Why Your AI is Targeting Men, Ignoring Languages, and Boring Your Audience (And How to Fix It) - Episode 179

Some language models like ChatGPT, they favour English content. It's the language bias. Based on the language model that just gives you answers, the ones you use, it varies. You can widen your footprint, your digital footprint, by having an English version of your cornerstone pages. They noticed that there was an increase in AI citation because of that on ChatGPT. It's an increase, with the test they did, of, like, one hundred and twenty two percent. So that's a lot. But for other language models like Copilot or Google, it was less. But there was still an increase.

Okay. So if your site is predominantly in a language that isn't English, it's worth having English language versions of pages.

That's what they concluded, because it's not going to harm you if you actually commit not to have, like, "zombie" sort of content.

Oh yes.

Because otherwise it's going to retrieve outdated information and that will harm you.

Hello and welcome back to Digital Marketing From The Coalface with Julie and Leslie today, Dave's on holiday, so we thought we'd just do one anyway because who needs, you know, guys. We'll just, do a girl podcast. It'll be absolutely fine. We'll maybe talk about slightly different things today because we're not going to get into, like, SEO nitty gritty stuff, but Leslie's more, like, video and visual and creative side. So we might sort of go off down some slightly different routes.

We'll give it a go. We'll give it a try. So. Well, if you want me to just start.

Yeah, you had some really nice ideas about some stuff to talk about, so go for it.

Yes. So basically when I went down to France last month, my brother lend me a book which is called Savoir vivre avec l'IA, which is written by a researcher, a French researcher (Laurence Devillers). She's a professor at the Sorbonne. She's also a researcher at the CNRS. And she talks about a lot of things in this book about understanding how to live with AI. But there are two ideas that made me think of you and Dave and I want to talk about these topics.

Brilliant.

So, one of them was the sexist bias that turned out not to be a sexist bias because-

Yes, it was interesting, So discuss that. That's really interesting because it, it kind of is what you think, but it also isn't what you think.

Yeah. So basically she talks about a lot of, um, like disciplines in her book, like education, etc.. But at some point she of course mentions marketing. Because AI is so prevalent in marketing for many different reasons. And so I think she mentioned sort of like a study or a research she was doing with her students, and about the fact that basically they asked, uh, an AI language model to target- The idea is to run an ad, an ad for the higher education sector, I think. So they wanted to attract women, uh, within this degree that is, I think, traditionally focused on attracting men. Engineering, maybe.

And AI and computer science-y things, which does tend to be fairly male dominated.

And so basically what they did was, like, use AI to assign a budget to run this ad.

Yeah. So it basically created the campaigns and things, didn't it? It kind of planned the campaign, planned the whole campaign. They automated it as much as possible.

I will have a look at it again when my brother will bring me the book that I left in France. I will look into it to, like, maybe add some more context. But basically, yeah, the AI basically ran the thing. And so they realised that it didn't target a female audience at all.

And that was the kind of general aim. It was to get more women.

Exactly. So they were wondering, is it because we're, well, we're living in a sexist society, and so the AI is trained on data that is biased.

Yeah. And you would have thought maybe they see, okay, these degrees attracts men, and it would have kind of the messaging. You'd think maybe the messaging would have like innately targeted men, because that's how the AI has been trained, which you think, yeah. Well, in certain respects it is. And I think there are a lot of cases where that's been happening. But it wasn't that.

No, in the end it turned out that because, um, the cost of targeting female in general is higher because, uh, women are, like, the target of- maybe because we are looking to buy things for our family, like brothers, partners...

Women do more shopping. Women make more buying decisions. But women are more often than not the people that actually, you know, click the button and make the purchase. So yeah, the cost of advertising, I didn't know that, but the cost of advertising to women you're saying was higher for these ads.

So, based on the, probably like, the guidelines that the AI received not to maybe go over a certain spend per day or whatever.

Cost-per-click or something so that-

It didn't, like-

The cost of a man was-

It targeted men.

The click of a man was worth less than the click of a woman because, yeah, because of supply and demand and the way these auctions work, which is really interesting.

It is interesting. I don't know if we can discuss a lot around that, but I thought, I wanted to mention it to you.

