Survive and Thrive with Paul McNally

Janelle Dumalaon: Welcome to Survive and Thrive, where we explore the challenges of independent public interest media in an era of turbulence.

Around the world, free media are fighting on multiple fronts. Repression, censorship, shrinking funds, competition from digital platforms, and a deepening loss of public trust.

Last season, we focused on individual outlets, how they endured, and what lessons they offered. This season, we zoom out: What will it take for us to survive and thrive collectively in a world that keeps testing journalism's resilience and relevance as a cornerstone of democracy? In other words, we're not talking about how media outlets can stay viable, but how media at large can. 

And that brings us to today's episode. Big Tech has already upended traditional media models. With AI, that disruption is only accelerating. So, can an ecosystem where Big Tech does not squeeze out all the profits be created? How would that look like? My guest today is Paul McNally, founding director of consultancy Develop AI, teaching user rooms how to use AI. Welcome Paul. How are you today? 

Paul McNally: I'm great. Thanks very much for having me here.

Let's start from the beginning. Public interest media, especially in the Global South, have had a hard time being viable, generating their own revenue to finance their journalism. That was already happening in the age of social media. Why is that and how has AI changed that picture further? 

I think it's fascinating that we've had a sequence with AI now where we have been excited about it and now, we're starting to slowly get sick of it, especially in the social media landscape, just as consumers.  

Journalists have been putting their content on social media for 15 to 20 years now. And I actually think AI is creating something of a reckoning around the question: Do journalists want to put their time and effort into producing content for these other platforms? 

We've always felt a little bit conflicted about it. I think AI is creating a sort of an inflection point where journalists can decide if they want to go all in on this. Do we want to create more AI generated content? 

The biggest request I get from newsrooms around the world is that they don't want to create their own social media content. They want to create a 3000-word article, and they want an AI process to chunk it up and put it on social media. And I think if you speak to people that are good at social media and they have good followings, that is the opposite of how they interact and how they build the following, right? You speak to influencers; you speak to these people that post all the time. They are not just giving it over to an AI process necessarily and just not thinking about the content that they're producing. 

I think journalists, as a generalization, have not felt comfortable or really interested in creating content for social, and they're hoping AI will help them with that. But I think it's going to do the reverse. I think it's not going to create content that is engaging or that people want. And I think the platforms also run a risk if this content becomes too widespread on somewhere like LinkedIn or Twitter or Facebook, then people might just turn it off and not even come back. 

It sounds like what you're saying is that media outlets never really found a way to make the age of social media really profitable for them. And it really does sound like AI is set to make things worse in that vein. Is that correct to say?

I think so, if they keep going down the same path.  

I wrote something for my newsletter a few weeks ago around media literacy – and it isn't that I'm anti media literacy, I do think it's very important in the age of AI – but if we keep doing media literacy in the same way as we have been doing it, I don't think it will work. 

And I think if we approach producing content for social media in the same way in the age of AI, that will also not work. So, I think each aspect of production for journalists is going to take not just a tweak, but maybe a revamping or a complete change of how we operate. 

Again? 

Yes, exactly. [laughs] Well, that's it. If you're in a big newsroom, you feel completely traumatized by the fact that didn't we just do all this like five or ten years ago? 

I thought we were still doing it. 

Yeah, exactly. I think journalists by design are often – me included – people that like systems. They like to learn that system and then they like to stick with it. They do want to learn new topics and do new stories and accelerate their brains in that way. But in terms of how they operate, I don't think they're really innovative. I don't think the industry is by design, and I don't think a lot of journalists are.  

I think journalists, the good investigative journalists, are being pushed into a place where they're not comfortable. People who are more product related, more interested in tech, they are rising up, but you want to still nurture these hardcore journalists.  

The thing I worry about now is with all these new tools and stuff that I teach newsrooms to do, you don't want to lose the enthusiasm of your hardcore investigative journalists because they are working every night. They are working every weekend. And if you lose them, you're going to lose an insanely good resource to uncover corruption and do all the good things that we actually want to happen in journalism. 

I guess the question is how do we keep fulfilling this democratic function?  

