Episode description

Artificial intelligence has made it easier than ever to produce polished content—but that doesn't mean the information is accurate, useful, or worth reading. In this episode of the RDH Podcast, Jackie Sanders, Andrew Johnston, and Amelia DeStefano discuss the rise of “AI slop” and the warning signs that can help dental professionals distinguish substantive, credible content from writing that simply sounds convincing.

The conversation covers fabricated references, overstated research findings, repetitive AI language patterns, promotional content disguised as education, and the responsibility of both readers and publishers to verify information. The hosts also discuss why dental hygienists' firsthand knowledge and expertise remain valuable—and why aspiring writers shouldn't be afraid to put their own ideas on the page.

Additional reading: Clinical Tips: Understanding AI in dental hygiene practice

AI slop, smart readers, and the value of human expertise

Jackie Sanders: Welcome back, everyone, and thank you for joining us. My name is Jackie Sanders, chief editor of RDH Magazine. Today on the RDH Podcast, I am joined by my co-host Andrew and my partner in crime, Amelia DeStefano.

Hello, you guys. How are you today?

Andrew Johnston: Howdy, howdy, doing good.

Amelia DeStefano: Hi.

Jackie Sanders: This is kind of a bittersweet moment because we are slowly but surely wrapping up the RDH Podcast. And what better way than to hang out with two of my favorite colleagues who helped me work the magic at RDH?

Andrew, do you want to jump in here and tell us what we're going to talk about today?

What is AI slop?

Andrew Johnston: Yeah, I mean, I think one of the things that we, all three of us, definitely encounter is the submissions that we get from writers. Maybe they've been established writers, or they're newly minted writers or up and coming, and they submit some content to us that looks just a little bit janky.

There's just something that's a little bit off, and you can tell it's either been lightly edited by an AI tool or it's been completely written by an AI tool.

Probably the three of us, Amelia has one of the best eyes for kind of detecting what is termed AI slop. And so that's what we're gonna talk about today. We're gonna talk about how dental professionals can be smart readers in the age of AI slop.

I'm really excited about this. Should we get going on the first question?

Andrew Johnston: And guess what, listeners? We're actually gonna ask the questions.

Jackie Sanders: This is absolutely a first.

Andrew Johnston: I know. We're not going to go down any rabbit holes this time. All ready?

Andrew Johnston: I'm locked in.

Amelia DeStefano: I have one question.

Jackie Sanders: What was the word you used? Swanky, slanky, cranky. What was janky? Does janky equal slop? Are those synonymous for each other? Could you use both of those?

Andrew Johnston: I think janky is the—I think there's something not quite right with it.

Jackie Sanders: Okay, not quite right. Okay, ask your first question, Andrew.

Andrew Johnston: Okay, so I think the place that we need to start is just kind of definition. So, Amelia, what have you seen, and how would you define AI slop?

And then, kind of as a follow-up part to that, because you've been doing this for a little bit, when did you start seeing this kind of creep in? Has it been getting worse, or is it getting better? Like, where are we at on the whole timeline of this?

Amelia DeStefano: Yeah. So AI slop is a term that gets thrown around a lot now. I don't know if there's a specific definition, but slop is definitely not content that's just been maybe touched by AI. It's something that is either significantly edited or completely generated by AI.

Maybe if you use a really good prompt or something, you could get some output that does have a little bit of quality to begin with. But for slop, it's like you just kind of do a quick prompt and put it out there as if it was something that you wrote.

AI writing is always really polished. It doesn't have typos or grammatical errors, but once you kind of start to poke at it a little bit or maybe question some of the sentences, you're like, there isn't really anything here. It doesn't actually make sense.

It sounds like a coherent sentence, but once you start to dig into it, it's not.

I saw something on LinkedIn that's kind of stuck with me. Someone said that if you die a little on the inside when you read it, it was probably AI-generated. That's pretty harsh, but it definitely stuck in my mind that that's what it is.

I think this really picked up about two years ago. In the beginning, we tried different ways to sort of recognize when something that we were looking at publishing or that we saw online was AI-generated. We tried using AI content detectors, which have turned out to sometimes not be that reliable.

So we've especially just moved to looking at the quality of the writing and the information. I think humans are really good at pattern recognition. You can kind of know that, okay, I see this in every single ChatGPT-written article that I've ever looked at, so this is probably AI-written.

Sometimes there can be what people might call hallucinations. It'll just make up a clinical study that doesn't actually exist once you go check it.

I don't know. You could compare it maybe with wood furniture. A dining table made out of oak feels really different than one made out of particle board, and AI slop is definitely the particle board of the situation right now.