Well it's weird because, um, we've got this really, really old blog post called why do marketers target females? Which was years and years ago, we did a like an advent, um, like twenty four days of blogging and they were a little short blog posts. They weren't even really blog posts. They were just little pieces of content. And we did one a day for advent and it was, um, it was something or nothing. It was like basically one of the team having a bit of a rant because her husband wasn't doing any Christmas shopping mainly. So it wasn't really anything, but it's, you know, it's our most visited page on our website because it got picked up somewhere on YouTube. I think we've mentioned this before, but it's got picked up somewhere on YouTube and, um, you know, it, um, it gets loads and loads of visits. We then rewrote it and it's actually a really good piece. Alex rewrote it and it's actually a really interesting piece about um, the female orientated advertising and stuff like that. But um, you know, females are, women are a major audience for any sort of selling anything that, you know, around the house. Like most sort of day to day purchases is just inevitably the women that bear the load of doing the shopping and making those decisions. So, certainly for B2C products, marketers do target women more. So that would make sense that their clicks are more expensive.

Mmh-mmh.

And it's obviously a topic that's well researched on the basis that the number of visits that blog post gets. It's obviously something, maybe it's something that is in marketing courses and it's students are writing essays on it or something. It's most likely that, it's something like that. It's an essay topic for a university course or something, which is probably why it's getting clicks, but it is a really interesting subject. And it's also how these, um, you know, like Facebook Market, um, Facebook Ad algorithm, um, the, it's like an auction thing. So the more bids that go in, the higher the price gets. So there'll be more bids for, like, the women audience than there are for the men audience. So that's why the price keeps going up because of the way the the auction runs. So yes, so the AI, that doesn't prove that it isn't male bias, but it does prove that. Um. Yeah.

There is a sexist bias, because it's still, like, a sexist bias, but not the one they were expecting.

Yeah. Not the one they thought it was.

Which was very interesting. Which leads to another bias that, um, she talks about in her book. And that also was explored in an article I sent to you and Dave earlier this week. It's the language bias.

Yeah.

So some language models, like ChatGPT, they favour English...

Content, yeah.

Content online when they're- it's very interesting. They go into more details in the article, but basically what Laurence Devillers talks about in her book is that, because she is French and she's a researcher in French tech, like French institutes around- like, exactly the name is the Interdisciplinary Laboratory of Digital Sciences. So she's focused on,

OK! Yeah.

you know, uh, how to make the French language and like French tech, like competitive, um.

With, like, the American Silicon Valley type stuff.

Absolutely, yeah. And so that is a thing, like with AI, there is like a hegemony, maybe the risk of hegemony of the English language, which is already used in a lot of places. So how do you, like, study that, and maybe counter that, and these sort of things?

Yeah. And I guess the training data for these AI engines is going to be more English language stuff going into their training, I suppose.

Well, yeah, that's very interesting because in the article that is from Search Engine Land, if I'm not mistaken, "Should multilingual websites add English pages for AI visibility?" So they ran sort of an experiment.

Right, yeah.

Because they wondered, like, they said that two questions led their research. The question was: how much does AI rely on multilingual sites' English versions?

Okay. Yeah.

And the second question is: for sites with an English folder. Are they losing visibility because of it? Or for sites without an English folder, are they losing visibility because of it. Without, sorry. So, uh, and so they realised that based on, uh, language models that just give you answer, the ones you use, it varies. ChatGPT is very biased, apparently. It's going to favour English, uh, content.

Even if you ask in French, or whatever?

If you ask in French, for example, it is possible that it's going to crawl a specific multilingual website. And even though there are pages in French or in Japanese or whatever. Apparently, it will favour the English version of that page.

And then will it answer in French, but using the information on the English page.

That's a good question.

Yeah.

Usually when I ask something in French and usually I don't really use ChatGPT, etc., but you know, it's like the AI excerpts from when you like when you do a research in a browser, just like- the Overviews, sorry.

The Google ones, yeah.

Um, sometimes they answer me in French and sometimes they answer me in English, though I asked in French. So I guess it depends, but I don't think there is a mix. I think there is. When I ask something in French and sometimes it answers me in English, the answer is fully in English.

But maybe because it knows you're in the UK?

Maybe. I'm wondering to which extent, because I do research in English. Sometimes I use a VPN as well. So I wonder to which extent I confuse browsers.

I don't know where you are. I don't know what language you're speaking! Yeah.

So maybe, yeah.