There are some really obvious ways that AI and big tech have made life harder or less financially secure for media outlets. Say you're a user: You want to look up an article, you get an AI summary from Google and stop clicking through to the news websites where that information actually came from. Perhaps that really good piece of investigative journalism you were talking about. Are there any solutions to be had there? 

I gave a talk actually the other day for UNESCO around this and sort of "the death of the web", right? It’s sort of been the buzz phrase and it's been kicking around, I'd say for about a year now.  

But it’s only now with those Gemini Google rewrites that really alerted everyone. If your business model is about people going to your website, then they're not going to go there if they can just get it from Gemini straight off the search. 

We are so focused now on doing the job that we're doing today better and faster with these AI tools but that's just not going to be enough.  

On the one hand we are scared of being scraped by LLMs and losing all our content without being compensated for it. I think that's a valid concern. The BBC have released big studies around how inaccurate these scrapings are when they're sourcing their own BBC material. You might look up something in ChatGPT, it'll say this was written by the BBC. But when you click through the article is about that topic, but the quote is entirely made-up. 

But not to the extent that you...  it will pass a courtesy check. So, you'll be like, oh, yeah, the BBC did write about that topic. Cool. I'm sure it's all fine. But actually, if you go a little bit deeper, it's full of mistakes. 

I think we are at this point where we don't necessarily even want to do a deal with the LLMs or the big tech companies producing them. News Corp or The Atlantic were a big get for Open AI when they did a multi $1,000,000 deal. And they've made, I guess, good immediate money from that. But for the rest of us there is a kind of feeling that are we going to have to pull people away into a different space or are we going to have to not think about casual traffic.  

The way I think about a news website is that they've got their loyal readers. We got very intoxicated with the idea of newsletters a few years ago. I sang the praise of newsletters myself and went to workshops about it and got very into this idea of building relationships with the audience. And to some extent, that was entirely true. But it’s also true that not everyone can monetize their newsletter. We can all have newsletters, and we can build this connection, but the conversion rate of financial security from your newsletter is only for a few. That didn't quite work as much as people wanted. 

But the thing that I think we can take from that is this connection with your audience that we've always been talking about. I think AI is going to absolutely accelerate that. If you get, say, 50% of your audience to your website from people that love you and come to your website every day and 50% from people just Google searching, the latter 50% is going to vanish largely. That is going to be the short-term effect, and we need to think about that as quickly as possible. 

That's not really a solution; it’s just knowing that the only audience you're going to have going forward are people that are coming directly to you for what they are trying to find and they're not doing a kind of courtesy search through Facebook or anything else anymore.   

You need to find a way for users to bookmark your page instead of Googling to find you because they won't click through. Is that what you're saying?

Exactly. I know that sounds very old school and very basic. Even the smallest digital platform is now going to realize that no one else is coming. 

Because the thing that I think about is how engaged you are with an LLM when you're chatting to it, right?There's been a lot of talk this week around inappropriate relationships with teenagers - Meta is being sued for that AI. I think it's something that if you are over 40 like I am, it's a little bit hard to understand. This idea that people are having full on relationships with an LLM is something that is going to keep growing in the media and in this younger generation.

And I think when you think about news consumption, that is the absolute opposite of a Google search where you search for things. That gives you 10 items that you can choose from. And then you go to those websites with an LLM, you're chatting to it, even if you're debating a topic or, or whatever it is. And the idea that you're going to click out and read a static website is the complete opposite. No one is going to want to do that. 

There is probably a reality where without clicking out, you get money just to be talked about in the conversation as a news website. I think there is probably something like that. There's probably some kind of agreement that already if you get scraped or if you give your data, there is a financial incentive. But the trouble is: Those payouts have been very few, if any, for Africa, for example.  

I think the bigger sites are kind of motivated by the fact that they're getting some money now because they worry that if they wait five years, they'll be no money. It's not a sustainable business model in any way. 

Are news outlets all going to need to be like ChatGPT? Like perplexity? 

I think that there is a feeling that everyone should have a chat bot, right? And I think that's fine for people to do that. But anyone that has sort of thought about that more in terms of a business sense will realize there is no business model there. You're not getting any kind of financial incentive. It's really like a cherry on the top of the cake kind of situation where maybe you have subscribers and you're like, OK, you can have this chat bot for free and now you can chat to bot which has access to our archive and etcetera. 

