In episode 178 of the Content Strategy Experts podcast, Sarah O’Keefe and Christine Cuellar perform a pulse check on the state of AI as of December 2024. They discuss unresolved complex content problems and share key considerations for entering 2025 and beyond.

The truth that we’re finding our way towards appears to be that you can use AI as a tool and it is very, very good at patterns and synthesis and condensing content. And it is very, very bad at creating useful, accurate, net new content. That appears to be the bottom line as we exit 2024.

— Sarah O’Keefe

Related links:

• 
  
• Pulse check on AI: May, 2024 (podcast)
  
• AI in the content lifecycle (white paper)
  
• The future of AI: structured content is key (webinar)
  
• Savor the season with Scriptorium: Our favorite holiday recipes
  

LinkedIn:

• 
  
• Sarah O’Keefe
  
• Christine Cuellar
  

Transcript:

Christine Cuellar: Welcome to the Content Strategy Experts Podcast brought to you by Scriptorium. Since 1997, Scriptorium has helped companies manage, structure, organize, and distribute content in an efficient way. In this episode, it’s time for another pulse check on AI. So our last check-in was in May, which in AI terms is ancient history, so today, Sarah O’Keefe and I are gonna be talking about what’s changed and how it can affect your content operations. Sarah, welcome to the show.

Sarah O’Keefe: Hey Christine, thanks.

CC: Yeah. So 2024, as we’re currently recording this 2024 is winding down. People are preparing for 2025. Throughout this year, we went to a lot of different conferences and events. Of course, everybody’s talking about AI. So Sarah, based on the events that you like just recently got back from, you finally get to be in your own house. What are your thoughts about what’s going on with AI in the industry right now?

SO: There’s, still a huge topic of conversation. Lots of people are talking about AI, a huge percentage of presentations, you know, had AI in the title or referenced it or talked about it. With that said, it seems like we’re seeing a little more sort of real world, hey, here’s some things we tried, here’s what’s working, here’s what’s not working. 

CC: Mm-hmm.

SO: And I’ll also say that we’re starting to see a really big split between the AI in regulatory environments, which would include the entire EU plus certain kinds of industries and the sort of wild, wild west of we can do anything.

CC: Yeah. So do you feel like it sounds like, know, when AI first came onto the scene, there was mostly, you know, let’s just all adopt this right now. Let’s go for it full steam ahead, especially marketers as a marketer. can I can say that because we’re definitely gung-ho about stuff like that. It sounds like, the perspective has shifted to being more balanced overall. Is that what you would say?

SO: Yeah, I mean, that’s the typical technology adoption curve, right? You know, have your your peak of inflated expectations, and then you have the I think it’s the valley. It’s not the valley of despair, but it’s something like that. But you know, you sort of go from this can do anything. This thing is so cool. Go, go, go, go, go to a more realistic. Okay, what can it actually do? And what you know, does the and this is true for AI or anything else? What can it do? What can’t it do? What does it do well?

CC: Mm.

SO: Where do we need to put some guardrails around it? What are some surprises in terms of things that are and are not working?

CC: Yeah. And at some of the conferences we were at this year, our team had some things to say about AI as well. So we will link some of the recap blog posts we have in the show notes. Sarah, what are some of the things AI can’t do right now? are the still, what are, Sarah, what are some of the big concerns about AI that are still unanswered, unresolved?

SO: So in the big picture, as we’re starting to see people roll out AI-based things in the real world, whether it’s tool sets or content ops or anything else, we’re starting to see some really interesting developments and some really interesting assessments. Number one is that when you look at those little AI snippets that you get now when you do a search and it returns a bunch of search, well, actually it returns a page of ads.

CC: Yes.

SO: And then some real results under the ads. And then above that, it returns an AI overview snippet. So those are surprisingly bad. You do a search on something that you know a little bit of something about and see what you get. And you will see content in there that is just flat wrong. I’m not saying it’s not the best summary. I’m saying it is factually incorrect, right?

