google caught wind of it. But many leaders latched onto the first part and didn’t follow the aftermath.

Metrics

Basically, there’s a delta between where we all think this is likely headed and where we are today. We’re all simultaneously trying to work with today’s reality while also preparing for the future. And let’s face it: Execs are not in the weeds of the day-to-day. Their job is big picture. So, it’s easier for them to read these headlines, jump to grandiose conclusions, and assume they can slash their content and SEO budgets. 

Meanwhile, SEO and content teams (especially ICs) live very much in the day-to-day and understand that, executionally, things haven’t changed quite as much as leaders might think. But very often, their “hands-on-keyboard reality” is landing with their leaders as change-resistance or defensiveness instead of as a sincere reflection of where we are now. 

The Chatter: C-Suites Versus Content Marketers

Okay, so let’s look at the perspective of the people who have their hands on keyboards. A recent thread on Superpath’s amazing community (run by the ever-insightful Jimmy Daly) posed a straightforward question:

Superpath-1

There was a mixed bag of answers, showing a range of use cases and skill levels (no doubt connected with the time people actually have to experiment). But overall, there was no huge consensus around time savings, budget reduction, and certainly no headcount reductions.

Content marketers are generally using AI to find efficiencies in the parts of the writing process they dislike or feel stumped. For example, outlining is popular because it gets people past the blank page. But no decent content marketer is saying, “Yeah, I used ChatGPT/Gemini to write this blog post for me, and it did a stellar, hands-off job while I just sat back and drank a mai tai.”

There isn’t even consensus on time-savings; it’s more often adjustments to how time is being used. So you might spend less time on your first draft copy but more time than before on editing. Or you might rely on something like SurferAI to write an article but spend substantial time fact-checking and supplementing with data and authority.

How We Produce Content at D*S 

At Digital Sisco, we love building automations and finding efficiencies we can pass along to our clients. But we’re also a pretty risk-averse team. This isn’t (just) because we’re Canadian. It’s because we work with our clients for years rather than months. So if we build results on shady tactics today, we’ll be the ones digging ourselves out from under them later.

We produce content. And we’re damn proud of the content we produce. Here are the ways we currently think about content:

We Don’t Distinguish Between SEO Content and Content

SEO is a distribution channel for content. It might be your main distribution channel. But your audience doesn’t think about that. They think about the information they’re seeking and the quality of the response. While there are things you’ll do to ensure content ranks (TF-IDF, etc.), we’re firmly against any distinction between SEO content and, for example, brand content, thought leadership content, product content, etc.

It’s ALL brand content, really. And SEO deserves your best foot forward as much as content for any other distribution channel (PR or social or base newsletter). In fact, given SEO is many content marketers' primary distribution channel, it really deserves special effort and consideration.

We Use a ‘5 Levels Model’ to Produce Volume at Scale

Last year, there was a lot of talk about the amount of ‘grey goo’ content on the internet. There is indeed a problem: Most of what is published is absolute junk. But let’s be honest. The grey goo content problem existed before AI was even a twinkle in Sam Altman’s eye. Content marketers made this problem, and AI is making it easier for companies to compound the problem.

We do not make grey goo at Digital Sisco, but we do make Level 1 content. This is our “get it out the door and starting to rank” content. We also use AI, but more than AI, we use a whole lot of judgment and experience. 

Over time, Level 1 will perform or not. If it performs, we optimize and uplevel it continually (we’re talking every 6-12 months). It becomes richer and more robust every time we touch it.

Levels

We are an SEO and Content agency, so we’re also always thinking about the SEO distribution channel when we produce. Some ways that manifests:

  • Starting with keywords for most (but not all) of the content we write
  • Using a TF-IDF to enhance our human storytelling content (Surfer is our rec)
  • Analyzing the SERP for guidance on the go-to level for each KW 
  • Helping clients understand the advantages, but also the limitations, of these tools

Even at Level 1, We Go *Way* Beyond the TF-IDF Basics

Could you get AI to write a decent TF-IDF Level 1 piece of content? Maybe. If we only planned on working with clients for six months, we might spin up a bunch of this stuff, publish it, take the cash, and walk away. But we know that’s just creating a different problem for them to solve.

