Meet the viral journalist whose AI app seeks to combat the rising tide of disinformation
Sophiana app empowers journalists and experts to make vertical videos in minutes
MULTI-AWARD-WINNING journalist Sophia Smith Galer believes she has an answer to one of the most pressing problems of our time: rising levels of mis- and disinformation on social platforms. Her aim is to encourage a tsunami of trustworthy vertical videos that will drown out the slop. But there’s a problem: making quality content that the social algorithms reward takes time and levels of know-how that many fellow journalists and subject matter experts simply lack.
Step forward Sophiana, the AI-powered brainchild of Smith Galer, whose pioneering TikTok videos while at Vice News and BBC News won her the innovation gong in the 2021 British Journalism Awards. Further recent awards have cemented her international reputation as one of the most innovative viral content creators of her generation.
Within minutes of downloading the app I’d made my first 43-second video ready to engage audiences on TikTok, Instagram Reels, YouTube Shorts and LinkedIn. As a test I uploaded our Quick Take on the US judge who sided with Anthropic in his fair use ruling — a complex story in which Sophiana detected three key hooks that would grab attention. I selected one and, as if by magic, there was the distilled script plus captions and hashtags. And below that, a button inviting me to film my script. Clicking on that booted up my camera and an on-screen teleprompter, and I was all set to record. It was that simple.
Sophiana launched this week as an iOS app having been user-tested earlier this year at the International Journalism Festival in Perugia. As Smith Galer told me, Sophiana is her second generative AI project as an innovator. The current version has been developed by her digital start-up Viralect working with direct-to-consumer (D2C) video app and AI specialists Leo J. Barnett and James Yorke. Right now OpenAI’s GPT-4o and a core prompt she developed are doing the legwork but as Smith Galer explained a future version will be trained on her scripts which have fuelled over 180 million video views.
Listen to the full interview (with transcript below) in which Smith Galer reveals:
How she built Sophiana and why you should be using it.
How as a journalist and author she uses generative AI.
Her ambitions for her AI creation and where it’ll go next.
I started by asking what was the problem she was seeking to solve?
Smith Galer: A lot of journalists are still struggling to convert their written work into vertical video that can fly online. That was important six years ago when I made my first TikTok but it is even more important now. We’ve seen only in the last couple of weeks from the Digital News Report from the Reuters Institute for Journalism that social media has taken over from TV for the first time ever as a source of news and information worldwide, so we really need to be in a position where journalists and anybody with expertise and high-quality information to share should be able to put content out on these platforms. Vertical video took over the internet and it’s a particular kind of vertical video. It’s one that puts people at the front of it speaking directly to camera. A lot of news publisher content — because of course, lots of news publishers are theoretically on Instagram and TikTok now —a lot of them are putting out what I call identikit news B-roll which gets lots and lots of views but doesn’t really do much to build community long term. And really the people who are delivering most peer-to-peer, speaking to camera vertical video are content creators with zero journalism experience. So the problem I’m trying to solve is this is digital territory, we should be engaging and more journalists should be speaking to camera and turning their work into video that performs well.
What do you think has held them back? Why aren’t journalists doing this? Why aren’t they embracing this, as you have done?
I think there are two pots of journalists. Let’s first take the journalists who sign up to training of mine because I actually surveyed over 100 of them back in 2023 to try and figure out the answer to this question. And I was wrong in some of the preconceptions that I had. I assumed that maybe some journalists aren’t doing it because their bosses told them not to, or perhaps some journalists aren’t doing it because they just don’t like this kind of stuff. And actually there were two major reasons that came out in the surveying. One is that the journalists felt like they didn’t have enough time and the other is that the journalists felt like they didn’t have the required video skill set. And I completely understand this because when I made my first ever TikTok back in 2019 I did it while I was a BBC video journalist. I was in the perfect position. I had the perfect skill set to feel comfortable giving it a go. I do have to say that there is another pot of journalists who still believe that content on social media, especially video, tends to prioritise mis- and disinformation. It tends to be wholly occupied by an influencer class who simply want to sell us stuff and that it’s not the place for journalism. And that is only an attitude that comes top down and is extremely hierarchical because that’s not what audience data tells us. Audience data tells us that social media has become an overwhelmingly important and growing source of information and news, and that is what I am always guided by as a journalist. It’s where are the audiences, then finding news and information to be the most accessible, and am I putting my stuff there? If I’m not there’s a problem.
I remember last year when we first met in person that you were talking then about playing with your own ... was it a small language model, it wasn’t a large language model I don’t think. But you’d already started to play with AI. Did this idea come out of that?