But I mean, I typed something in German into Google today because I was trying to translate a page on a website into German. I did German at school, but technical German, not a clue. There's all these long words that are about three hundred letters long. And I mean, I'm not understanding. So I put one of these words into Google because I was just trying to make sure it was in the right place on the page and... But it didn't. I thought if I just typed it in, it would know that I wanted it translated, but it didn't. It gave me a great big answer in German to whatever. I don't even know what the word was I typed in. It was something about pipelines. Um, but.

So it answered you in German.

It answered me in German. Even though, well, I suppose just because I typed a German word in, it answered in German. So it didn't guess that I wanted it translated.

And maybe the information you got that the, uh, AI crawled was only available in German anyway. So since it identified as such in German, they didn't have like an English page available on their website. Only German. But if they had an English page available on their website, maybe it would have-

Yeah it's possible. Maybe it would have shown it to me. But yeah, anything-

Again, it depends on the-

Yeah, anything containing this word about pipeline, whatever it was, maybe because it's such a very specific word.

Yes that's true.

It could only find German content, I suppose. But yeah, I never really thought about it.

And, uh, yeah, what's really interesting again, is that, um, so it depends on the language model that you actually use. So they say ChatGPT is very biased. They create- So basically what they did is they turned the bias they noticed into, like, a single index and the... What they did was they took the AI citations, like the number of AI citations they had, divided by their share of traditional search impression, for, I guess, the same query. And so they, well, by dividing them, they just happen to have an index. And 1.0 as an index is neutral, and I think Google is, uh, I think it's Google that is neutral, uh, or Copilot. Copilot is neutral. Google is going to favour apparently, uh, not necessarily English language, but ChatGPT has a very strong bias of, they say, 2.6. So, uh...

And I guess Google's just been searching the web for longer. It probably has more background. It's been, it gets it more. I don't know. I mean, ChatGPT is still relatively new and maybe just hasn't been trained on as much foreign language.

Yeah. That's true. And one of the things they talk about with ChatGPT as well is that, uh, it actually has like three crawlers. So there is the crawler that, uh, they call it the training one.

Yeah. Yeah. There's the training data, which is the, you know, how do people speak, what sounds like human voices. What words go with other words and that sort of thing, isn't it?

The indexing one and the retrieval one. And apparently it's the retrieval one that is very biased. So it might crawl and train on like varied data, but the answer it's going to give you, apparently it's like-

It's more likely to find out an English language page when it, I mean, it's like up to date, like I need to know this thing.

Yes. Because the thing they mentioned is that, uh, it can be interesting for websites to- Basically the word they use is that you can widen your footprint, your digital footprint by, uh, having an English version of your cornerstone pages. So they noticed that there were an increase in AI citations because of that on ChatGPT. It's an increase, with the test they did, of like one hundred and twenty two percent.

Okay.

So that's a lot. But for other language models like Copilot or Google, it was less, but there was still an increase. It was, I think between like under ten percent or around fifty percent.

Okay. So if your site is predominantly in a language that isn't English, it's worth having English language versions of pages.

Apparently, yes. That's what they concluded because it's not going to harm you if, if, they say, you actually commit and you update- Not to have, like, zombie sort of content. Because otherwise, it's going to retrieve outdated information and that will harm you. That's what they say.

Right, so if you update one page, update the English version as well. Yeah. Don't just leave it sitting there and be completely out of date, because that isn't great.

Because even, like, a forgotten PDF on a drive or somewhere on your website that is out of date, they're still going to crawl it. So you really have to be, you really have to commit. It's not a one-off sort of thing.

Yeah, it should be updated along with the rest of your website, which makes sense.

So that is it. I found it very interesting, and I found it interesting because for a long time we've been discussing, like, the idea of maybe creating a French version of Red Evolution website, or some pages at least, because I think there are, uh, in our industry, in the industries that we are targeting, people... Like French speaking companies who would be keen to penetrate the English speaking market to just- Because... I mean, we see it with STATS, you do a campaign in German just now. They're based in Scotland, but they operate...

Yes. They're a Scottish company, but they operate worldwide. They've got offices worldwide. But yeah, they're breaking into the German market.

And we had Master Industries, a French company.

Yeah, they're a French company. And we worked with them for a long time because they were trying to break into the American market. So they wanted an English speaking agency to, you know, help them basically create a decent English language version of their website. And they had like, I think they had an English language page or something, but they wanted to properly target the American market. So they needed a full English language website. And they wanted a, you know, English speaking agency to do that for them. And that worked really well for them. You know, we did manage to, to help them, um, get some business in the States. And I think they're doing pretty well over there now.

