Written By Jon Paget.
Leveraging AI in marketing
Inspired by a number of conversations among the team, we recently held a webinar on whether AI – and specifically ChatGPT – could, and should, be used in marketing. Jon Paget and Bethan Vincent from Open Velocity were joined by Katie Thompson of Katie Lingo to discuss just 4 questions:
1) What are our initial thoughts on ChatGPT?
2) How might it be used in marketing?
3) What are its limitations?
4) Where might it go from here?
You’ll find the answers to these questions in the recording of the webinar below, along with a list of what we feel are some of the best articles on the topic.
Interested in our events and future webinars? Want to speak to us about how we might be able to help with your marketing challenges? Drop us a note to email@example.com
Articles that have helped shaped our current thinking about ChatGPT
https://www.forbes.com/sites/bernardmarr/2022/12/21/chatgpt-everything-you-really-need-to-know-in-simple-terms/?sh=f677b4ccbca3 – a good all-round introduction
https://www.smartinsights.com/managing-digital-marketing/marketing-innovation/the-best-prompts-for-using-chatgpt-for-digital-marketing/ – an interesting article that mentions Prompt engineering
https://www.nvidia.com/en-gb/geforce/news/jan-2023-nvidia-broadcast-update/ – real time eye tracking and video editing using AI (watch the video to see how it works)
https://www.searchenginejournal.com/google-e-e-a-t-how-to-demonstrate-first-hand-experience/474446/ – asks the questions, can machines have first hand experience?
https://developers.google.com/search/blog/2022/12/google-raters-guidelines-e-e-a-t – Google reacts to AI more broadly by adding an extra ‘E’ to its EAT update, focusing in on both Experience and Expertise.
https://www.vice.com/en/article/93a4qe/conservatives-panicking-about-ai-bias-years-too-late-think-chatgpt-has-gone-woke – an article by Vice on the inherent bias of AI tools like ChatGPT
https://openai.com/dall-e-2/ – OpenAI’s image-based program that Jon mentions in the webinar
Social media posts we like
Jon Paget 0:09
Okay, hello, very warm welcome everybody to our leveraging AI webinar. We are going to be packing a lot into the next 30 minutes. And we’re not going to be hanging around. So I’m going to jump straight into some introductions and a bit of housekeeping. So my name is Jon Paget. I’m senior partner at Open Velocity, a marketing consultancy, joined by Bethan Vincent who is managing partner of Open Velocity as well as Katie Thompson, who is managing director of Katie Lingo, a marketing services agency.
So we are going to be talking about AI in marketing, and specifically about chatGPT. As a bit of background, as I’m sure everybody will know, chatGPT is a chatbot that’s powered by artificial intelligence. It’s been around since the end of November 2022. And it’s caused a bit of a stir, everybody, in fact, is talking about it. There’s some big numbers attached to it. So the company behind it is reportedly worth $29 billion. It’s got some big backers in the sense of Microsoft investing reportedly $1 billion And in the first five days, it had 1 million users.
So this led us to have a conversation around its use in marketing, and whether it was all hype, or whether there would actually be some practical uses for it. And so that’s what our ambition is over the next 30 minutes or so. I should add, if you’ve got any questions, please pop them in the chat. If we have time, which might be ambitious, we’ll try to come to them. If we don’t, everything that we do today is going to be recorded and made available on the Open Velocity website next week. And we will endeavour to answer the questions when we pop the content up on there.
Okay, so hopefully with that understood as a bit of context, Bethan, I’m going to come to you with the first question. So what are your first impressions?
Bethan Vincent 2:15
Yeah, great question. So I was probably quite an early adopter of AI content generators in general. So even before chat GPT I was playing around with chatbots and content generators; things like Jasper copy AI. And that was really just because I was interested in it. I was like, oh, maybe this can, you know, speed up content creation, you know, do things like keyword research faster, all of that kind of stuff.