Yeah, it was when people were first starting to talk about AI that I thought, ‘OK, this is a fun new shiny thing that I’ve never made myself before”. I’d been aware of it in the ether and I’d also reported on its misuse, that was something I had done as a tech and platform accountability reporter at Vice. And I think this current AI revolution, where it’s different, it’s that people without technological backgrounds are now being empowered to try things out. So I first had the idea for this back in 2023 and as someone without a coding background I used a third party no-code tool to build what I was limited to do which was a web-based chatbot. So I made a kind of version of this that just lived online and was a chatbot that people could pay for monthly access to. And I did it to see whether the idea would work, like in theory could a chatbot help someone turn their work into a video script as if I were there in the room with them. And it did, it worked, but it used a third-party tool, it was on a website and really this needs to be something that was on the phone, and was all mine, that didn’t have to rely on any third party anything. And building technology takes thousands and thousands of pounds. So, I was like, ‘Cool, I’ve proved the idea but I can’t take it any further at this stage, so I’ll park it’. And I am simply very fortunate that in a space of a couple of months I won the Women in Journalism Georgina Henry award in November and by January I’d also won a spot on the Disarming Disinformation Solutions Challenge from the International Center for Journalists in the US who gave me additional AI training, took me to Istanbul to do a design sprint where I built the first simulation prototype of Sophiana. And we did a pitch at the end where we competed for funding, so between the funding I got from that and the funding I got from that Georgina Henry award I was able to have a budget that I could hire a developer and an engineer to actually build this. I hired Leo and James and they’ve done a fantastic job.
They have indeed. In that process then was it trained on your own TikTok scripts, was it trained on how-to scripts? How’s it actually been trained?
The first ever tool that I made was trained on my scripts and Sophiana will be trained on them again. In order to get the scripts in the perfect way that I want them to I actually have to do a lot of data preparation so the next whatever instalment of cash whether it’s revenue or funding that’s one of the key priorities is to fund that data preparation so that the scripts can be part of it again. As it currently stands it’s a core prompt that is running everything and that core prompt has been tested endlessly for about six months and it’s based on my knowledge base.
Just for other people haven’t heard that phrase before, core prompt, what do you mean by that?
Yeah so AI tools a lot of them work from prompt engineering which many of you will have done if you habitually use ChatGPT when you’re speaking to it, and you’re trying to improve the answers that ChatGPT gives you. There’s a sort of a special way of speaking to it in order to get it to deliver what you want, and so a lot of tools like mine will have a core prompt in which it’s given a full-bodied, tight, incisive instruction that makes sure it does what you want it to do. And within that you can put in proprietary knowledge and that’s what I’ve done.
I see. And the teleprompter. Was that a brainwave of yours or is it something that exists elsewhere? I mean that is just ... that’s a breakthrough!
As far as I’m aware there are teleprompter apps on the market but none of them will help you write the scripts, they are just for you to input text on. And Sophiana can also be used like that, so if you’re someone who maybe you are very advanced in scriptwriting and that’s not what you need help with, you need help in the filming and the delivery, you can still use Sophiana, you just put the text into the teleprompter. So that’s unusual for it to be both a scriptwriting assistant and a teleprompter. I wanted it in there because I’m trying to reduce friction. When those journalists were telling me that ‘We don’t have time, we don’t have the video skills’, and they’re talking about all of these obstacles that are set up in front of them that mean they can’t make a video. And if in one app you can do the script and then film it straight away that’s reducing friction and reducing time.
And it’s not just journalists that you’re aiming it at, is it?
No. Obviously I am a journalist and most people who know who I am are journalists, so it’s naturally the first market to expand to because I’ve witnessed this desperate need within journalism and especially the news media space to really get a lot better at countering misinformation online with high-quality engaging video and amplifying our work. I think, and I know, that a lot of news and information today is not coming from news media. I for example get most of my news about feminism and women’s health from a really good GP and a really good gynaecologist that I follow on Instagram who know how to make video content. So I’d love to see this in the hands of experts. I’d love to see Sophiana in the hands of doctors, in the hands of people in academia who have written research papers that they’re desperate to amplify. I also think that there are people who are offering amazing solutions and they’re trying to change the world every day in their business but people don’t know about it because they’re not making video. So I see a lot of use cases.
As a journalist yourself what tools are you using? How are you using generative AI?
I think [like] a lot of journalists we’re looking for solutions that help us do some of the boring or sort of tech-heavy bits more quickly, more accurately, and preserving what we believe should be human-based. I think it’s actually quite interesting looking across the portfolio of work that I do. I’ve just finished my second book and not a single word of that second book has been put into anything connected to AI. That’s a part of my work that absolutely must be my original thought, and the time investment, the thinking — that all has to be me. And that’s quite apart from that being what I in my heart and mind feel, that is a sensible legal choice because the book is my copyright and you can’t hold copyright over stuff made with AI. On the other hand, ‘emailageddon’ right, and the fact that I am trapped by my own email inbox every day. I manage three different email inboxes — a personal one and two business ones — and on one of the business ones I started using an AI email assistant, and it’s transformed my working life frankly. I’m spending I reckon at least an hour less a day writing emails because what it’s doing it’s learned how I write and it’s presenting me with draughts.
OK, final question: where are you going to take Sophiana next? What’s the development plan?
Well first of all people need to use it and enjoy using it! I would love for Sophiana to be used in newsrooms. There is an option for newsrooms to work with me to effectively have in-house Sophianas that can do a little bit extra compared to the D2C app that’s available for people at the moment. I would love Sophiana to be on Android, at the moment it’s only on iOS. I want Sophiana to be multilingual, at the moment she is English-only. The priority really for me is just to see people using it and using Sophiana to elevate and amplify factual and high-quality storytelling vertical video.
◼️ Sophiana is available via the Apple App Store.