So yeah. So that was like setting up pages for them, and also running -I remember- running campaigns on LinkedIn as well.

Yeah.

English speaking ones.

We did, yeah.

We stopped working with them before the AI thing came along.

Yes, that was during the pandemic. Then it got difficult.

Sorry I interrupted you, but they could now be cited in ChatGPT, for example, if someone is looking for a solution based on what they can provide, and if they update their content in English, hopefully they do. But the thing is, now it is a proper strategy we could have suggested to them.

Yes, that's right.

It should work in their case. So that's just very interesting.

Their English language site is actually, or while we were working on it, the English language site was a lot bigger than the French language site because we were adding blog posts and case studies and things, and they were doing the French site, but they weren't sort of keeping up at the same rate. So the French site was actually a lot smaller. So that's presumably a good thing for AI. The thing with doing, um, like foreign language pages or versions of your site, you've got to do it properly because otherwise, um, if Google is able to sort of compare them, it's like, well, that's the same page. So you kind of have to tell Google it's the same page in this language. So it kind of knows, but that's just like a bit of code on the back of your website. And that tells it, you know, this is the French version of this page. This is a German version of this page. And you know, as long as you do that properly, then the search engines, and presumably the AI engines probably work the same, they would understand that it's the same page. This is the same information. Show this one in this country. Show this one in this country, or to someone asking in this language. Then it would know that, you know, pick this page rather than the original, but it would know it was the same information. So, you just, you know, you set that at the beginning and do it properly so that you're not muddling everyone up.

You know, that's when Amy's input would have been very nice because she's upstairs and she's a developer and she could have given information about that. But yes, you have to set up things correctly in the background.

Properly, yeah.

It's not just a...

Don't add a new page and translate the content and add it as a separate page on your website. You know, just set it up so that it's the same page with a language version. And then that means everybody, as in every machine, understands what it's doing, what it's for and everything. So that's just a little tip. It's um, most systems do that kind of pretty easily. Like, in HubSpot, it's really easy. You basically just press a button.

Okay.

It's very nice! You just press the button and it kind of makes a stab at the translation as well. So when we were doing this German's page we were speaking about, so we, um, we built the English language page and then you basically tell it's a multilingual page and you tell it to do German and it does most of the translation. And then we got an actual German speaker to go through and check it, but he said it was pretty good. He did make a lot of changes, but, you know, he wasn't like, this makes no sense. What a load of rubbish. He was improving it and making it sound more natural. Particularly when it was, like, really technical stuff. So he was going through and saying-

Yeah.

You know, yes, it's correct. But actually people probably wouldn't say that. And this sounds better. So, you know, he went through and made the changes, but it was much easier to get HubSpot to do the translation in the first place and get him to check it than asking him to go through and translate it, which would have taken him a lot longer.

I remember when, uh, you worked on the French landing pages for the same client.

Oh yeah, that's right! And you went through and checked them, didn't you?

Yeah. And I think it's on standby for now, that project, or maybe they went live in the end, but-

They did go live for a while. But then I think they changed direction slightly. Yeah.

And, uh, yeah, my impression was that it was pretty, uh.

It was French but...

Apart from the technical lingo, like the vocabulary, sorry, that I had to look up to make sure.

Yeah, that's the thing.

It was honestly, surprisingly, like, okay. In terms of like construction of sentences, this sort of thing.

You think back maybe what, five years if you asked Google Translate to do something, or back ten years. It was pretty awful. I mean, it's, you know, the difference is amazing. You can get a really- You could get a translation that you could probably put out into the wild and nobody would be absolutely horrified.

Mhm.

It's not that bad.

I've been experimenting with, uh, I think it's still in HubSpot. There's the possibility to generate, uh, a French version of an English post, for example. Like, just translating it. I think it was... Oh. No, sorry. It was in YouTube. You have the opportunity to, like, publish a video and a description of the video and a title in another language, any language. So I did that with French. I have to catch up actually. But I did that with French on some of our reels...

So does it change the subtitles to French then?

Uh, no. I have to upload them.

Okay.

Um, because I always upload them separately.

Right.

But the titles and description, it did a pretty good job translating it. I barely touched them, but I still take the time to go through them because... It's what I told Dave at some point when he started experimenting with the translation of some blog posts a long time ago, from English to French, I said: it's pretty decent. It's not our tone of voice.

Ah, okay. Yeah.

And I know that in French with the, you know, the formal and informal yous...