I think maybe this analogy works, ChatGPT is kinda like junk food on the surface, you’re kind of like, Ooh, this is good. Like, this is great. This is producing some stuff, you know, I can kind of see how I can use it. But actually, when you dig into it further, you realise the limitations of it, and it doesn’t leave you fully satisfied. So for example, you know, a lot of people are using it in terms of creating fully fledged blog posts. So they might do that through one prompt or a series of prompts chain together to kind of expand on things as it goes. And the ultimate result is kind of, okay. It’s like a lot of kind of mass produced SEO content. It answers questions in a very surface level kind of way. But it’s got no depth. And we’ll probably come on to this in a little bit around Google’s treatment of AI generated content and some thoughts around what’s going to happen there. But I think this kind of surface level, junk food nature is going to be an issue. It does require human intervention on top of it. We can’t just rely on it out of the box.
Jon Paget 3:37
Katie, what are your initial impressions?
Katie Taylor-Thompson 3:41
Well, I love that junk food analogy there Bethan. If I was talking to my husband about it last night. And I said, well, you know, we can look at any technology like a microwave, I could microwave lasagna, or I could cook it from scratch, and it’ll be delicious. I’ll be full either way. But is it a better experience? And I think yeah, and there are a lot of parallels there.
Obviously, as a copywriter, I’m going to feel very strongly about it, I’m going to feel very emotional about it. I feel like you can’t really be on the fence about ChatGPT. And, just because it’s everywhere everyone is talking about at the moment, we’re probably all sick of it.
But I think as with any new technology, there’s this kind of sense of trepidation, but also excitement at the same time. And I’m really seeing strong polarised views. From a technical standpoint, loads of people say how much time it saves them, how you can generate code and similar stuff. But then using the copywriter’s perspective, which is saying you can’t replace humans and it can’t give you those creative outputs, those personal touches. But I think realistically, we can’t pretend it doesn’t exist.
You know, you can look at things like let’s say Clubhouse that was the next big thing. Voice Search. That’s the next big thing. Some things are the next big thing maybe like Tick Tock but not so much things like Clubhouse so I think we have to acknowledge that it’s going to be part of our world now.
As copywriters we’re going to have to use it to our advantage rather than just pretending it doesn’t exist. But I’m seeing a lot of opinions on it. And there are other considerations too, like ethics and fact checking – stuff like that. So I think there is a long way to go before we can fully use it to our advantage, but don’t be threatened by it, I guess would be my overall summary.
Jon Paget 5:25
Great – I’m a bit conscious that for anybody tuning in who hasn’t used, it may not be obvious as to just how simple it is to use. It feels like such a step change in technology, because it’s the first chatbot that allows you to talk to a computer, in the same way that we might be talking to each other on WhatsApp. And the quality of the response that quite often comes back is impressive. But I think it’s that ease of use which enables everyone to interact with it, which perhaps is why everyone is talking about it.
This feels like an evolutionary step rather than a revolutionary one. But this is crossing over from things powered by AI to tools that are very much in your face, that you can interact with, in a way that we haven’t had before, certainly at a sophistication level that we haven’t had before. And I think that’s what’s got everybody really jumping up and down.
I guess as a segue into our next question, then in terms of practical marketing uses, what can we actually use it for? Katie, let me ask you first.
Katie Taylor-Thompson 7:05
Yeah, sure. So I actually think, in terms of marketing, in my head, I sort of separate marketing and sort of creative and the technical side of things. And again, I would say this as a defensive copywriter. I think it’s much better for the technical output. So you know, I’ve seen a lot of use cases of LinkedIn, people even who are absolutely diametrically opposed to it saying, Oh, actually, it has been really helpful for writing things like schema markup that’s gone from 10 hours to 10 minutes, and things like that.
In terms of writing codeI have no idea. But I know, my husband is a coder. And he’s said that he’d use it before to write things like regex statements, which we’ve then used to kind of plug into like a lead generation tool. So I think for more sort of stuff that we want to automate anyway, that is just generally quite dull and very time consuming, and probably quite costly, then it can be really helpful. It’s kind of a stepping stone to creating bigger, better things.Other examples are people using things like V lookups, for SEO. So certainly for the technical side of things, maybe for some creative things that aren’t necessarily going to excite readers, for example, meta descriptions, product descriptions, I’d be a little bit cautious with them, particularly with product descriptions, you know. But I think, well, it’s probably better just for lots of, you know, small, quick tasks, rather than anything that requires a kind of creative inspiration.