When polished writing has no substance

Jackie Sanders: Yeah. Well, I was just going to say that explains some of the things that we have received over the last 12 months. We receive content that we can't find the references that are used in it, or, as you said, the content almost reads robotic.

Especially if it's a writer that we've known for a while, and then they just—what I tell our writers is, even if you're just doing a grammar check, AI is really bad about taking credit for everything it just read and grammar checked. And then that's where we were getting some of those results that were saying it was high in AI content.

Andrew Johnston: I think, at the risk of—you know how you don't want to highlight the bank robber, right? Because then they give them all the tricks to robbing a bank and stuff like that. So you don't want to talk about that kind of stuff.

I kind of feel like that here. I don't want to necessarily give away all of the patterns or the things that I'm seeing, where, like, the thing that makes me stop and question, because I just don't want people to start trying to defeat that and then I have to find new patterns and new ways.

But I will say, you're right that humans are really good at recognizing those patterns. There's a lot of, "Well, it's not this, it's actually this." Or the word "actually" is used a whole bunch of times.

Amelia DeStefano: Or "quietly," that word.

Andrew Johnston: Oh my gosh. There are certain ones like that we just kind of have in our back pocket that we're like, okay, well, this is obviously garbage.

But at the same time, I know I've been fooled. I absolutely had to have been fooled. I don't think that there's a way that we couldn't.

You even look at—I think this might be more of a generational thing than anything—but they have those videos where people are like, "Oh, is this AI or is this not?" Half the population gets it right, and half the other population thinks that this is like the best, most smartest little puppy that knows how to ride a tricycle.

You know, it's just like, no, the puppy doesn't know how to ride a tricycle. That's not real.

Jackie Sanders: It's like the puppy's left foot looks like a frog.

Andrew Johnston: Exactly. It's like, what is a frog's foot doing on that puppy? It was such a cute puppy.

I guess where I'm going with this is the readers are kind of having an uphill battle facing the same kind of thing that we encounter as we're reading it.

Are there things or signs that a reader could look for that would make them stop and be like, "This probably is not something that's credible," or, "I just need to double-check this"? Like, what should they be looking for?

What should readers look for?

Amelia DeStefano: That's a really good question. I would say, just to walk it back a little bit, I know that you all share this passion that when we publish something for our audience, we think that they deserve content that is of really high quality and helps them do their jobs better.

I sometimes use AI tools in generating ideas or checking something, but we don't want to put out something that had really limited human involvement and has a lot of factual errors because our audience is taking care of people and making important decisions. I think that's something that we take really seriously.

I would say look for maybe the volume of content that might be coming from somewhere, because good writing takes time. If you see, maybe we'll just say it's an influencer online who's talking about clinical stuff and they have a new article every single day or two new articles, that might not necessarily be evidence that something's wrong, but I'll definitely take a look and say, like, how many research-based papers could one person put out in a week?

Additional reading: How artificial intelligence is changing periodontal assessment: Enhancing diagnostic accuracy in dental hygiene practice

I would also look for whether someone is suddenly writing outside their area of expertise. If someone's always written about anesthesia, for example, and all of a sudden they have all these thoughts about biofilm removal or something, maybe that would have been outside their original expertise. You might ask, okay, I know this person is really experienced in X, but do they actually have the qualifications to be talking about Y?

You can look for those patterns, those language patterns that are probably really familiar to us at this point.

I've found a lot of times that AI is not good at nuance. It tends to overstate study findings sometimes. And then if you push back on it, for example, if you're using a chatbot or something, it'll say, like, "Oh, I'm sorry, that's my bad. I should have actually said this."

Then you realize that you're not talking to a person. You're just talking to a language generation model.

I think probably this is just more of an overall content thing, but when it comes to any kind of articles that are talking about products, if you feel like all of the content is sort of funneling you toward buying or using a product, that might be kind of a red flag that maybe the point of this isn't to be educational.

That's what I've noticed or what people could look for. What about you all?

Andrew Johnston: No, it's a good point. I mean, just like the intention behind the piece, right? Who, what, when, where, why—asking the questions behind why this was created and what it is leading me toward.

I'm right there with you. I don't know if I actually have that much more to add as far as giving them moments of, "Oh, this is when it's always this, then you need to be on the lookout."

Trusted sources, I think. Just taking a step back into the Googling days, you're not going to go down to the 18th source and quote them. You're going to look at the ones that got filtered up that are not the sponsored content ones, but the links that are just down there a little bit more because, from what I understand, the model before was they pushed things toward the top that most people found helpful and useful.

Obviously, going to your PubMed and your kind of research-based search engines and things like that are always probably good places to start.

I just think it's gonna be a rough one. I don't know, Jackie, if you have anything else to add to that.