Yeah. Vous and tu, yeah.

Yeah. I felt like Red Evolution would use the informal you because we want to, like, convey close proximity, this sort of thing. So it's just, like, adjusting this sort of thing. Honestly, other other than that, it's pretty decent.

I know, it's amazing.

So yeah, all of this to say, Yeah, that's, uh...

That's interesting because, you know, a lot of companies operate in, you know, in more than one countries. So, um, you know, if it's much, much easier to create, like, language versions, you know, why wouldn't you? Interesting actually, because we've got another client now. Their main market is actually, like, mainland Europe. More so than the UK, even though they are a UK based company. Um, so they target like, you know, fleet management and stuff in, in Europe and we've been running very small scale Google ads, just a little budget just to sort of, um, you know, put them out there. It's quite a niche market. So they don't need to, you know, spend masses of money. They rank pretty well organically anyway. But they wanted these Google ads. They had them running before we started working for them. And we sort of tidied them up and they were keen to just keep them ticking over. We're going to do something different with them in the next couple of months. But so far they're just ticking over. And they were running French and German campaigns and English language campaigns. We paused the German one. Then we're running French, and then we did a Spanish one as well. Um, and the French and Spanish ads were doing nothing. So, um, my phone's ringing. Sorry, my phone's ringing on my watch. I don't even have my phone, but my watch is ringing.

Technology.

I never answer numbers if I don't know who they are, I never answer my phone. Actually, even if I do know who they are, I usually don't answer my phone either, so don't call me. Um. WhatsApp me. So what was I saying? Yeah, we, um, we paused the French and Spanish because basically they weren't doing anything. Um, we kept the English language ads running. Like, it's very niche, but so, you know, we go in and check, you know, which keywords are getting clicked on and things like that. Um, so it's things like tachograph reader. So, um, looking at it going, right, you know, we've got certain number of clicks on tachograph reader, you go and look at the search terms. So you see what people actually typed in when Google decided it matched that keyword. And um, we get um, Italian, German. Something else, mainly Italian and German, but like the, um, people typing in tachograph reader in German or Italian, they're still getting shown these ads.

Okay, in French, for example, or in Spanish.

No, they're getting the English language. We've only got the English language on running. But if people type it in in German or Italian, Google still knows that they're looking for a tachograph reader. So, I didn't know it would do that. I thought if you were targeting, you know, if you wanted somebody in Italian, la torre de tachograph or whatever it is called, um, you'd have to use that as a keyword, but Google knows that it's the same. So it's showing your ads even if you're not running the ads in that language. Which, well, you know... So you don't actually have to, I mean, obviously it's better if you're really targeting that audience to have ads in their native language, to have a page in the native language, and which we did for the French and Spanish, but for, um, you know, if you're just experimenting it and see, you know, is there an audience out there? Are people typing that in? Even if you don't have that language, Google's picking it up.

Okay.

So if you type in in German, the German version of tachograph reader, you're getting shown this English ad for a tachograph reader. Even without a German campaign running. So Google translates it sort of on the hoof and understands that it's the same thing.

And did it give you ideas to propose to that client to maybe create campaign in German, consequently, because of that result you weren't expecting?

Well I was able to go back and say, well, there's obviously- The German and Italian audiences seem to be the most active in Google.

Mhm. And so in Google specifically, did you have other results from other... Because it's Google Ads, they show up as, like, sponsored results. You can't find them anywhere else?

No. That's right. No, no, just just in Google search. They're not the ones that sort of follow you around the internet or anything. Um, just in Google search. But, um, yeah, Google can understand and shows your ads, even if somebody types it in in a different language. I mean, we were targeting the whole of Europe. We didn't say just target the UK, but yeah we were targeting the whole of Europe with them. But you don't need to, you know, you don't need to type it in in English. You can type it in any language.

Okay. And there were like a French campaign running at the same time.

We had paused I think all the campaigns. All the others. We paused French and Spanish. We just went for the English because they were the only ones getting clicks. And they seemed to be doing the job of all of them.

Yeah, yeah. They're just redirecting...

They're just, yeah...

People to-

To the English version.

I was just wondering if there were like a, like a French campaign running if they would, if Google would just direct Italian.

To the French.

To the French. It would make sense to me because it's close and we share a border. But at the same time, the languages are very different and it makes more sense to direct to the English campaign because English is the global language.

Yeah. I don't know what it would do if they were all running. I'm not very sure.