Jon Paget 8:45
I guess, Bethan, that’s what we’ve certainly been talking a lot about. Right. I think we would all agree with that, for the more technical side of the content, this could be really useful. It’s the more contentious area of content creation where it gets murky. I mean, at the basic level, subject lines, blog titles, but also then, actually, paragraphs of copy that can be used. And I know, you’ve got some strong views on that. So what’s your view on that?
Bethan Vincent 9:16
Yeah, I mean, again, like it’s back to the junk food thing, actually, you know, I used it for the title of this webinar. I’m not saying the title of this webinar is the most creative title ever. There’s definitely more that could be done there. But it was a rapid way for me of trying it out – but for me the quality wasn’t quite there. I think if I spent, you know, a good couple of hours thinking about it, I personally could have come up with something better. And I think it’s a really interesting question from a creative point of view. Also, I don’t know if anyone else struggles with this, but I’m a writer who loves having written and I hate writing. I love the end result, I hate the process. And the worst thing for me is a blank page. And actually, where I found it helps me from a creative point of view is not actually writing your content. But is that initial idea generation? So giving it a prompt something like, you know, what are five different ways you could think about lead generation in this particular vertical, then it will spit some stuff out. And I’m applying my knowledge, my expertise, my kind of contextualization of what it spits out and building something off of that. So, I guess we’re using it as building blocks.
And we’re using it to automate stuff that people were already automating through scripts, it just allows it’s almost like a no code tool in a way because it’s allowing you through text instead of a graphical interface to build code. But from a creative point of view, it can just do the building blocks, that’s really where I see it. I would be extremely uncomfortable, for example, producing a piece of content for a client or giving the client a piece of content that has been purely written by it instead of using it just as inspiration and maybe leveraging it in that initial ideation phase. Okay, if I were a client, I would be running it through an AI detection tool, to be perfectly honest.
Jon Paget 11:09
We are dangerously entering territory of limitations here, which is exactly the topic that we’re about to come on to. But before we do that, I wanted to jump out of chat GPT territory and stay in AI territory, which is to mention Dall:e, which is the image based AI that OpenAI launched before they released ChatGPT and this for me as a marketer is really exciting.
Quite often, for me, the marketing challenge whether it’s putting a web page together or building an email is actually the images being more of a headache than the copy. And the idea that AI might be able to help generate images that I’ve got permission to use, images that exactly fit the description of the content that I want to create. That is really exciting. And I think, again, I’m going to jump ahead and say that I think Dall:e combined with AI like ChatGPT is a really interesting space. But I’m going to leave that for further down the track.
So before we get any further, and I maybe should just say, if you’re watching this, and you haven’t had a look at Dall:e, or ChatGPT, if you can get onto it, definitely go and take a look. It’s the best way to experiment.
So let’s come on to the limitations. So we’ve touched on it there Bethan with you talking a little bit about that. Do you want to maybe pick up on what you see as the limitations and particularly perhaps mention Google.
Bethan Vincent 12:55
Yeah. And so what’s about to happen is the internet is going to be flooded with even more mediocre content, because the barriers to producing and the speed of producing content has been massively lowered. Right. And Google, you know, rightfully so is going to, you know, have to have some way of combating this because, especially from a fact checking point of view, that’s going to be really important. And you’ll notice they recently updated the webmaster guidelines for the first time in quite a long time. And they’ve also introduced the extra E in their EAT update. So they’re optimising for both Experience and Expertise now, and I think they’ve done that as a preemptive measure. In fact, it’s not even pre emptive, because the AI kind of content revolution is here. Right? But they’ve done that, obviously, to have a measure of, you know, what authority what, what kind of extra contextual information is this content covering? You know, and I think the challenge is that producing good content to a standard that is adding something substantial to the user new knowledge, new information, new contextualization, that’s a really high barrier. And not many businesses were doing that anyway, you know, they were producing this very surface level content, often just for SEO – a headline, keyword stuff, quite thin content. And actually, whether it’s produced by a human or produced by an AI, that style of content is going to be penalised.
