What is it actually like running ChatGPT ads?

Transcript

Charles Travers: Hey guys welcome back to another week of eCommerce marketing chats with myself Charles and Jimmy. Jimmy what’s going on mate? I know we’ve just been speaking before the call. happening.

Jimmy Dalgleish: Yeah, there’s a I think I think there’s yeah, there’s a lot going on. like always in e-commerce. But yeah, I think one of the things I’ll be chatting about today is ChatGPT ads. ⁓ so I have so when I say I’m testing them, ⁓ basically if you are looking to get into open AI or ChatGPT ads, whatever whatever you want to call them, ⁓ I will say like straight away up front that when you actually launch the ad, it goes through like a

Charles Travers: Yeah

Jimmy Dalgleish: two week kind of approval process. So even though I started like ⁓ kind of a a week and a bit ago, ⁓ I w we still don’t have anything approved yet. And it’s not it hasn’t been disapproved. It just saying like pending approval. It’s been saying that for like a week and a half. So yeah, if you are planning to get into it, just ⁓ just start now and put something up. And then when you are wanting to fully test it, at least you won’t have to wait another two weeks. But ⁓ yeah, there’s some some interesting well actually we’ll just jump straight into it. So a couple of interesting things on there. So ChatGPT is recommending like a a cost per click of between three to five dollars to kind of play. which for a lot of e commerce stores is is very high. so but what I what I have noticed on like Chat GPT is the conversion rate is much higher. So it may be justified, but the thing is, will it still be as high with the with an ad placement versus kind of a native insertion, which we’ve seen anyway? So it’s yet to be seen whether the ads will have as high a conversion rate as kind of those native suggestions, I guess. But there’s a there’s a there’s another part that you can experiment with the ads like down the down the bottom, you kind of have to Give the context in which you want your ads to show, which is interesting. So the first thing that I’ll be testing with is whether you just dump a whole bunch of keywords in there and then it shows up for that, or whether you’d give ⁓ the AI like more context about like a a really detailed description about the types of conversations you want it to show up in. Which one’s gonna work better? No one knows at the moment. So Because this thing’s literally brand new. So the you’re gonna have a lot of people coming out saying, You need to do it this way, you need to do it that way. I’m very much of the opinion of test, test, test, see what, see what works for you. Take some learnings from what other people are testing and then implement it into your own testing at the moment. So I think there’s gonna be a l a land grab quite early for I guess the guru scene, people who wanna be seen as, you know, the Chat GPT ad person or or this and that. So you’re gonna get a lot of, you know, this is this is one hundred percent how you do it. ⁓ and just with a platform that’s so new, you just have to be you have to be very mindful of all that kind of rhetoric and yeah. And I would love to have some more data for you guys about what we’ve tested and what’s come back so far. But to be perfectly honest, we’re still in the ad approval process. ⁓ and I don’t have any Don’t have any live yet. I obviously have watched a few videos of some people that were in the the the beta testing to see kind of how they use the platform and how else you would test it. But one so another thing that you’ll want to implement, ⁓ if you’re not already across UTM tracking, you want to implement that in your ChatGPT ads because apparently the tracking is Quite hard, ⁓ but you can put your own UTM campaign parameters in there, which makes it much easier to track, both on the Shopify backend or whatever other platform that you may be using. So that’s