Yeah, maybe. Maybe Google doesn't know either. Maybe they're just like...

I suspect it makes up as it goes along, much like the rest of us.

Yeah. So okay.

The other AI thing that I was watching, reading about, and it was a webinar I was doing. So, um, Otterly AI. I think we've talked about it before when they do their experiments and they try stuff out, I think I spoke about some of them in the last one, but they did, uh, um, that sort of experiment to see if AI could judge art. Or if it could judge designs or anything. So it had like, um, various sort of abstract-y paintings and it was like, you know, they asked people, do you like A or B? You know, with the first design, most people liked one of them. And then the second one, a lot of people liked the other one. And, but with the AI, it pretty much just, if they kept asking it, it just was more or less fifty fifty. It was just guessing.

Okay.

And it wasn't coming down on one side or the other. It basically hasn't a clue, so.

No, it can't. Well I guess- Sorry, go ahead.

It has no way of judging because it's not data. So, I mean, I'm not sure why they tried it, but, I guess, things like, you know, designs, there's no use saying to AI, if you've designed two adverts or whatever. It's not going to know which one's nicest or best or whatever, it has no idea when it comes to images and designs.

Yeah. I guess what would be useful to query them for about in this context is to ask them about user experience based on where you want this design to be. Like if it's a landing page, then...

It might be able to.

It might give you advice on, like, other content that it sort of scanned and the result that this content had. And I feel like in terms of design, like the user experience part.

Yes, it might be able to say it's getting in the way of the, the words, or...

We could rearrange that so that it can be more effective or it's too long, shorten it, or this sort of thing.

Yeah. It's um, yeah, things like that or it's um, it doesn't show the product properly or the words are in the way of the picture. It might come up with things like that. But in terms of, you know, which painting is nicer? It does not understand.

Yeah. And yeah, you're right. And going back to like brand voice or tone of voice that I mentioned earlier with the translation of a post or, I'm sure Dave mentioned it in a podcast episode at some point, about the fact that in, like, an era where AI is everywhere, it cannot replace your brand tone of voice. And it's very important to have a solid one and to carry on investing in it.

Yeah. Be almost recognisable. If somebody read that piece of writing, would they know it was you without saying it was your name or your logo or anything else? Would they know? And then there's some brands you would know straight away, like, who they are or, you know, roughly who they were, whereas others, you just like, well, you could be anybody. Now I'm trying to think of examples and I can't think of anything, but, you know, things like, um, Lidl and Aldi on their socials and stuff. You know, they have a very distinct tone of voice. Um, you know, just, you know, coming out with, um, kind of random stuff and um, the way they speak is really quite distinctive. Whereas a lot of other brands, it's just, you know, could be anybody.

Yeah. That's true. Duolingo as well.

Yeah.

With their owl, uh, mascot.

Yeah. Although they've, I don't know, they've gone, I don't know what happened. They were very distinctive in terms of like, you know, do your Duolingo and, you know, the bullying owl. But then they went a little bit weird and I'm not sure what they're doing there.

I know that they disappointed a lot of people, and rightly so, when they replaced most of their translators with AI.

Yeah they did. Yeah, that was a big backlash then.

Maybe that happened around that time as well. I don't know, I have no idea.

They changed their advertising and it all went a wee bit bizarre. Yeah, it went a bit weird.

Like a fever nightmare.

Basically, yeah. I don't know what they're doing now. But yeah, there's some brands like that, they have a very distinctive voice and, you know, they're really good at sticking to it. And I think it's probably more important than ever to try and, you know, have a way that you, you know, communicate and have a very specific tone of voice that is yours because then, yeah. Even if AI is helping you write stuff, you still want to make it stand out and not just be the yeah, here we go again sort of thing. I mean, you know, LinkedIn's a complete nightmare at the moment. I mean, you just look down and every single post, it's like it's not a such such it's a such and such. And then there's like three bullet points and it's like, oh my God, they're all the same. Just, you know, just either write your own LinkedIn posts or edit them before they go out, but they're all the same. And they're all long and they're all, you know, really tediously similar. And there's some really obvious things, not just the dashes, but, you know, AI is very much like they have like lists of words and, and yeah, it's not this, it's that.

You can really recognise it, yeah.

If you read enough of them, you're like, haven't I just read that? Oh, no, it was something else. It was on a different subject, but the sort of layout of them is so similar. And, you know, you can get AI to do it, but tell AI how to make it sound like you. And tell it not to do all that nonsense because it's just boring.