Why? Because tools like Surfer are great, but there are lots of things you need to do with content that Surfer will not recommend or score you on. And this is not just the ‘storytelling’ stuff content marketers bang on about (rightly so, btw). It’s also stuff that matters to Google, and that matters to revenue, too.

These are the things you really want a human adding to your content:

EEAT 

You know the acronym: Expertise, Experience, Authority, Trust. To be honest, they all melt into each other. But the key takeaway is, “Why should I trust this content?” Trust like this can be built in a number of ways:

  • The publication/domain (e.g., people used to trust certain sites for product reviews and recs, but that may be waning)
  • The byline (literally the qualifications of the writer, reviewer)
  • Timeliness (is the content up-to-date)
  • Opinions/expertise incorporated (beyond the writer, is there authority in people quoted and interviewed? We’re much more likely to trust an article about the best hummus recipe if it interviews Yotam Ottolenghi than some random TikToker’s recipe that doesn’t even include tahini—yes, this is a bit of a rant topic for me!)
  • First-hand experience expressed (is there first-hand experience, or am I retelling a story told already on the SERP?)

When we write for you (even at Level 1), we add as much EEAT as possible. We’ll ask your SMEs to review. We’ll do research and incorporate data from reputable sources. We’ll tap into your customer reviews, your webinars, your community, and your data to imbue as much EEAT as possible into every piece. This is all above and beyond what even the best AI prompts deliver.

We’re Obsessed With Outlining (For Very Good Reasons)

In our opinion, outlining isn’t talked about enough. A blog post is not just an assembly of stuff in any order. There should be flow. Points connect and build on top of each other. Even in a listicle, where each item is a standalone point, there’s a relationship between each point, a logical order, and a hierarchy of information.

Let’s compare outlines that might be relevant to one of our favourite clients, Knix. (Bonus: you might learn about women’s health issues, which are shockingly underfunded.)

DS-Blog-Ai-Infogrpahic-1

Think of an outline this way: When you furnish a room, you need a sofa and coffee table, a TV, some side chairs, side tables, lamps, and other accessories. But the room doesn’t come together until you place these items. The right placement can give you a room that feels welcoming and practical. There’s a place to sit that’s comfortable, a place to put your drink. You have a light if you want to read, and you’re oriented to a focal point, whether that’s a TV, fireplace, or other people.

AI just throws the furniture in the room. It gives you the components (or most of them) you need. But you still need to spend time arranging them. We’ve outlined articles using AI where the same point is made three different ways, where the main point is nested between two minor points, and the edge-case explanation is the second item on a list. You need a human to arrange the furniture: to take that outline and make it make sense to the reader, convey the salient information, and thereby drive them to the appropriate takeaway and action.

We Incorporate Your Brand Voice, Tone & Product Positioning

You can teach AI a lot about your brand, your products, and your positioning. But there’s still a lot of nuance in what you’re selling, who you’re selling it to, and the voice and tone that will resonate with them and reflect your brand as a whole. Even in a world where literally every Voice & Tone guideline is some version of “Friendly” and “Straightforward,” there are moments when AI will dramatically miss the mark.

When you use SurferAI, for example, there’s a dropdown menu of tones to choose from. 

  • Tone-Of-Voice

Transparently, we haven’t played with them all because SurferAI credits are not to be played fast and loose with, but we can tell you that none of the ones we’ve tried really nail it. The language is often opaque, repetitive, and fluffy. Metaphors are belaboured and overused to the point of being cringy. Word choice is sometimes try-too-hard-I-got-a-thesaurus-for-Christmas verbose.

Let’s stick with the same example as above and show you just how 😬the language can get with AI:

DS-Blog-Ai-Infogrpahic-3

When we use SurferAI for first drafts, it’s just that: a first draft. The time we’ve saved is spent instead on rewriting and editing. We consider the voice and tone of your brand and of humans in general. We think about sounding smart, real, respectful to the reader, and on brand. We know AI will get better and better at this, but unless you’re willing to invest days in prompt writing, the faster route is still a decent writer.