In my view, the bar has just been raised, frankly, and that’s quite exciting. Because I think from you know, as a user of the internet, what’s disappointing is when you’ve got a question and you Google it and the top 10 answers are all these SEO kind of listicle style pieces, and you’re like, but this is giving me nothing, you know, this is giving me no extra information. And, there’s talk about obviously being or incorporating it into their search functionality.
I think In terms of limitations, it exposes a limitation in content and the way content has been produced in the past. Hopefully that kind of makes sense. That’s quite a rambling argument. But I think what I’m trying to say is basically, that it’s not necessarily fundamentally changed the game. It’s just sped up and accelerated. A lot of things were happening already. And a lot of kinds of ways Google was trying to understand quality content. They’ve had to adapt to it. But I think they were going to do that anyway, to be perfectly honest.
Jon Paget 15:32
Yeah. And that feels like a really important point. Because I think there is lots of really bad content out there already. Right? We don’t need ChatGPT to help with that. And so actually, our abilities to find and to curate the best sources of information is just going to become even more important, and whether that’s we do that personally through the people that we trust and the authoritative sources that we know. Or if it’s Google, improving its ability to sort through the rubbish and the clutter and surface the really good stuff. That’s clearly going to be a challenge.
Katie, what are your thoughts?
Katie Taylor-Thompson 16:15
Just to kind of build on what Bethan said obviously, with all the Google considerations as also the helpful content update, one of the things that Google cracks down on is non unique content, I think some of the examples they talk about is sort of movie reviews. If it’s just the same thing, you’re not getting new content. And that’s essentially not helping users. And obviously, any AI content is built on stuff that is already out there. And a huge floor within chat GPT and open AI is that its data essentially out to date. By its own admission, it’s based on information up to June 2021. So today, for example, I was playing with it for the first time, I could get on it in ages. And I was asking questions about my favourite singer, God rest his soul Meatloaf. And I was like, please tell me about him. And it was like meatloaf is this many years old and blah, blah, but meatloaf died last year. And if you ask about Queen Elizabeth, who will say Queen Elizabeth is this age, it will say, you know, Donald Trump is president or whatever. It’s not up to date. So if you want to write about anything that’s happened after June 2021, it’s not going to help you whatsoever. So again, with the fact checking, that’s a real issue, particularly if we’re thinking about stuff like your money or your life content.
Also, just things like ethical considerations as well. This is something I talked about a lot at Brighton SEO in October, which was before ChatGPT even came out, but it was just like AI will take content writers’ jobs. And there are ethical considerations like the biases, things like auto suggesting if I search on LinkedIn for Stephanie says, Did you mean Steven, or like Google word mapping studies say that woman is homemaker and man is computer programmer and stuff like that? Because it’s all built on human inputs, garbage in garbage out. So it’s going to have those inherent biases. And I think, well, in a world where we’re trying to be more inclusive, what kind of output we’re going to get if it’s if it’s been trained on biased inputs. There’s a lot to think about there’s so much to unpack
Jon Paget 18:24
I’ll tip my hat here to Dr. Dave Chaffey, who I think among many others has been writing some really good content on ChatGPT. He wrote a post recently and he was talking about prompt engineering. And Bethan touched on this earlier about the sense that it’s just like any other tool, right? When you learn to use Google, you learn the keywords that you’ve got to put in to get the kinds of responses that you went out. And that’s effectively what people are starting to do with ChatGPT, by understanding its limitations in order to use it most effectively as a tool. And I think that that’s a really interesting way of looking at it. And though he didn’t describe it in this way, it very much felt like he was briefing it,, as you would an agency or a freelancer; he effectively made seven or eight iterations of the content he’d entered in and created a brief for it.
And I think that if you are going to use it and get round its limitations. You’ve got to be prepared to invest a bit of time to understand what its limitations are and Katie, I totally agree, on anything that’s written in the last couple of years, it’s clearly not going to be able to help and it’s going to be just flat out wrong. And then to be able to take the time to be able to massage it into the kind of content that you want at the end of it by all these prompts and these little nudges to try to engineer the solution. which then begs the question, what’s the point? Why not use a person in the first place?