Politico came quite early out the gate and offered an absolutely outrageous subscription where you could use a chat bot that was linked to the archive. The idea wasn't just to ask it questions about the news, but to help you with your job, right? So to help you write political PR statements. And within six months, it kind of came out that it was hallucinating at the worst possible rate. A product department had put this in and published this. And the way that that was exposed was from their own journalists who were like, we don't like this. This is exposing our journalism as being incorrect when actually if you scaled back to the actual clicks, the journalism was fine, but the way it was referencing it wasn't. And that was a paid subscription perk for like thousands of dollars a year. 

So, I think we are a little bit far away from being able to do that confidently. On the other hand, I think we also don't think about how people want to consume news. They don't necessarily want to sit there chatting relentlessly all the time. People also just want to watch something. YouTube still grows every quarter – it is the juggernaut of the media landscape in the world. If you look at the landscape of media consumption, YouTube is the winner, right? It won and a lot of people use YouTube everyday just for their news. 

And that is a relatively passive platform. I mean, there’s comments, sure. Google have talked about how they're going to release something where you chat to the video. But how many people really want to do that 8 hours a day rather than just watch something? That’s debatable, right? Also, we have been trained for decades to be passive and maybe we enjoy that sometimes. 

I want to go back to one of the points that you were making about outlets working with or pushing back against, against big AI companies. 

Obviously, you have this issue where they train their models on media content on the web, and then the media outlets themselves lose traffic. So, there is this talk about the "three L" of pushing back against AI companies: legislation, licensing and litigation. You have the likes of big outlets like the New York Times suing open AI.  

But of course, we also have smaller outlets, say in the Global South, that have less leverage, less resources. Where does all of this leave them?

The smaller outlets in Africa are sort of left, in my experience, wondering if there's any money in this. They've seen the big deals, and they are aware of the ethical considerations. They're aware of everything, but they're also thinking, is there going to be a payout for any organizations on the continent as well? I think that is keeping people interested from talking out as much as maybe they should.  

I'm actually speaking at the M20 next week in Johannesburg around AI and media. And luckily there is this growing swell of small organizations really wanting to kind of bandy together and work on the legislative side. It’s quite encouraging. People are really trying to figure out a model to put laws in place which will protect them not just from a disinformation or harmful point of view, but also to protect journalism and their business models. 

So, there are things happening in Africa for newsrooms of all shapes and sizes at a higher level. I think if you are a small organization and you're not connected to that advocacy; you’re left a little bit in the wind and you're not quite sure where you should go. Like I think you want to be ethical and you're aware that your audience wants you to be ethical. 

But I think a small news outlet is conflicted because where the money is often coming from is from Microsoft, or the Google News initiative. There’s a huge multi-million-dollar grant for African newsrooms by Google that's rolling out in community media just this week. And I think you want that money. You need that money to survive. 

I often think when I talk about ethics that it’s a resource game in some ways. The more resources you have, the more like you're in a luxurious position to be more ethical. It's like eating well or being more environmentally friendly. You can get very wrapped up in it as an ethical issue, but it's also a resource issue. And I think it's the same for news to some extent. There’s not a newsroom that I've ever come across that wouldn't want to take the ethical route. But they also are aware that there's cash on the table from these big tech companies that could transform their businesses.  

So, let's pretend you have like this golden green fielding opportunity, right, where you could create a profitable newsroom from scratch that's also fit for this age of AI. How would that look like? [laughs] I feel like I should be paying for that insight. 

Yeah, you want me to solve media now in the next two minutes. [laughs] 

Exactly!  

So, if we take say, South Africa as an example, where I'm based, the capacity for people to pay for news here is so low. The major news platform, News 24, charges barely $5 a month, and you get access to a huge amount of content and news for that. And everyone else that's tried to paywall anything hasn't really succeeded to some to any extent. 

So, I think, if you're talking about a fantasy operation, the idea of charging your audience even in a substack model, I don't think would be the way to go. And from what we've seen in terms of Trump's proclivities, the idea of being donor funded does work for a lot of newsrooms, especially if you're in exile. I'm busy coaching a whole bunch of newsrooms for Thomson Reuters who are in exile from Russia and Belarus, and they are all donor funded. 