Most Importantly, We Drive Conversion & Microconversions

Okay. Let’s not forget what we’re all ultimately here for: We need to shift some product, sign up some users, and collect some email addresses. There’s always a conversion, or at least a microconversion.

You can ask AI to incorporate your product, but will it get the nuance right? (Yes, nuance is a big part of this.) Maybe you can get it there with the right prompting and inputs about your product, but will it weave in the right message at the right time? Will it think about microconversions along the journey (connecting to other pieces of content, to communities, to your upcoming webinar)—not just the ultimate conversion? 

This is something we consider a lot when we’re writing at D*S. We think about how to connect a keyword with your brand in a meaningful (and by meaningful, we mean “revenue”) way. We do it in ways that are sometimes subtle and sometimes overt. We also consider what’s going on around the finally published piece (banners, callouts, related link CTAs, etc.) It’s hard even to begin to articulate the rules behind such judgments. 

We Test All the Tools, so You Don’t Have To

Much as we’ve been critical of ChatGPT and SurferAI outputs in this article, they are actually tools we use, recommend, and enjoy using. They’re just not tools we’re hands-off with. Sometimes, they get it more right than wrong, but other times, we intervene and correct A LOT. 

This post isn’t supposed to convince you that human writers are better than AI. (Remember, the grey goo existed before AI.) We certainly don’t think you should ignore AI. And we do think it can offer efficiencies in content production, especially when content teams have the time to invest in learning how to leverage these tools best. 

We just don’t think AI’s good enough to be used hands-off. In fact, we believe that borders on negligence.

AI Risks: Google’s Not Going to Take AI Lying Down, Nor Should You

Maybe you’ve read all this, and you’re just like, “I couldn’t give a sh%t about any of this.” Maybe you have a quarterly number to hit, and believe a bunch of grey goo TF-IDF AI content will be just fine. Maybe you just want to land your bonus. Maybe you think all this sounds like a luxury you can’t afford.

We get it. And you’ll certainly find people willing to help you with that kind of AI content (just not us). But beyond quality, there are reasons for concern, and we’ll just float them out there and let you weigh them up:

  • Plagiarism: We have used AI software and discovered sections of articles directly and verbatim plagiarized from sources outside the SERP. Obviously, this is a potential legal landmine.
  • Misinformation and disinformation: Rely on AI, and you could get it wrong. We use AI every day, and we fact-check it all damn day long. We’ve seen it get legal and medical information wrong. This is not the kind of stuff you want to play fast and loose with.

Bet you’d like an example right now. Let’s learn (or try to) about the role hormones play in the development of uterine fibroids.

DS-Blog-Ai-Infogrpahic-2

  • Algo updates and penalties: As we saw with the content heist above, Google is not rolling over and dying any time soon. While the latest guidelines are tolerant of the use of AI, Google will keep coming after fully AI content that’s produced with the intent to manipulate rankings. This is changing fast, so we won’t cover it in-depth here, but here’s a link to the latest update, covered by Animalz.

So, What Do We Recommend for Smaller Budgets?

It’s really a no-brainer: Quality over quantity. Maybe SEO will not be as important a distribution channel in the years to come. Maybe it will still be key. But people will still need information, and they will prefer good information.

To be clear: This doesn’t mean every blog post needs to be Pulitzer Prize quality (which is also something we encounter), but it does mean that appropriate effort should be put into meeting and exceeding the audience's needs, making content trustworthy, telling the story in a way that’s easy to read, digest and ultimately that encourages people to try your product, subscribe to your newsletter, whatever.

If you’re on a budget, be ruthlessly prioritized about what you spend your money on, go to Level 3, 4, or 5 instead of sticking with Level 1, and consider your distribution channel(s) before you produce content. And keep on playing with AI. It does make many parts of the writing process easier and faster. It just doesn’t get there on its own—yet…