Bethan Vincent 20:06
Exactly, yeah. And that, you know, some of the best things I’ve read on the internet are very clearly written from someone’s personal point of view. They’re in their style, you know, there is a joy to good writing. And I won’t go on my rant about how, you know, we’ve lost our focus on literature as a society, that’s a whole different conversation for anyone who wants to go to a pub, for a pint with me. But you know, that’s the thing, writing is a craft at the end of the day and doing it well is engaging us and building a cohesive argument. Because that’s especially when you’re wanting marketing content to be…is persuasive. Right? That’s what we’re trying to do. We’re trying to nudge and persuade and doing that yourself is going to be much more effective than a machine, in my opinion.
Katie Taylor-Thompson 20:54
Absolutely, that storytelling element isn’t there. And I used jasper AI version in an experiment for an article I was writing. It was about marathon running. And obviously, I’m gonna drop the fact that I’ve run six marathons here, but when I scored my content against the AI’s, my one came out much higher than the AI one.
Jon Paget 21:39
Right, time is whizzing by and I’m really interested in the last question, which is, where do we go from here? So a bit of context before maybe Beth and I’ll come to you, which is that there are rumours circulating that an updated version of chat GPT is going to be out later this year. So we can consider if you’re lucky enough to get on and be able to use it, you’re actively contributing to its future efficacy. It’s an experiment. And they’re learning from that. So we’re in effect, all helping to improve ChatGPT for the next iteration. Also, some really interesting comments from Sam Altman, the CEO of OpenAI, who’s essentially said that it’s free currently, but it’s not going to be free forever. The costs to just keep it running on the web, are going to be eye watering. And so there’s going to be some kind of commercial model that’s going to come in that opens up a whole swathe of different questions, which we won’t get to today, but I guess, Bethan, where do you think this might go? And is it going to become more useful for us as marketers in the future?
Bethan Vincent 22:44
Yeah, interesting question. So if you believe the rumours it’s going to be priced at $49 per month for individual user access, which is an interesting price point, right? It’s probably too expensive for a kid to use it for their homework, so it’s priced at a business user. And I think at that price, I would purchase it for our business, just for the very fact that it is quite a useful sounding board to throw some ideas at and get something out, I find value in that.
So I think it’s not going away, people will pay for it, they’re clearly going to make money out of it. I think commercially, and I alluded to this already, if you’re working with copywriters at the moment, I would be looking at clauses in contracts that stipulate whether AI can or cannot be used, and some of the boundaries of that. And I would also be putting the output through a checker. And that is protecting yourselves as a business, you know, if Google is going to be clamping down on it, you want to know what your exposure is to that risk. And also, from a copywriter perspective, I think you’ve got to be really clear with your clients, whether you’re using it or not, as well, I know, as a client, I would want to be told, and I’m not necessarily against you using it, but I want to understand those boundaries. In terms o where AI takes us, my fundamental view is that it’s as you said, it’s not revolutionary, it’s evolutionary, it’s going to just accelerate a lot of the things we’re doing at the moment – there’s some really interesting stuff around people using AI to auto create marketing campaigns and marketing collateral, it’s just going to speed up the creation, it doesn’t fundamentally change the work we’re doing.
And maybe it’s quite an exciting opportunity, talking about using it for things like schema markup, and quite boring, repetitive tasks. Maybe it frees us up to be more creative and put our kind of creativity into the things where we really add value where it’s real people with real ideas and real thought so, yeah, I’m kind of excited to see where it goes commercially.
I was reading that someone who had been using it to produce content asked it to put in quotes from Hitler and Mussolini and it refused. There are clearly ethical considerations here. But then you can say, attribute this quote to someone else. And it will input that Mussolini or Hitler quote into the content. So, who’s got the liability for the hate speech there? I think you probably as a publisher do. So that’s why I’m quite cautious about this and the fact checking element anyway, that’s my view in a nutshell. I think, again, we could do a whole half an hour on this subject alone.
Jon Paget 25:33
Totally. What do you think Katie?
