2025: Making Sense of a Year in AI
Not to be indelicate, but shit a brick. It’s been a right old year. Apologies for being crass but a) have you seen 2025? Totally warranted. And b) how else to demonstrate the mix of shoddy executive function and utter exhaustion that is one hundred percent me and not the confabulation of an LLM server farm.
If 2023 was the year that AI brought a shonky new toy to advertising and generated a bunch of wonky experiments, 2024 was the year that the industry started to have lots of lengthy, serious-sounding conversations about the not-to-distant future. And 2025? 2025 was bananas. Nano Bananas, in fact.
This year, artificial intelligence (and generative AI in particular), came out of the shadows as more brands gained the confidence to go public with the use of AI in their ads. No longer the odd shot here or there, or limited to targeted digital ads, AI has thoroughly embedded itself in production for many major brands.
Towards the latter half of the year, AI-generated and AI-enabled commercials proliferated. There were the publicly debated campaigns for Coca-Cola and McDonald’s, of course, but also brands like Fiverr, Wizz Air,IAMS pet food, Longchamps, Lamborghini and more.
Tech Acceleration Reaches Terminal Velocity
The AI video arms race heated up in 2025. The headline fight was between Google’s Veo 3 and Open AI’s Sora 2, with both AI video generation models integrating synchronised audio into their tools and offering up significant leaps forward in various aspects of physics. Sora 2, Veo 3 and Midjourney V7 all introduced character consistency support - though wrangling perfectly consistent characters from scene to scene isn’t without challenges. Sora 2, which bundles together video generation tools with its own social platform, is only available in the United States, Canada, Japan, Korea, Taiwan, Thailand, and Vietnam -- data privacy regulations across Europe are slowing its rollout there.
While most platforms still require users to work with fairly tedious language-driven prompting, Runway Act Two launched in July, letting creators use simple, real-time performance capture to drive their generated characters directly, giving immediate oversight of expression and movement. And in September, Luma AI released Ray 3, a tool that allows for multimodal input and that uses ‘reasoning’ to exert greater control (and to allow filmmakers to draw instructions around framing, camera movement and character placement).
But it wasn’t all about video. Google’s Nano Banana Pro launched in December, turning out images that even seasoned AI-watchers struggle to distinguish from reality.
The year closes out on a ‘code red’ at OpenAI, in response to developments on rival Google’s Gemini 3, a sign of just how quickly things change in this madcap arms race. It also added air to the AI bubble chat that had been murmuring along for the past few months, as incestuous investing practices and valuations that outstrip the promised productivity gains raise questions… but that’s a 2026 problem.
Production Finds a Hybrid
As more producers and directors get their hands on AI tools, the reality is proving to be more complicated than you might think. The popular misconception is that you chuck a prompt into a blackbox AI platform, and out pops an ad -- for those working with AI, that characterisation is the source of much frustration, just take this piece from Made By Humans’ Jesus Plaza.
Each project requires its own workflow, pulling together different tools – AI-driven and more traditional. Perhaps as technology advances further, the balance between traditional and AI tools and processes might change but for now, many producers are favouring an open, hybrid approach. In the VFX world, always an early adopter, people are using AI or machine learning tools to speed up aspects of the craftlike lighting CG objects. At Whitehouse Post and Carbon, they say AI is a ‘turning point’that allows artists to focus on the creative aspects of their work.
That hybrid isn’t just about tech, it’s about behaviour and communication. It’s about weaving in familiar steps and stages from traditional production into even fully AI productions, ensuring client, agency and maker are all on the same page. As Toby Walsham and Guy Soulsby from Made By Humans explain, this means that steps like casting or location scouting can be just as useful as virtual exercises.
This Lamborghini film from Televisor Studios and director Iryna Nalyvaiko, demonstrates how AI and classic CG and 3D modeling can work together. While AI was used for previsualisation and planning, the 3D modeling was used for the car to maintain control. AI was used to handle texture and mood.
Director Julie Reali suggests that success with AI will be defined not by being the fastest, but by cultivating the taste and understanding to know when and how to use it and -- just as crucially -- when not to.