one tip. Use UTMs in Chat GPT. The other thing that we’ll obviously be testing is what’s going to be the best strategy? Do we want to create an ad that’s really specific for one product, or do you actually want to go at the collection level? Or do you want to just advertise your whole site as kind of the generic thing or generic site in that kind of space? Like and for example, for the supplement space, do you just want to advertise ⁓ you know, where we’re a supplement store and we sell all these different supplements and have that as an ad? Or do you want to look at a particular collection? So for example, you might have ⁓ your magnesium supplement collection or creatine’s quite hot at the moment too. Do you advertise the whole collection for those products? Or do you go really specific on like one particular product for people looking for the highest quality of this particular supplement? ⁓ and then you find your highest quality product and you’re like, this one’s the best. This is the one that I want to push people to. So they’re kind of the I guess the three tiers that you’d want to be testing in e-commerce, which one’s gonna work best? I don’t know. I think natively the one that’s gonna work best, well, the one that is working best on Chat GPT is natively is the specific product. ⁓ but from an ads channel, I don’t know yet. So but I just wanna communicate all this with everyone listening because if you are jumping onto Chat GPT ads, like one of the one of the things that you’ll want to be thinking about when you’re launching straight away is like, okay, what what do I actually test when I jump on this? Or what do I want to test? So these are all the different test variations that I’ll be running through and am running through when all these l ⁓ kind of ads go live. So and I’ll have more data to share with you guys after kind of those results come back. But yeah, I think just knowing what to test or having different ideas to test to start is ⁓ A better place to be than just going on there whipping up a random ad and then seeing how it goes. Because the cost per click is going to be high. So when I put the cost per click at $2.99, it says something like unlikely to serve. And then you put it at $3, so one cent more. And it’s like it goes from the red bar to the double green bar. So They very much want you in that to be bidding in that three to five dollar per click range. So you’ve also got to consider that when you’re thinking about what products that you want to promote, because you’re not going to be able to you’re not going to be able to promote a a $20 product if or if your average order value is under $50 and you’re bidding five dollars a click, you’re gonna have to have like a probably a conversion rate of twenty to thirty percent just to break even. So ⁓ which is outrageous. You wouldn’t have conversion rates that high. So people’s strategy will change a little bit based on how much they’re actually having to pay per click. ⁓ and then yeah, be interesting to see what data comes out. But yeah, chat chat GPT or open open AI ads, whatever you want to call it, is Well and truly open in I think it’s Australia, the UK, New Zealand, Canada, and the USA at the moment. So yeah, if you’re in any any of those markets, even I don’t think we haven’t well, I don’t think we’d have any service-based businesses listening to this. ⁓ but it may be a good play for service-based businesses. I’m thinking like legal or something like that at the moment, because they pay way more than three to five dollars a click on Google anyway. It’ll be an interesting to see how it plays out for e-commerce. I think ⁓ if you’re a higher priced e-commerce store, like furniture or ⁓ really expensive parts or maybe even saunas or spas or pools, for example, ⁓ you ⁓ yeah, it could be a really interesting kind of advertising platform to play on when you when you do have those higher priced products.