CC: Yeah, I hate them right now.

SO: So those are surprisingly bad. And talking about search for a minute, which ties into your question about marketing, there’s some real problems now with SEO, with search engine optimization, because if I’m optimizing my content to be included in an AI overview that is A, wrong, and B, doesn’t actually give me credit, Pre-AI, those snippets that showed up would say, I sourced it from over here.

CC: Mm-hmm.

SO: And in many cases now, the AI overview is just like the sort of summary paragraph with no particular, there’s no citation. It doesn’t say where it came from. So what’s in it for me as a content creator? Why am I creating content that’s going to get taken over by the AI overview and then not lead to people going to my webpage, right? How’s that helped me? 

CC: Yeah. Yeah.

SO: So there’s some real issues there, there’s a move in the direction of thinking about levels of information. So thinking about very superficial information. How much does a cup of flour weigh? That type of thing. That’s just a fact and you can get it pretty much anywhere, we hope. And then there’s deeper information. Why is it better to weigh flour than to measure it? By volume, if you’re a baker.

CC: Yeah.

SO: And what does it look like to use weights? And are there differences among different kinds of flours? And what are some of the things I should consider when I’m going in that direction? So one of those, know, flours, a cup of flour weighs 120, sorry, a cup of all-purpose flour weighs 120 grams is a useful fact. And I don’t know if I really care if people peruse that further or come to my website for more about flour. The deeper information, the more detailed discussion of, you know, whole wheat versus all-purpose versus European flours versus American flours and all these other kinds of things, that requires more in-depth information and that is not so subject to being condensed into an AI summary. So that distinction between, you know, quick and dirty information versus deeper information, information that goes into a topic,

CC: Mm-hmm.

SO: We have a huge problem with disinformation and misinformation with information that is just flat out not either not correct or because of the way AI tools work, is trivially easy to generate content at scale. Tons and tons and tons and tons and tons of content. And because it’s trivially easy,

CC: Mm-hmm.

SO: That means it’s also trivially easy for me to generate, for example, a couple thousand fake reviews for my new product or a couple thousand websites for my fake products. It we can fractionalize down the generation of content. 

CC: Yeah.

SO: And the you know, the interesting part of this is that it implies that you could potentially, you know, we talk about doing A/B testing and marketing. You could do A/B/C/D/E/F/G testing pretty easily because you can generate lots and lots of variants and kind of throw a bunch of stuff against the wall and see what works. But the bad side of this is that you can generate fake news, fake information, fake content that is going to be highly, highly problematic from a content consumer trust point of view. And so that I think is the third piece that we’re looking at now that is going to be critical going forward. And that is information trust, content reputation or the reputation of content creators and credibility. 

CC: Mm-hmm.

SO: So for those of you listening to this podcast, how do you know it’s really us? Do you know these are live humans actually recording this podcast versus you know there’s now the ability to generate synthetic audio and you can create a perfectly plausible podcast which is really hard to say unless probably your AI and then it can probably do it perfectly but our perfectly plausible podcasts are you know how do you know that what that what you’re receiving in terms of content, digital content in particular, is actually trustworthy. And so I think ultimately there’s going to be some, need to be some tooling around verification, around authenticity, around, you know, this was not edited. You know, in the same way that you want to be able to verify that a photo, for example, is an accurate record of what happened when that photo was taken.

CC: Yeah.

SO: And if I went in and photoshopped it and cleaned it up, then that’s something that should be acknowledged. By the way, for the record, we do record these things and we do edit them. We try to stay on the right side of just editing out dumb mistakes and not editing it in a misleading way. 

CC: Yeah, ums and ahs and yeah.

SO: So it’s not like we record the whole thing from soup to nuts and never, you know, never break in and never edit things out because believe me, I’ve said some stuff that needed to be taken away. If you ever get the raw files, they are full of, I didn’t mean to say that. you might want to take that out. 

CC: Me too, so many times. Let me start over, that’s me a lot all the time.