I think it's naive to say you have to always get away from donor funding entirely because that does exist and it will continue to exist, especially if you're in newsroom in exile or you, you you've got safety issues. But I think you can realistically say it's not going to exist for everyone depending on the work that you're doing. 

I think, the way that you would think about your business model in terms of not relying on those two things, I think advertising can still work. But I think you will have to pull away from the traditional models on Facebook and Google. And I think that's going to take years of training the audience in where they go for news. 

News outlets were complicit with training people to use the social platforms. We helped them do that, right? We linked to Facebook as a little icon on every news outlet in the world and that was just free advertising for this big platform that we did for years because we wanted them to go to our Facebook page. And I think we will continue to. 

But speaking of an ideal business model, you would not do that. You would try and pull away and make something as a community. If you think about the media development community, they could probably get together and think about a platform where people put news and people could go to that for news. That would be like a training exercise: You'd have to train your audiences to know that that's where they go for news, globally. A social media platform for news could be something that exists if it was communally owned by all the top 6 to 10 media development organizations.  

So, I think we can't keep playing to these people and advertising for them. I think you should take the money if you can get some money out of any of these people. But I think the idea that they're going to be there for us in five years just because they scraped our content and people might see a random link to you in a Google, in a chat with an LLM, I don't think that's a path to profitability at all. 

Taking that vision that you described, what kind of regulatory environment has to exist to bring that into being?  

Personally, I'm very pro regulation. I think there's good from that. What we've seen with the EU’s AI Act, is that you could debate certain parts of that legislation as being good or bad. But at the same time, I think, the failing of it isn't really its fault. It's the fact that it's not global but these technologies and platforms are. 

I think the failure with big tech and media was the lack of regulation. Even with the monopoly stuff: The fact that we allowed Meta – before it was even called Meta – to buy all those other apps and grow them into this huge monopoly and then try and break them up after the fact, which obviously they refused to do. I think it's a regulation issue and that's clear. 

The trouble is that getting the world to all work together hasn't really ever worked for anything. But when it comes to tech and the Internet, you can or should try and do that. 

But I'm also not oblivious to the fact that regulation can be turned on its head. You often find countries where they've created laws about hate speech that are supposed to protect people but where governments twist those around and push them against journalists. They’ll accuse journalists of using hate speech and try and imprison them based on that.  

That kind of leans me off into this idea of not doing it at a legal level, but at a self-regulation level. I think the most successful press codes haven't really been laws, right?  They've been rules that huge amounts of media have agreed on and kind of told their audiences that people who are not adhering to these rules are bad journalists. 

When I think about press codes example, like the audience has kind of lost that. They don't often know if they're part of a press code or what that even means or if that's a good thing or if they should even trust the press code. I mean, people are more suspicious now, right? So, they're not sure if this press code that they're meant to be considering a good thing is actually a good thing. 

Around AI, in most workshops that I do, if you bring in a bunch of newsroom representatives from across Asia or Africa, there is usually an end result where people say let's all bring at least our editorial policies around AI together and start building that out into some sort of agreed idea of what we're doing. And I think it will take big organizations to do that. I think you need to work on both sides, but I think we can control the self-regulation side where the actual legislation side is more complicated. 

Just as an aside, I find it interesting that you said people are more suspicious now. Do you find that they're more suspicious or that they're more credulous? Because this is one of those issues that we talk about in terms of whether users can identify what is good or bad information, and how AI could exacerbate that. 

I think people are more suspicious to the point where they've had almost their trust in information shattered. 

And if you talk to news organizations, they take the shattering of truth and building truth very personally. They think about it in terms of them and their relationship with their audience members and they've somehow failed or this connection has been lost, and it somehow needs to be rebuilt. 

That's admirable to think of it like that. But picture this person who's just consuming content from all sorts of directions. And as a result, they don't really trust anything. It's got nothing to do with you! It’s a "it’s not me, it’s you" kind of thing. 

We need to kind of take people at that value to some extent. We got to be like, OK, these people don't trust anything. Is this something we can fix in a person just as one news outlet when they might be coming into contact with 50 different outlets a day? Probably not. 