Charles Travers: Yeah, that’s really valuable insights there, Jimmy, especially from someone who’s, you know, seeing them for the first time. And I think probably many brands listening and probably haven’t. or they may have started to just have a look at the platform, but it’s useful to get those insights. Yeah, you answered my first question, which is where are ads being rolled out? Because actually you talked about non-e-commerce sites or B2B, and I actually had an Australian client of mine who randomly reached out.

Jimmy Dalgleish: Yeah.

Charles Travers: Last week and I’ve message you didn’t I Jimmy because he was ⁓ asking me like did I know much about chat GPT ads? So Yeah, it’s it’s certainly on the radar I’m gonna be interesting to see Do you get is there much data you can see like can you see even though your ads are not live yet? Can you see any data in there? like I’m really curious to know from an attribution perspective how things are gonna work because Yes clicks are one thing like when someone clicks through to a merchant site, but as we well know Often someone will do research in chat gbt and they may then go on to google to search the brand directly So therefore the attribution is not Directed to chat gbt or assigned to chat gbt is is effectively branded searches. So are you when you’re When you’re building out these campaigns, are you going to be looking at? correlations between chat GPT ad exposure and branded searches on Google search for example.

Jimmy Dalgleish: Yeah, no, it’s it’s an interesting question. And so I have in I have installed the you can install a a chat GPT pixel through the ⁓ platform. I don’t know how accurate it is yet. So ⁓ my question to that at the moment is I’m not sure. ⁓ in terms of tracking, I’ll do be be be doing basic UTM tracking. I think from if organic AI search is anything to go off, I think they have a really high click through to purchase rate, mainly because of the quality of the ⁓ chats that happen within ChatGPT before they go to the site. I just don’t know if the ads are going to correlate the same way as kind of those native recommendations ⁓ will. So yeah, in terms of the pixel being able to track kind of multiple touch points. Again, I’m not sure at the moment. I don’t have any click data yet. But yeah, they do make you they do make you install ⁓ a pixel. ⁓ it does measure conversions. ⁓ but yeah, I just don’t know if the conversions are tracking last click straight to site or whether it is going to track ⁓ first touch points or anything like that. I haven’t I haven’t I don’t have any ⁓ data on that at the moment. So yeah, more interesting things to think about and contemplate. So ⁓ yeah, but I I think the the first thing when you get set up is you wanna in you want to install that pixel ⁓ to get to just to have it on there to get started. And yeah, we’ll have a we’ll have ⁓ hopefully more more answers than questions in a in a couple of weeks when we we’ve actually done a a fair bit of testing on this. But Yeah, I just a lot of people are going to be, especially in e-commerce, are going to be jumping on this. And I just want to give them some some different angles to go in with to test and to think about when they jump on there. So I think the the main big ones are use UTMs on your final like URL that you’re using. And then you’ve got the options, I guess, to test one product specific, a product collection, or like the the site as a whole. Like they’re all different ways that you could add run ads to. to see which is the most effective. The ads are the ads are super simple. It’s like a it’s like a Google text ad, but like sh shorter. ⁓ the the ads manager is absolutely it’s like version one of Google ads back in the day. So I think a lot of people are looking in the back end and just kind of laughing for you know AI in all its wisdom has produce like the absolute basic bare bones of a an ad platform, basically. But yeah, it’s ⁓ there’s obviously a lot of questions, but yeah, I know. So it’ll be interesting. I think ⁓ you know, there’s rumors around open AI’s ⁓ like ⁓ IPO potentially and then wanting to get in advertising dollars early, so them just basically putting together something with sticky tape and

Charles Travers: I wonder if that’s intentional.

Jimmy Dalgleish: Hoping that ⁓ they can get as many dollars in as possible. It was funny when I initially set up the account, it it had ⁓ that had a budget set at a billion dollars. I was like, ⁓ yeah, well, if they had to run that, I would have just done a chargeback if it spent a billion dollars. ⁓ but yeah, there’s there’s a lot, there’s a lot to test at the moment. So ⁓ but yeah, the hard bit is just getting approved and getting on there and ⁓

Charles Travers: Yeah.

Jimmy Dalgleish: seeing what it can produce. So yeah, I think a lot of people are interested in it at the moment. And yeah, I’ll be I’ll be testing it across a few different accounts. So hopefully have some data for everyone in a couple of weeks.

Charles Travers: Nice, nice. Nice, interesting, interesting for sure. On other news with regards to on the organic side of things, over the last week, Google have come out and said that they’re gonna be adding AI search performance within Google Webmaster Tools. So three months ago Bing did this within their Webmaster Tools where you can see what pages on your site have been cited by AI, the amount of impressions for certain ⁓ search queries. And it looks like Google are gonna follow suit. It’s actually gonna be rolled out in the UK initially as well. Something to do with, yeah, data protection or something where they’ve been, Google’s hand have been forced to launch it in the UK, which surprises me, because usually it’s always US first, isn’t it? And then the rest of the world get it. So yeah, that’s gonna be really useful to

Jimmy Dalgleish: Really? Mm.

Charles Travers: because many within the SEO space were kind of waiting on that data. Interestingly, they’ve not mentioned that they’re gonna show clicks from AI searches. It’s only gonna be showing impressions, which I found was quite interesting. Whether that’s…

Jimmy Dalgleish: Okay.