Getting Real About Artificial Intelligence
The AI hype was unavoidable but beneath the noise, as more people got the chance to get hands on with AI tools, the genuinely useful applications became apparent -- as did the tech’s current limitations. At Ogilvy EMEA, for example, where 80% of the people have adopted AI, they’ve found countless use cases, from allowing strategists to work through and ingest more research to testing scripts with virtual audiences. But in their research, they’ve found that there’s one thing that AI is just terrible at: coming up with creative ideas.
Agentic AI was one of the buzziest of buzzwords this year -- AI agents are systems that can perform specific somewhat complex tasks. Across the industry, everyone from new business leads to strategists to UX designers to producers have been training agents to speed up tasks. However, taken in the round, there’s some evidence that AI investment hasn’t been quite as lucrative as businesses might have hoped. The State of AI in Business 2025 Report from MIT suggests that 95% of businesses are seeing little financial return on the tens of billions businesses have invested in AI. AI pilots are stalling and not scaling. However the report does note that media is one of the two industries that has seen structural change due to AI (the other being tech). So while AI may not be having the impact tech giants had hoped for in other industries, it's impact in advertising and production can't be written off so easily.
Human oversight, it also turns out, is vital, a lesson several organisations learned to their expense in 2025. Deloitte had to learn it twice. On two separate occasions, they produced government reports that contained fake, hallucinated citations to academic research that didn’t actually exist -- once for the Australian government, the other for the government of Labrador and Newfoundland in Canada. It’s not just Deloitte that’s been left red-faced, law firms, news publications and even US government departments have all been caught out with fake citations.
The Ad Business Leans In Hard
Inevitably, AI loomed large over the holding companies. In a year where share prices were downwardly volatile, the hold cos have been keen to demonstrate to shareholders and the markets that they’re ahead of the game when it comes to AI.
For Publicis, who announced a €300m investment in developing their Core AI platform and AI capabilities in 2024, that early spend seems to be paying off, with strong revenues and a bullish confidence from CEO Arthur Sadoun that they’ll “outperform the industry”. Judging by their surreal AI New Year ‘s greetings video, starring Edith Piaf and founder Marcel as an AI lion (who takes on the Nazis no less), they’re positively giddy.
WPP brought tech expertise right to the top of their organisation with their new CEO, former Microsoft exec Cindy Rose. Their controversial but daring move has been to launch WPP Open Pro, which grants client subscribers use of the WPP Open AI platform (which bundles together tech from top suppliers), without requiring that they pay for agency services. The goal is to create tools for in-house agencies or SMBs that would not normally engage with agencies at the size and scale of a VML or Ogilvy.
Omnicom, well, Omnicom’s been focused on bringing the IPG acquisition to completion, but a key part of that deal was IPG’s data business Acxiom. Bringing together data expertise and billions of identities, Acxiom will help power Omnicom’s AI platform Omni and, it’s hoped, give it an edge in AI-powered personalisation.
It’s not just the holding companies that have been shoring up their AI capabilities and offerings. Silverside, the AI lab from Pereira & O’Dell has partnered with business consultancy Vivaldi Group to unlock new opportunities.
And on the independent production front, there’s been a fair bit of activity. New companies have launched -- the likes of Anima, and an AI Candy in Belgium and one in Australia. Existing production companies like The Sweetshop and Knucklehead have launched AI divisions. And then there’s players like Made By Humans, a spin-off from production company Imagine This, which has been in the AI space for nearly two years but has seen a marked uptick in business in 2025.
One topic to keep an eye on in 2026 is the impact of AI on the industry’s talent pipeline as it’s seen as a tool that can take on the brunt of repetitive tasks often carried out by juniors. In some cases, such as WPP’s creative technology apprenticeship, there’s a chance to throw new entrants straight into the wild, exciting, experimental stuff. But in an age of ‘efficiencies’, these junior roles could appear an easy cost to cutand that repetitive work is where experience is earned, and instinct is honed.
An industry that believes in its future will be looking to nurture tomorrow’s tastemakers and AI-wranglers.