Katie Taylor-Thompson 25:34
There’s definitely lots of things like that, such as the editors code of practice or Afghan things; like who has the responsibility for making sure that content isn’t spouting hate speech or whatever. And again, at Brighton SEO, I think one of the things we talked about was a twitter bot that started to spout anti semitic hate speech because it was trained on human bias. But in terms of the future, I’ve got a few different ideas. One talk at BrightonSEO said that ChatGPT four is coming out, it’s going to be 571 times more powerful.
So it’s got 100 trillion parameters, as opposed to 175 billion, which is what ChatGBT three has. Does that mean it can mimic human emotion? Absolutely not. Are we also going to see it regulated? Probably, there is already talk of watermarking assets. If you look at Google Trends, people are typing in a lot more often now “AI content detector”, there are already 10 results on Google straightaway. Happily, I put in my own website copy yesterday and it told me it was 100% human. So I know it’s working. So I think the fact that people are already searching for AI content generator terms suggests a general lack of trust in it. But it is going to get more powerful. So does that mean that the watermark detector is going to be less powerful? I’m not sure. But yes, there will probably be some kind of regulation within it. And like I say, if they’re going to monetize it,there are risks with that. Look at what happened to Wordle when the New York Times bought it. So I think the buzz will die down when it becomes a bit more commercialised and invariably regulated.
Jon Paget 27:18
Yeah, just on that, I think maybe we can put some of the best articles that maybe we’ve read on the website along with this recording next week, so that anybody can go and check it out. But I think on that regulation piece, if there’s a lot of money to be lost by industries being disrupted, and I think we’ll see a lot of coordination happening to make sure that regulation comes in. The more organised a sector is, I think the more likely that is to happen. You can’t imagine the legal sector or finance industry being railroaded by this because they will simply put the barriers up – they’re too well organised. There’s too much money involved.
I mentioned it earlier, but we’ve currently got voice AI. We’ve got image AI with Dall:e. and then we’ve got ChatGPT, which is the first chatbot that is effectively able to have a reasonable conversation with us and provide impressive answers. I think if we get to a point where those start to merge, and rather than having three distinct AIs or three distinct programmes, we start to see one that can do everything you can speak to, and it can spit out really sensible, accurate copy, and match that with image creation, which is effectively what they’re trying to do. But in distinct silos. I think that’s where it starts to get super interesting, and potentially very helpful. How they’re going to commercialise it and how they’re going to make money from it is also really interesting.
Bethan Vincent 29:00
And they’ll make money from your data as well. And again, you know that when you sign up to the tool, it’s kind of saying, Look, I’m you’re adding to the dataset that the tool is using.
With my legal and commercial hat on, don’t feed it any commercially sensitive information, because it’s definitely storing what users input into it. That’s a minefield too.
Jon Paget 29:34
Unbelievably, we are half past four. And we said we will keep this short and sweet for a Friday afternoon. So I don’t know whether either of you have any final words that you want to just add in. Otherwise we’ll wrap up.
Bethan Vincent 29:50
I think we’ve heard how we’re all clear in our own opinions, the fact we have a tool getting much more powerful in mimicking our own conversation is a fascinating subject to debate from all sorts of angles., Hopefully people have found it interesting. I definitely have.
Katie Taylor-Thompson 30:25
Beautiful. And just to add to that, I think specifically for the content writers, I don’t want to be ignorant or myopic, but I don’t think we should be threatened. I think we should be aware of it. But like you say, if it is saving us time, so that we can put more energy, more love into creative outputs, then great. And if anything that’s going to make us even better at our jobs. And you know, we’re not going to go anyway.
Jon Paget 30:48
Brilliant, well, maybe we should do this in another 12 months and review what 2023 was like.
For now, let’s wrap up there. I think I will just end by saying that, again, the recording, and some useful articles that the three of us have found useful will be going up on the Open Velocity website next week. I think we’ll be sending out an email with the link to that too. So you’ll get that in your inboxes next week, but thanks very much, everybody for joining in. Katie. Thank you, Bethan. Thank you, it’s been great.
Stay tuned in
Sign up to get our latest content and priority event invites.