Charles Travers: Intentional because I mean probably e-commerce hasn’t been as widely affected but there’s many industry especially, you know within SAS publications where organic traffic has significantly dropped because a large chunk of traffic has been taken from kind of AI overviews So maybe they don’t want to show that tater they’d rather show, you know visibility and get Website owners and brands more familiar with okay being cited and visibility as a key metric to track opposed to like clicks But yeah, we don’t know the exact date in which it’s gonna be rolled out to the rest of the world. I think in the UK it might, I believe it’s this month, it’s starting to show up. But yeah, who knows when, I assume it’d be pretty quickly after once they’ve rolled out the initial beta launch. yeah, exciting stuff for brands to actually get a bit more AI search visibility data now.

Jimmy Dalgleish: Okay. Yeah. ⁓ they’re gonna need it too. So but it’s it’ll almost be ’cause a lot of people are putting out a lot of content guessing. ⁓ I think if Google bring out something, at least you’ll have a metric to start measuring by like the things that you’re implementing, what kind of impact it’s having on impressions. ⁓ I think once people can kind of see ⁓ kind of a direct correlation to the effort that they’re putting in. I think it’s it’s honestly just gonna steamroll the speed in which it’s kind of rolled out even even faster. So yeah, the people that have been putting in the work early is probably gonna see the benefit of it for ⁓ what’s coming round, I guess.

Charles Travers: Yeah, for sure. think really up to this date, we’ve just had these kind of third party AI prompt tracking tools, which, you know, I’ve used a little bit. In my experience, they’re kind of wildly different in terms of the results you get and wildly, some of them can be pretty inaccurate. I think in time they’ll get better. Like if Google open up some of their kind of… first party data to some of these tools that they can kind of plug into, the data might be better. But yeah, up to this point, it’s not been easy to kind of track AI search visibility, particularly with Google.

Jimmy Dalgleish: Yeah, hundred percent. I think ⁓ something else that we chatted about just before the call, Charles, is like I’ve been working on kind of different different blogging formats and kind of high converting blogs and potentially even advertorial ⁓ pages as well, but I always used to say the lo the longest part of doing one of those is like the f the formatting of it and getting it right. So it’s it was fine to have the AI write the content, but then the formatting of it would be the longest part. I’ve seen now that AI can handle the formatting too. So that’s ⁓ I think ⁓ I think a high quality content kind of onslaught is about to happen. And when I say onslaught, we also spoke a little bit about, you know, not overdoing it. and you coming at it from a a cautious angle too, Charles, about just spamming kind of your website with hundreds of blog articles. It’s probably probably not the approach we want to do, but potentially to start adding kind of one or two blogs a week and ⁓ and seeing what if any is getting traction and kind of tailoring the the template and the approach ⁓ based off that. But yeah, I think there I think there’s gonna be a lot more content come out for e-commerce stores. ⁓ and I mean how many people have gone to an e-commerce store and the first blog that they’ve seen was was written in twenty twenty three? Like the business hasn’t blogged in like three years.

Charles Travers: Yeah.