Yes. Yeah, yeah. It really like must be a tool to widen your footprint or amplify your voice, but your voice is yours.

Yeah. And speed up your work. Speed up your work, not, completely replace your work.

Otherwise. Yeah, it's going to do a lot of harm.

Yeah. And you can train it. I mean, you can, you can write stuff and then then show it. You know, like read this and and use this and create a tone of voice and it has to sound like me and, and it'll do it. If you, if you give it enough background information, it will do it, but just without any briefing, it'll just give you the same thing it's giving to everyone else and it is just boring.

Slop. Yeah. I haven't checked this information and I saw that on different, like, occasions. But basically the Internet is now apparently more bots than humans anyway. And it's like AI, just-

Robots are writing for robots.

Yes. In the end, it feels like that a little bit sometimes. Like, when you mentioned LinkedIn I thought about that immediately. Like, people sharing, like, using language models to share ideas with the world, but they're going to be read by other bots rather than humans, because humans will skip-

Yeah, because LinkedIn, it's like...

Information like that.

It's like the second most cited source now in some of the AI engines. Anyway. Um, Reddit has- because Reddit was always like, oh, you've got to be in Reddit, you've got to be there, because that's where AI is getting all its answers from. Reddit has blocked all the AI engines except Google. Google's paying a vast amount of money to Reddit to have access and, um, all the other ones haven't paid Reddit. So Reddit's blocked them all. So if your entire AI strategy has been about going into Reddit. Bad luck, bad luck, because, um, that'll only get you into Google and it won't get you into all the others. And now, looks like LinkedIn's actually your best bet for all the others. Okay.

And at the same time...

But it's kind of tedious because it just means everybody's using LinkedIn as a way to get AI visibility and...

So in the end, are you really visible? Because there is so much noise.

So marketers are going to- well, marketers already broke LinkedIn really, because LinkedIn's never really been much fun. It's always been people doing what they've been told they should be doing. Like, there was a phase of heartfelt sob stories, and, like, personal pictures and the, you know, the really crappy side of Facebook appeared in LinkedIn. And then now it's, you know, people doing great long posts and, you know, trying to get picked up by AI by using LinkedIn, which will just at some point break LinkedIn completely and nobody will use it because it's just so tedious and it'll just be like robots writing for it. So you'll automate the LinkedIn post, so Claude will go away and write a LinkedIn post and probably post it for you. And then the other robots will read it and regurgitate what they wrote in the first place.

Yeah.

Oh, God.

Yeah. That's just-

It's really quite depressing, isn't it?

I mean, it's very fascinating. I'm just going to, uh, mention- I actually don't know if my voice is picked up. Hopefully it is picked up. Um. It made me think about something. So basically the fact that it's probably just bots talking to other bots and the fact that marketing and, like, social media... Like, there's the word social in it, so it was supposed to connect us with one another. And of course, for different reasons, we feel more disconnected maybe when we use it, I don't know. I wonder if-

I wonder if at one point we'll actually all start going out and meeting people in person again.

Yeah, probably. I was talking with my best friend yesterday about- he brought up the topic of... He said: I read something really sad on Reddit and it was about, um, people already talking about their AI boyfriend.

Oh my God.

Like, I'm not judging at all because we all have like need and we feel lonely, this sort of thing, right. We talked about a lot of things regarding this topic, but one of the things I raised up was the fact that it's supposed to be here to help you address things like, um, this sort of thing. And, I don't know, it just makes us, like, more and more disconnected from one another. I feel like we isolate ourselves using it.

I was talking to somebody yesterday. Um, and they've been in a, some sort of meeting or group with a lot of people who had maybe just graduated and just joined the workforce. And they were saying that they didn't speak. They were like, and they didn't want to speak. They just, you know, they were really uncomfortable in that sort of environment. But then they probably had most of their lectures online, particularly if they've just- people who have just graduated, probably had a couple of years of online lectures. So they haven't had that sort of environment where you're in a big group of people and you participate and you, you know, you bounce ideas around. They've been sitting in a Zoom call with a camera off, you know, not making a noise, muted. And then they go into a real life situation in a big conference or something, or around a big table. And, you know, they're probably really uncomfortable because people can actually see their faces. And, you know, they're, they're in real life and they're not accustomed to speaking out in that environment.