SO: Yeah, sorry. Starting over. OK, but the point is that when we put out a podcast, we are saying this is our opinion, this is our content, and we’re gonna stand behind it. Whereas if it’s synthetic or AI generated or AI generated by these non-humans, you can do these weird, let’s make a podcast out of a blog post, well, okay, but what’s the value of that and why would I trust that content?

CC: Yeah.

SO: So that I think is going to be the big question for the next couple of years is what does it look like to be a content creator in an AI universe and to have the ability or sorry to as the content consumer to have the ability to validate what you’re listening to or reading or seeing.

CC: Yeah. And a point that you had brought up in, I believe it was the white paper that you authored back in 2023. One of the points in there was that, people are going to, because of this trust and credibility issue, people are going to have to start relying on companies and brands that they’re already familiar with for the information that they’re looking for rather than a search from scratch because, you know, search is so messed up right now. And that is something I’ve seen personally, like myself, I do it a lot more. I’ve seen that with friends and other contacts and stuff like that. That’s really what people are doing is they’re going to, you know, the source even for recipes. Recently, as I was looking for a recipe and instead of just Googling it like I used to because I’m so sick of the summarized AI search, I went to all recipes, you know, a place that I knew that I liked the recipes or I think Sally’s baking addictions or something like that. There’s a lot of different places like that that now I’ll just go there instead of, you know, a search from scratch. That’s… I don’t know how we’re gonna fix that problem, yeah, trust and credibility, that’s gonna be a huge one.

SO: It’s a really good example though because if you search for a particular recipe, even say two years ago, you would get a certain set of results and then you would say, I’ve heard of that website and I’ll go there. Now you search on a recipe, I’m getting 20, 30, or 40 websites that I’ve never heard of that all seem to have posted exactly the same recipe.

CC: Mm-hmm.

SO: I, you know, do I trust them? Do I trust them not to be AI-generated? Do I trust them to remember to not, you know, recommend that I put gravel in my recipes? You know, maybe not. And so I’m doing the same thing you are, which is, you know, reverting to trusted sources, trusted brands that I know that have a reputation for producing good recipes. Now, the flip side of this is that content is disappearing. 

CC: Hmm.

SO: So, I have an infamous triple chocolate cookie recipe, is really if you’re looking for a chocolate bar in the form factor of a cookie, that is what it is. It’s just stupid amounts of chocolate. 

CC: Mm-hmm. yes, that sounds amazing.

SO: It’s they’re delicious. And I think we’re putting them in our our holiday post, which may or may not have gone live already. So keep an eye out for that. But here’s the thing. I have the recipe because I got it out of Food & Wine about 20 years ago and I have a paper cut out of it that I wrote, hand wrote Food & Wine 12/01 on. So it was December of 2001 and so I went to Food & Wine. I went searching for this recipe knowing that it was originally published by them. I can’t find it. It is not there.

CC: Hmm. wow.

SO: It is not in their database, or at least it didn’t come up in their database when I searched on the exact name of the recipe. I then searched that exact recipe name, you know, just generally on the Internet, and I found three or four or five different places that had it, but none of them credited where I got it from 20 years ago, which I’m pretty sure is the original, right? Because these are all much more recent sites. So there are digital copies out there floating around, but they are not the original recipe and they didn’t credit the original publisher. Now, I don’t know exactly where Food & Wine got it because all I did was cut out the recipe. didn’t cut out the article. It was probably the context around it. But what I’m now reduced to is that I have a paper copy stashed in my paper recipe book, right? And I took a photo of the paper copy and put it on my phone. So I have a sort of digital version, but it is literally a photograph of a printout, which is, it is 2024 and we are doing photographs of printouts, but I can’t find it or I can’t find the original online.

CC: Yes. Yeah. That’s interesting. Why do you think that content has disappeared? Do you think it’s because of the breakdown of the content model where the AI engine is just eating what it’s already regurgitated a bunch of times? Do you think it’s that? Does an org pulled it for some reason or what do you think is the cause?