Jimmy Dalgleish: I mean, if you if you can solve this for businesses and make their business look active, ⁓ I sometimes the the actual content might not even matter. You can just see that the business is actually active and sometimes that’s enough of a factor to ⁓ get the conversion across the line. So ’cause I I don’t know about you, but when I when I go to e commerce stores and they’ve got blogs on the home page and it the last one was written like three or four years ago, you’re like, ⁓ you know, how involved in is this brand? ⁓ currently. So I think just even having some up to date content might might help in all facets. But yeah, there’s ⁓ it’s something that I’m really interested in at the moment in terms of getting co work to put together these blogs. But it’s it’s one of those things again. It’s ⁓ you can’t just go into Claude and say write this blog and then post it to your site. Like we’re actually we’re actually training it up to write a blog how we would write it, format it how we would format it. ⁓ and even then there’s still testing and iterations that go into it. So it’s one of those things with AI that ⁓ it’s only gonna be as good as the level that you train it at. So I think the the people that have really good systems for these for these blogs ⁓ could potentially do quite well in terms of generating some kind of evergreen traffic. But I still think if you go in and your your strategy is, ⁓ I need a blog now, so I’m gonna go into Claude or ChatGPT or whatever you’re using at the moment, say, hey, write me a blog on this, and you cut and paste it into ⁓ Shopify w without formatting and without you know making it look good. I I just I just don’t think that’s ever gonna work. So I still think with AI, the level that you can train it is kind of the level that there is of the results that you may potentially see as well. So Yeah, I’m all for AI, but ⁓ yeah, I’m learning very quickly. ⁓ because I’ll I coach ⁓ a lot of businesses in e-commerce and I see how they use AI, and then I’m like, I 100% can see why you’re not getting results for this. Because it’s a beginner trying to use AI to do advanced things, but they’re missing so many parts. I see it in see it in the Google Ads, for example. They’re like, Jimmy, I changed something on the account and I was asking AI about it and told me to do this, this, and this. I’m like, yeah, you changed that one thing, but you also changed this, this, this, and this. And the AI is only looking at this one part. It’s giving you advice on that. And it’s taking you down the complete wrong path. And then once you explain to them exactly how the different bidding strategies work and how different things work on the account, all of a sudden they go like, ⁓ yeah, I can see that now. This wasn’t actually the issue, but because they think it’s the issue, that’s what they keep telling AI about. And that’s where AI keeps pushing them down kind of this completely wrong path of trying to fix what they want to get fixed. So I am seeing is yeah. The better you can train your AI or the more expertise you have or the person that you’re getting to do things for you has in a particular area, you’re gonna get the better results quickly. If you’re trying to do it yourself from the start, you’re likely going to be learning the hard way, which is fine. ⁓ but it just might take you a lot longer to to get to the point where you want to get to.

Charles Travers: Yeah, for sure. mean, you know talking about blogs for example, there’s a big difference between you know, write me a blog about a particular topic And then it’s got and you use a bunch of stock images Versus you know giving AI detail brief, know Making sure the site and the the page is structured correctly with a linkable table of content adding an author box So you’ve got that authority that Google loves linking from the author box to the box of the the author, you know all these little things adding a comparison table if you’re comparing two products for example adding custom images that you own within your business with it, know These are things that will actually make a difference to how well the page performs and ranks on Google and also from an AI search visibility perspective as well. So You know just creating a blog every single day And dumping a bunch of text on a page with some stock images is here is a complete waste of time in my experience so

Jimmy Dalgleish: I agree. And I think even even when I like have got this template and system done for blogs, like I’d be reaching out to someone like yourself, Charles. I mean, like, what am I missing here from an SEO perspective? Like, what do I need the AI to look at from an SEO perspective? so I I think I think ⁓ almost you people’s expertise at the moment’s gonna be heavily leveraged in cons consultations. ⁓ because if I’m building something then I want someone ⁓ like I’ve I’ve done a template based on something that I’ve seen work for client side but there’s obviously things that I’m missing and then speaking to another student that I was working with about this, they’re like, I use the bloggle app, which I am using at the moment, but I want to move away from that to have kind of Claude format the blog posts rather than bloggle. But The thing that they like is it gives a like an SEO score for their blog article before they publish it or or things that they can add. So you could almost build that into the AI system. Hey, this is from this article, this is how I want to rank it based on like these SEO factors. So after I finish that, give me a ranking of it. And does it include the the H1 titles? Does it include all these these different kind of SEO facets that should be in an article? So if you just Tell AI to write an article, all that stuff is going to be missing. So this is where you get expertise from someone in the arena and say, hey, what what am I missing here? What do I need to train my I AI on to do a better job, whether it’s from the SEO side or helping in ⁓ I guess, AI searchability too. Like, what else do I need in here? So I think it’s gonna be quite powerful. But then when you do get something that works and you can roll it out, like, one blog a week and it’s actually the formatting’s great, the blog articles great, the it’s built to for conversions for your products. It’s ⁓ it’s done the keyword research so you know what keywords you should be targeting. So you’re not just blogging about anything. You’ve kind of got an SEO kind of ranking on the article for things that you might have missed or you should be adding in the article that the AI can add. ⁓ I think once you’ve got that sorted, you’ve got a pretty good system to At least roll out some some blogs across across your e-commerce site.