Why authentic expertise still matters

Jackie Sanders: The only thing I would really add is authentic content is just so valuable and more read because it's personable.

What I'm afraid of is all of these new young authors who are eager to write something for us are thinking that they have to use AI to generate a quality product. That's not the case at all.

I tell all these new authors, put your own thoughts out there. Put your own ideas, your own research. We have an amazing editorial team that will polish it up and pull it together properly.

But you're the one with all of the dental and dental hygiene knowledge, and you know what you're trying to say more than AI knows. Just let the editors be the ones that say, "Let's use this word instead of that word. Let's put the period here instead of there."

And just lean into the own massive amount of knowledge that you have in your head from working with patients and going to hygiene school. You don't need AI.

Amelia DeStefano: I think editors have this reputation of being grammar Nazis, for lack of a better word, but it's not really true. I know, especially now, every editor would rather get something that has substance and then maybe isn't perfect structurally or grammatically. That's something that you can fix.

AI writing is kind of the reverse. It looks perfect, but there's nothing there. And once you try to start fixing it, you just really can't.

So a lot of times, we just decide not to publish these things because it's just not worth our audience's time.

I think as it becomes more and more common, people will be looking for content that sounds like a human being because that's what we want.

Andrew Johnston: I really like when people submit mostly just an idea of the content. Don't write it completely out. Don't finish the narrative of it. Tell me what you're trying to write and then let me help you tell the story. Let's create something together.

We've gotten lots of pieces where we're just like, hey, there's something here, but we have to do a complete rewrite because it just does not sound good or flow or whatever.

So, I mean, Jackie, right? Let the editors. There are people in this world who will help you do the thing. You don't always have to use AI.

I also don't think that you have to be completely ready to—you don't have to be the expert in the field necessarily. You have to be the expert in the one piece of content that you're trying to push out. I think that's really important too.

Is AI creating a new version of an old problem?

Andrew Johnston: But I guess I have a question for both of you on this one. The background of this question is influencers, social media influencers and stuff.

I've kind of always had a problem with social media. It's just, in general, I think it's more of a personal Andrew problem than it is anything else. But it's because these people get on and they try to be experts in, like, literally everything.

So, Jackie, you're talking about the new writers that are coming in. I'm wondering, is AI enabling—or is this the same problem that we had with our content creators and influencers or clickbait-type material that was going out before?

Or is this a new problem that we're going to have when we have AI-generated content? Is it kind of mirroring the same thing, or is it two different problems?

Amelia DeStefano: Yeah, I mean, it's like it took everything because weak writing obviously has always been a problem, but it just took everything and inflated it 1,000 times because now you don't even need to spend time writing a bad article. You can just have ChatGPT do it for you in five minutes.

I think it can make people feel like they're more knowledgeable on something than they really are.

I remember when I started working in this field about 10 years ago, we had a much bigger problem with plagiarism. People were just outright stealing things that other people had written.

I haven't seen an actual classic case of that since AI became heavily used because now people don't need to do that. They can just use ChatGPT to make something new from what the other person wrote.

Andrew Johnston: That's an interesting new problem though, isn't that?

Amelia DeStefano: Yeah. I mean, I hate that people feel like they have to steal each other's work because I know in this profession, there's so many ways to be knowledgeable that people really can carve out something for themselves.

Jackie Sanders: Yeah, and I think there are the individuals who do take other people's work. I think there are individuals that aren't knowledgeable enough in writing that they don't realize they can't use somebody else's work without recognizing the other person for their work.

I've run into that before. When we do push back, somebody will—this has probably happened to me, I don't know, half a dozen times—somebody I've never heard of before will reach out to me and say, "Would you need an article on this topic?"

That's a great topic. I'd love it.

And within 45 minutes, I have an article come back from that person.

It's like a red flag, but what it's doing is giving me some really good article topics to share with people that do not actually know how to write. So people, just keep sending me those ideas.

But it's like a whole new array of problems.

Those are usually people that aren't even in the dental industry. They have no credentials. They just, "I read your editor's note. This would be a great topic for an article."

And I'm like, "Yeah, go ahead, write it."

Okay. Forty-five minutes later, here's your article.

Andrew Johnston: It's like, no, I'm not going to say that we do that at DentistryIQ at all. However, there may or may not have been one on a very specific topic done like, "This is a great angle for this. I've never even thought of this angle before."

But it came from a PR firm, people I don't even know. I'm like, "Oh, but I have a specialist who can write on this."

I mean, hypothetically, that may or may not have happened.

Jackie Sanders: But that's just one reason to read all the spam email you get.

Andrew Johnston: It really is.

Jackie Sanders: I mean, a new reason to read spam. It's great ideas.

