What next for AI and factual TV?

Much of the factual TV industry initially saw Artificial Intelligence (AI) as an existential threat, but now that companies have started adopting it into everyday working practices, is it turning out to be the sector’s saviour?

For many companies, AI is now gainfully employed in cutting back on the repetitive, boring everyday tasks of film and TV making, such as logging rushes. There are gains in development too. AI has dramatically increased the quality and speed with which materials for pitch decks and sizzle reels can be put together.

It can also help with storyboarding and scriptwriting. Ask a language model like ChatGPT for ideas around a doc and it can kick back multiple angles. Early adopters report that AI helps them think ‘quicker’.

German high-end doc producer Beetz Brothers, whose film Eternal You explores how companies can use AI to create avatars of deceased people (which aired as part of the BBC’s Storyville strand in 2024), has incorporated the technology wholesale.

“The first step was to use AI tools in post-production, especially in sound editing and colour correction,” says chief executive Christian Beetz. “We can use it to source more varied footage, which is very helpful in archive-based documentaries – we don’t need expensive recreations any more.”

Mafia Hunters, Beetz Brothers’ flagship 2025 project for ARD and Arte, also used AI in the post-production process.

AI and Machine Learning (broadly seen as a sub-set of AI) can also provide audio transcriptions, saving editors considerable time, while visual effects are also becoming more automated.

“VFX artists and VFX companies that are adopting these things and learning how to integrate them into their pipeline may be future-proofing themselves,” says director Paul Trillo, who spoke at the recent AI Creative Summit in London, organised by Broadcast and sister title Screen International. 

Late last year, he released a music video for the American band Washed Out that was generated through Open AI’s text-to-video generator, Sora.

Trillo – granted access to Sora prior to its release at the end of last year – says the lines between computer-generated graphics and real life are blurring. “Sora is very good at doing chimpanzees,” he notes.

“Currently, CG chimpanzees still kind of look like CG chimpanzees, even though the technology is very good. But the chimps that I’ve generated with Sora are undeniably good.”

Lost jobs

Improvements such as these come with a cost, of course. The expected growth of AI will lead to jobs being lost, particularly in post-production, and there are concerns that it will also impact entry-level posts.

Yet new jobs are being created, such as AI prompters, and there is a feeling among some that the ongoing shortage of work – largely due to the overall contraction in commissioning – has been mistakenly attributed to AI phasing out jobs.

“The demand for professionals skilled in AI technologies is growing,” says Nikola Kohl of Munich-based premium doc and factual producer South & Browse (part of All3Media). “While some roles may vanish, new opportunities will emerge, allowing creatives to focus on higher-value tasks.”

Neil Barrett, a British cinematographer who has worked on more than 100 documentaries for the likes of Netflix, HBO, National Geographic, the BBC and PBS, is also pragmatic about the role of AI and its impact on jobs.

“In an average documentary day, you will probably go to three locations, film complex human interactions while trying to minimise your presence and make complex editorial decisions [that may have social, political or cultural ramifications],” he says. “That’s not to mention the fact that many subjects would not be comfortable being filmed and followed by a robot.”

“There are areas where jobs will be hit, but we’ll also get productivity gains”
Neil Barrett, cinematographer

The technology, he suggests, would have to be approaching AGI (Artificial General Intelligence - a currently theoretical area of AI whereby software can perform any task a human can) before it affects the role of an experienced DoP.

“There are areas where jobs will be hit, and there are other areas where we’ll get much-needed productivity gains,” says Barrett, who is now designing software and strategies to help the cash-strapped and resource-limited documentary sector thrive in the age of AI.

Media coverage of AI has also arguably blown up perceptions of just what the tech is capable of. “While AI simplifies many tasks, achieving perfect results often requires multiple iterations,” says Kohl.

She has most recently been working on Unearthed – The Mystery Of The Shaman Woman for ZDF’s Terra X and Arte, employing graphics software such as Unreal Engine 5.0 and MetaHuman to create an accurate Stone Age avatar to appear in the show.

“The combination of different tools is key to achieving the best outcomes,” she explains. “AI tools are powerful, but the artistic vision and nuanced understanding of context still rest with experienced professionals. The best results come from collaboration between advanced AI and human creativity.”