Charles Travers: Yeah, for sure, for sure. I think one thing that’s massively overlooked as well is actually updating ⁓ old blog content. I was just working with a new client this week and one of the areas that we need to improve is their blog, because they published, I think their last blog was October 2024. But when I looked into the data, they had, I was expecting that actually none of those blogs were likely

Jimmy Dalgleish: Yes.

Charles Travers: driving any clicks or impressions but actually the contrary was true is that they were actually getting some decent clicks and impressions it’s just the content was a bit old so you know one of my job is to look through Google search console and see what keywords they’re showing up for for those for those particular blogs and seeing where the opportunities are there’s certain keywords that they may be getting high impressions for but low clicks for that we can include within the

Jimmy Dalgleish: Mm.

Charles Travers: content, refresh the content which Google loves and also I was reading a study the other week that actually LLMs like. fresh content. That’s one thing that they tend to cite more is more up-to-date content. So there is another reason to go back through some of your old blogs, analyze the data, see how they’re performing, and then if needs be, upgrade them and rewrite them. It sometimes actually gets you quicker wins than writing brand new content. Not to say I still think you should be creating new content on a regular basis, but don’t neglect some of the old content because you’ve already you’ve got some data there already on that content if that you know if it’s starting to get impressions and clicks and it’s it’s still getting clicks and impressions a year and a half later there’s a good indication that there’s there’s value there on the page therefore just giving it bit of a revamp you know would would supercharge it a bit more

Jimmy Dalgleish: What are your tips for revamps? Because I’m always scared to kind of change content on blogs that is generating traffic already. I’m I’m scared to touch them because if I I’m worried if I touch it then it’s gonna go back to zero. So what what are some things that people should avoid, I guess, when they’re updating old blog old blog content?

Charles Travers: Yeah. Yeah, I think first thing what I would be doing is looking into Google Search Console and see what the top performing search queries are that’s driving clicks to that blog. And I would be making sure that those are kept within the article. I would try… my best to be looking at like what can we add opposed to completely rewrite? Because if we’re completely rewriting, then there’s a chance that we’re removing a lot of the stuff that Google’s.

Jimmy Dalgleish: Okay.

Charles Travers: already ranking it for and driving clicks from. So like could we add a comparison table? Could we get in more custom images? Could we add an FAQ section at the bottom of the page and and style it in a way that it doesn’t look like it’s just a bunch of text but there’s actually a frame and a framing to it. Like could we all, for example, another thing that a lot of brands miss is having an author box. Again, from kind of a trust and authority perspective, Google do look at that. They’re the EEAT framework where they’re looking at the kind of trust signals. is again another key signal for Google. So yeah, I do proceed with a bit of caution for sure and I wouldn’t be upgrading and rewriting a blog that was maybe only a couple of months old. I’d be looking first at blogs that are maybe at least six months, more like 12 months old where I think there’s value in giving it a refresh.

Jimmy Dalgleish: That’s good advice. All right. Well, I think we’re at the the half an hour mark already. So I think we’ve covered a lot for open AI ads. ⁓ what’s coming in terms of the the AI ⁓ data that will be in the Google search console hopefully soon and the kind of blog systems, I guess, that we’re working on too. I think ⁓ I think that’s pretty good takeaways for a lot of e commerce ⁓ businesses today. And Hopefully we’ll have a lot more data on all of those things in the coming episodes too. So we can continue to give you guys the the best data that’s coming out. If you want to chat to us, head to Facebook on e commerce marketing secrets or or Shopify Marketing Secrets even. And then ⁓ Key Commerce Marketing Chats on all the podcasting platforms. And ⁓ we’ll chat to you guys next week. Take care, guys. All the best.

Charles Travers: Shopify marketing secrets. Yeah, we’re still here. Good stuff.