Andrew Johnston: It's amazing.

Who is responsible for verifying what we read?

Andrew Johnston: Well, I think probably the last question that we have, Amelia, for you—but also Jackie, I think that you have some ownership in this area as well—is that I personally put a lot of responsibility on the audience.

I think that we always have, throughout A Tale of Two Hygienists and all that kind of stuff, is you as the audience member, you as the person that's in the CE course, you, the consumer of this content, have a responsibility to play detective and make sure that the stuff that's coming into your ears and into your eyes is factual and accurate and whatever.

That could just be verifying the source, or that speaker was Tom Viola, a pharmacologist that knows stuff about stuff because they went to school and they kept learning.

Or it could be the piece of content you just go and verify through other sources later on. Going back to your anesthesia analogy, just verifying that the right dosages are—this is the maximum safe dose for this, for lidocaine or whatever. These are easily verifiable.

So I put a lot of responsibility on the audience member. However, what responsibility do you feel like a publication like RDH Magazine, or educators, professional organizations, probably even our writers—what responsibility do they have to make sure that they're handling this correctly?

The responsibility of publishers and educators

Amelia DeStefano: Yeah, that's a really good question. I think something that we've kind of lost in this world, even before AI-generated content was common, was that if you have a publication like RDH Magazine or Dental Economics or a newspaper or whatever it was, you had some kind of expectation that the content that was being published there had been vetted.

We're not peer-reviewed in a scientific textbook kind of sense, but we do have quality control to say, "This is not something that we want to publish," or, "This needs more work."

Now that the bar for publishing something is really low, I think that's kind of an expectation that has been lost. Especially now, you might feel like, "I have to question everything that I read online," which is really exhausting.

I wonder what the long-term effect of that will be. Will people stop just casually reading online? I don't know.

But as far as a publication like us, I think we have a responsibility to verify claims and citations that appear in our articles. We have a responsibility for editorial content to be editorial and sponsored content to be clearly labeled so that we're not misleading people if there might be a conflict of interest in, for example, the author or something.

I think we also have a responsibility sometimes—and this is another thing that has sort of become a lost art now—to preserve uncertainty.

Sometimes a study, we want it to be this big revolutionary thing, but when you start to dig into it, it's like, well, here are all the limitations, and maybe it didn't really find the things that either the PR is saying that it found, or that you would like to have found.

I think also one thing that we've done well is resist the pressure to just publish more because we could. We could just use AI to account for things, and we don't because, like we were talking about earlier, I think that anything that we do publish, we really want to be worth our audience's time.

Jackie Sanders: I think the only thing I really had to add to all of that is, I've been in this role for seven years, and when you're putting out content from dental professionals to dental professionals, you're publishing knowledgeable content to a reader who is also knowledgeable, if that makes sense.

In the seven years that I've been doing this with you, Amelia, I don't know that we maybe once or twice had an article that somebody has come back and challenged us on.

So we're working with really good people. Our editors make sure and check the references.

I'm proud of the content that RDH and DE put out because I know that it is being—we do, which our audience may not know, we do have—

Shoot, what is her title? What is Kirsten's title?

A technical editor.

So we do have an individual who does go through a lot of our content that is heavily researched and checks it all out for us.

That's always reassuring when she does come back and say, "We might want to check this. We might want to change this word to that word."

But we're doing our due diligence to put out the best content we can possibly put out. I'm proud of the hard work that we put into our content.

Andrew Johnston: You should be. It's solid, man. Everyone loves it.

A final message for aspiring writers

Jackie Sanders: So, Amelia, wrapping this up, if we do have some dental hygienists, dentists, dental professionals out there that want to submit content to us, how can they reach you? What is your email?

Amelia DeStefano: Yeah, you can email me at [email protected]. My email is also on our website, on the submission guidelines.

The submission guidelines also offer a lot of helpful information if you are interested in contributing.

Jackie Sanders: Okay, thank you.

Jackie Sanders: Well, for our audience, the RDH Podcast is wrapping up, but we're not leaving you completely. We are leaning into A Tale of Two Hygienists podcast with Jessica Atkinson and David Torres, and they will be leading and carrying the reins from here on out.

Andrew, do you have anything you want to add to that?

Andrew Johnston: No, I mean, just I think that kind of wrapping everything up, you know, podcasting and AI is really hard to do.

So you can trust Jessica and David to bring you some really good content with some really good guests. Make sure you're checking them out too.

Jackie Sanders: That being said, everybody, take care. Thank you for spending time with Andrew, Amelia, and myself. We hope to see you soon.

Andrew Johnston: See you.

Additional reading: 9 things every dental hygienist should know about dental AI