The current consensus is that AI is best employed as a creative ‘co-pilot’, whereby individuals or companies explore how AI is most productive for their specific requirements.

Marc Smerling, creator of HBO’s groundbreaking true crime doc series The Jinx, outlines his experiences. “I’ve only used [audio editing tool] Descript, but it was a gamechanger,” he says. Smerling has also “played around with software that does rough edits to pull selects”, the results of which he describes as “bumpy”.

While the applications of AI in production are myriad, the spider’s web of IP copyright issues is arguably yet more complex to navigate. Who owns

AI-created images, for example? The prompter, or the company who trained it, or the creator of the original image ‘scraped’ from the internet? Will fractional, multi-ownership of images be the new normal?

Fair use

Answers to these questions have not yet been thrashed out, but there are multiple cases in the US courts pertaining to what is deemed to be ‘fair use’. Elsewhere, countries such as China have a very ambiguous take on the issue, meaning that outsourcing to countries where laws are different may create further legal complications.

Until these matters are resolved, studios and platforms are likely to remain reluctant to use AI – outside of development and post-production – for fear of being sued.

But in the Wild West world of social media, individuals are putting their own content on platforms that take limited responsibility for what they host. It is no surprise that YouTube has experienced exponential growth in docs and podcasts.

“What I see a lot is the melting between audio non-fiction and docs – mostly in interview shows on YouTube. You’ll see podcast interviews being shot with two cameras put through an AI editor, touched up by the interviewer and posted,” says Smerling, whose true crime podcast Firebug is being adapted for Apple TV+ with Dennis Lehane (The Wire) as showrunner.

“I met one kid who travels around the country interviewing different people in the crime genre. He goes back to his hotel and feeds the interview into an AI editor and after about 30 minutes he tweaks the cuts and posts,” Smerling says.

“It’s his whole life. He says as long as he keeps posting, his audience grows and he makes money from YouTube and Patreon.”

Fake news

It is in this less regulated space where there is the strongest fear that AI could be an existential threat to the factual sector, with the technology having the potential to turbocharge the distribution of fake news and misinformation.

“You can make things quickly and sloppily and add visuals quickly and sloppily – no factchecking, legal or journalistic review. It won’t take long for docs to come out that promote pure BS,” says Smerling.

In the non-fiction world, realism is currency. Maintaining trust with the audience and being as transparent as possible are paramount, yet proof of ownership has arguably never been more important.

“You have to authenticate. It’s a simple concept but it’s sort of brilliant – if I can prove that I took a photo and it’s real, that’s going to matter going forward,” says Barrett.

Organisations are setting up processes in this regard. The Content Authenticity Initiative – whose partners include the BBC, Nikon, New York Times, Wall Street Journal and Microsoft – has created an open source standard (C2PA) that provides internal and external transparency for content in the age of AI.

“It creates a kind of nutrition label for a film, telling you what’s in it. So the network or platform is aware of the source of the content, and it also lets the audience know,” says Barrett.

“If I take a picture in the field with my camera, it says ‘taken by me at this time and date’. Every change to that image is recorded and stored in the metadata so there’s a fully trackable record of the image’s authenticity. It’s going to be as common and useful as time code.”

This enables platforms and outlets, particularly public service broadcasters such as the BBC or PBS, to stand for integrity and reinforce their status as trusted sources of information.

Respected brands that wish to build elements of trust into the supply chain are also being proactive.

“Companies such as Adobe and Moonvalley are focused on creating clean, fully licensed data sets on which to train their Gen AI models. You cannot be sued because their models are trained on fully licensed stock footage,” says Barrett. Lionsgate has also struck a deal with Runway to train a model customised to the studio’s portfolio of content.

Clearly, factual producers need to embrace AI to avoid missing out on the once-in-a-generation opportunities that the tech offers, and to ensure they don’t lose further market share to the social media space. Yet they must also maintain the basic values of non-fiction: integrity, trust and transparency.

“AI technology must be bent to the good,” says Barrett. “I don’t want to be a bystander on this. I want to make sure this industry that I love survives in the age of AI.”