SO: Well, I mean, my best guess is that their recipe database only goes back so far and they just said anything more than X years old doesn’t need to be in here. They had some similar recipes. So maybe, well, this one’s been updated. It’s a little more modern, whatever. But it was just, it was really troubling that I, even knowing what the source was, I couldn’t find it.

CC: Yeah, that is troubling. So how can companies prepare knowing that this is our context, this is our landscape? What should we do to prepare for 2025 and beyond? Because it’s not just like next year.

SO: Beyond yeah, okay. So first of all you have to understand your regulatory environment Because that is very different by country or by region the issues that the people in the EU are looking at or American companies that sell in the EU, right. 

CC: Mm-hmm. Yeah.

SO: There’s an EU AI act, and there’s a whole bunch of guidance that goes along with that. So there’s some concerns there. Whereas here in the US specifically, we don’t have a lot of regulation around AI, if any. Mostly we lean on, well, if you put out something that’s incorrect, there’s potentially product liability. If you put out instructions that are wrong and people follow them and they get hurt or worse, then the product owner is probably liable for putting out wrong instructions. That’s kind of where our stuff lands. But as a content consumer, I think you have to do what you’re describing, Christine, and become very, very skeptical about your sources methods, right? Where’d you get this stuff? And do you trust the source that it came from? 

CC: Yes.

SO: If you are a content creator, then looking at questions around AI, the questions become, how can I employ AI inside my content workflows in a responsible way that achieves the goals that I have and doesn’t get me in big trouble in whatever way? And there’s also the question of, if I’m a content creator and I know that my consumers, my customers, are going to be using AI to consume my content, then how do I optimize that for that? How do I prepare for that? So it looks very different if you’re a person writing, creating new content, versus you’re the person deploying a chat bot on your corporate website that’s going to go read through your content corpus versus the person actually using the chat bot versus you name it. So.

CC: Yeah.

SO: And then, you know, we’re talking about AI generally, but of course we have AI tooling and we also have generative AI and we have all sorts of different things going on. So it’s a very, very broad topic, but overall, you know, what’s the problem I’m trying to solve? Can I apply this tool in a useful way? And what are some of the guardrails that I need to employ to keep myself out of trouble?

CC: Yeah, in one of our webinars from this year, from 2024, depending on when you’re listening to this podcast, Kerry Hain mentioned something along the lines of like, you know, when you’re dealing with AI, it’s such a huge topic. You need to break it down by what’s the purpose of what you’re trying to do and then tackle the problem that way. Okay. So to wrap up, Sarah, what are your final thoughts, wishes and or recommendations for the world as we enter this new era? Or I guess we’re in it, but as we try to recover.

SO: So the very short, we’ll try and keep it short. I think when all this AI stuff hit us a year or two ago, business leaders generally were hoping that they could just use AI as a general-purpose solution. Fire all the people, use AI for all the things, cool. 

CC: Mm-hmm.

SO: The truth that we’re grasping towards or finding our way towards appears to be that you can use AI as a tool and it is very, very good at patterns and synthesis and condensing content. And it is very, very bad at creating useful, accurate, net new content. That appears to be the bottom line as we exit 2024.

CC: Yeah. Well, thank you very much for unpacking this with us because I know that, you know, things are changing so fast. It’s helpful to have people like you that have been in the industry, the content industry specifically for a really long time that can help, you know, figure out a way through all this and give some practical ideas.

SO: Well, you know, in six months, we’ll just feed this podcast into the AI and tell it to fix it so that it remains accurate. And off we go.

CC: Yeah, there we go. And then we’re done. 

SO: And we’re done.

CC: Yeah. Thanks so much for being here today and for talking about this.

SO: Yeah, anytime.

CC: And thank you for listening to the Content Strategy Experts Podcast brought to you by Scriptorium. For more information, visit scriptorium.com or check the show notes for relevant links.

The post Pulse check on AI: December, 2024 (podcast) appeared first on Scriptorium.