There was one more question that I had around regulation. And it has to do, again, with the outlets that you have from the Global South and the fact that the big companies that sort of determine the shape of the Internet, they're all in the US, famously not the biggest proponent of regulation. What effect does that have on the ability to create regulation that would apply to everyone? 

When I speak to newsrooms, especially in Africa and Asia, they are always thinking about how they want to make something for them. There is a kind of pull back to this idea, not just with regulation, but even with technology. People say, OK, we don't want to be just giving money to these big corporations all in America. We need to build one for Africa. We need to build one for Asia. There seems to be that sentiment. 

And now, you realize how much money, how much resource that really takes. There are obviously incredible start-ups that are doing amazing things around languages and building language models. But the low hanging fruit of that, well, it's still quite high. But the lower hanging fruit of that is the sense of building an API that can kind of plug in that can help you with certain languages. And I think there is amazing work being done with that and then newsrooms kind of utilizing that. 

But in terms of people creating almost like a walled garden for themselves away from the US, I think that’s not possible. Even if you wanted to do that, say you could regulate everything in Africa and it would all work perfectly and it would be separate, that wouldn't help, right? 

On the one hand, you don't you want it to be global. And on the other hand, there's still intense bias that favors white men on these LLMs. If you ask for images of someone in authority, an LLM will give you a white man because it scraped off the internet and the internet is incredibly biased not just towards English, but racially and according to gender as well.  

It becomes so obvious when you're in a newsroom, say in Kenya where was a few weeks ago, that they are battling with that bias, but are kind of having to live with it. And the reason why you don't want that walled garden LLM or technology is because people outside of, say, Africa won't even know that bias exists, right? If you let these things out just into Europe and the US with these biases that aren't being made less biased by input from Africa and input from Asia, they will get worse.  

I would say every African newsroom has had someone in there has thought we need to pull up the gates, we need to become more insular. And invariably it's not possible financially, but I don't think it's good for the greater ecosystem.

This has been fascinating, and we do have time for one user question from Aaron Tungaraza from Tanzania: "If AI can already perform many newsroom tasks faster and cheaper, what real incentive does major media outlets have to keep investing in human journalists beyond protecting quality and trust?" 

I'm glad you asked this because it comes up quite a bit. 

And I think it comes to the sense of like if you want to do things faster and cheaper or if you want to do things better, right? If you survey people in an open forum where they’re seen, they will always say I want to do things better. But if they've got it on a secret ballot, they might say, actually, I just want to do this much cheaper and I want to do this quicker. 

But why newsrooms can't do that is because they are ultimately competing with every single other digital person on the planet who also has access to all these tools. The gatekeeping around journalism and technology has been going down for a long time and it is now completely at Ground Zero. 

So, if you're an outlet that just goes, we'll do everything cheaper and faster and fire half our staff, then you're not going to have an organization very for very long because you're competing with a guy on Instagram who's using all these AI tools and actually probably making maybe very good content just on his own.  

So, you need to think about how many people and resources you have, at least for the time being. How do we use tools to make everything quicker and cheaper on the one side, but then also bolster the quality of the content in such an extravagant way that it blows away all these one-man shops? Because – with a reasonable number of exceptions – influences and people on YouTube are just one man shows.  

I think it should be obvious for a newsroom to understand that even if everyone's job must change in your newsroom to some extent, you must double down on producing much, much better content. The idea of old school journalism, of uncovering something. People still love corruption stories. People still love holding people to account. People love that stuff. It does work well in this environment. It has just become so prohibitively expensive. So hopefully you can use AI tools to help with those parts of your production. And do more of them and kind of make, you know, the world a better place. 

Paul, I think you did it. I think you solved the media in the age of AI. Thank you very much. 

It's a pleasure, Yeah, I'm glad there's a lot of pressure, but I'm glad it worked. 

Paul McNally, founding director of Develop AI! Until next time on Survive and Thrive, your media viability podcast. 

This transcript of "Survive and Thrive" has been lightly edited for clarity. 

Get in touch    

For questions and suggestions write to dw-akademie.surviveandthrive@dw.com.

Or find us on LinkedIn and Youtube.    

This podcast is produced by DW Akademie and is supported by the Federal Ministry for Economic Cooperation and Development (BMZ).