There is a comfort in delegating things to technology sometimes, but like, there is the risk of, well, not- I mean, it's like, it feels like a trap. I don't know how to explain that.

Yeah, and once you get out the habit, I mean, remember in the pandemic, you know, once you get out of the habit of leaving your house and socialising with people, it's hard, I find it. I do less socialising now than I did than before the pandemic, because sometimes it's like, oh, I'll just stay in. But before I was out more. And I think, you know, it makes it harder to get out of your house. And if you've got something, an excuse not to, and some sort of technology to do your socialising, to do your communicating for you, then you are going to let it. And then if you're less often meeting up with people, you're then, you get out of the habit and you're less comfortable with it. Then it becomes much harder.

Yeah.

So, we need to be training our young people to go out and meet and socialise and not hide behind screens. I mean, there is the thing with the ban of, um, social media and things for under sixteen, although they haven't decided on that, haven't they?

We have the same in France. They are debating it. Under fifteen, but yeah, yeah, yeah. All of this to say, like, sorry, just like when we talk about marketing, it's about marketing something to people who have a problem and you really have to connect with them. And there is really a problem of authenticity- when you received a call earlier on your watch and you mentioned the fact that you don't- I'm not picking up because I'm terrified now. There really was a before and after AI. I'm terrified now that when I speak, my voice will be stolen. So I definitely don't answer anymore. So, like, this need for, I need to know...

Never say yes, because that's the worst thing. I go: Uhu. Mhm. And they don't understand.

And so, like, really, like, there is this desire to have a proper, you know, you have a human being on the other side that can answer your problem. And so that is why some customers will really see in a very bad way when you're obviously using AI to do your posts, for example, or this sort of thing. And yeah, I just. It's just that that I had in mind when talking about the fact that we could use AI to amplify our voice.

Yeah. But not replace our voice.

Not replace our voice. But it feels like sometimes- And at the same time, you know, when you were mentioning people, young people entering the market, I know that now when you send a CV, it's first triaged by AI.

Yeah.

And so you don't even have the opportunity to put your foot in the door by talking to a human being, you know?

And sometimes if you're kind of- people used to make their CV look a bit different and stand out. And, you know, that's not even going to work anymore because that's going to get filtered out if it doesn't tick all the boxes from the machine.

And so, you might be losing the opportunity to hire a fantastic person that would completely fit the culture of your company because their CV wasn't formatted-

Yeah.

And organised the way... And I think we harm ourselves so much using technology, sometimes.

People, you know, people still have to, you know, use- my nephew was looking for a job. He's seventeen, sixteen, he's sixteen. He was looking for a weekend job and he, you know, he wasn't getting anywhere. Um, there were a lot of them that were saying you have to be over eighteen, or whatever. So, um, eventually he remembered that my daughter, his cousin, um, knew somebody that worked in a fast food restaurant in Edinburgh. So he contacted her, said, does your friend still, you know, work in this fast food restaurant? She said, I think so. She contacted the friend said, do you still work in the fast food restaurant and are you looking for staff? He said yes and yes. So she went back to her cousin and said, yeah, you know, they're looking for people and obviously, you know, connected them and he got a job and he's working in a fast food restaurant in Edinburgh on the weekends now. But that was purely down to actually using human contact, not going through the system. And I think people have to almost relearn how to.

It can be scary. Rejection is scary because you have the possibility to be rejected-

Yeah.

In direct, like, live.

By a person. Being rejected by a machine is not nearly as scary. But yeah, people have to go back to understanding where the human side is important and where the machines are useful. And, you know, the human skill is lacking. And we've got to kind of find a way back to, you know, using that when it matters and using the machines as assistance, when it doesn't matter, when the human side doesn't matter. So yeah, people, I mean, yeah, it's new and everyone's just trying everything out. But I mean, I'm hoping that the human contact thing will be the thing that makes people stand out in the future. And the people who learn how to really relate and really communicate in person are the ones that are going to win because everyone can learn how to use AI, but not everybody's got like the human side. And yeah, people need to put themselves in scary positions. You know, it's terrifying. It's terrifying going out in front of a room of people. But yes, you have to do it. You have to do it. People, get out there! You know, do stuff that scares you and just get out there. That might be a good place to end. Go and do the scary things.

Yes. Good advice.

Right. Well, thank you very much for listening. If you did. That was Digital Marketing From The Coalface. I've been Julie and that's been Leslie, and we will see you next time. Bye!