AI Dons a Human Face
They say there’s no such thing as coincidence, so the fact the AI giants all decided to lean on old fashioned live action storytelling for their big brand campaigns may be less fluke and more an indication that all of these platforms feel the need to earn our trust.
Rather than sample their own merch, or make a big play about the product promises of their gen AI tools, they’ve got a bigger strategic game to play. Perhaps using gen AI in their brand marketing at this point in time works against their mission to gain widespread societal acceptance. If other brands can shoulder the reputational risk of putting out AI ads and, in the process, help normalise AI content in the eyes of the public, well, double bubble.
September saw Miles Jay jump behind the camera for OpenAI’s ChatGPT with indie agency Isle of Any and Meta with Droga5; Daniel Wolfe and Mother positioned Anthropic’s Claude as a tool to amplify thinking with a grainy arthouse film; and in November, Perplexity leaned into human comedy with director Luke Allen-Thompson of Sata Studios, with some help from Lewis Hamilton and Eric Andre..
Budget Friendly with a Side Helping of Backlash
More brands did start to go public with their use of AI, particularly in the second half of 2025. For some, like Maserati and Autotrader, the novelty of AI helped convey a message of tech-forwardness. Many brands found that, whether making fully AI ads (like the aforementioned Autotrader or this Belgian milk company) or weaving AI-enabled shots into more traditionally-produced spots (like this Allegro ad from Poland), the technology allowed them to stretch their money further, whether that was making more executions or telling wilder, wackier stories than their budgets would normally allow for.
And for the most part, this work passed without too much comment, bar the odd industry nitpick. Mostly. But those campaigns that did find themselves in the spotlight… oof. It seems the bigger the brand, the bigger the risk. McDonald’s in the Netherlands found that out the hard way with their Christmas ad, which was torn apart and subsequently pulled. To an extent, agencies and marketers alike should understand the lay of the land and be prepared for potential negative reactions. If you’ve signed it off and it gets a negative reception, to an extent, suck it up. That’s the trade-off. Coca-Cola, for example, has fully embraced it. Their 2024 AI version of the classic Coke Christmas trucks ad attracted scrutiny and derision (Silverside AI producer Rob Wrubel finds the description of the ad as ‘pioneering’ amusing, reflecting the American pioneers were beset by arrows and disease as soon as they left their wagons). And in 2025, Coke and Silverside did a re-do. Again, it attracted criticism -- this time over inconsistent truck sizes -- but the brand went into it knowing what they were in for, and calculating that any negativity was worth the bigger goal.
Having said that, for every AI ad that received a pile-on, there were dozens that went largely unremarked. Whether that means that the AI filmmaking wasn’t that bad or just that the ads didn’t trigger some algorithmic cascade, who’s to say? Also, many campaigns are using AI to help with the odd shot or with localisation (where it can speed up processes, but where human understanding of local cultural and linguistic nuance is still key).
A Rat King of Controversies
In creative advertising circles, the main recurrent debate was, inevitably, around the quality -- or lack thereof -- in gen AI advertising. And then there’s the end-of-civilisation-as-we-know-it existential fears, which somewhat overshadow quibbles about what it all means for advertising agencies or production companies. It’s all too big to dig into here.
But on a more… cognitively manageable level, there have been debates, scandals, pockets of resistance and emerging problems that even the most breathless hype merchant couldn’t ignore in 2025.
First of all, there’s the brain rot and work slop of it all. In June, a study from MIT suggested that use of platforms like ChatGPT decreased critical and creative brain activity. It’s a much-cited paper and presents AI enthusiasts with something of a paradox. Those who embrace AI say that it requires critical human oversight -- but if it’s turning our brains to mush, then who’s going to be capable of that oversight? Later on in the year, came the Harvard Business Review-published research that revealed the phenomenon of ‘workslop’, reams of AI generated reports, memos, meeting transcripts that employees are generated to give the impression of business but that go unread and performative rather than productive.
2025 also saw the growth of AI influencers, models and actors, which also triggered debate. The UK’s Channel 4 drilled down into this with a documentary that, in a meta twist, featured an AI presenter. The most polarising figure was AI starlet Tilly Norwood, presented as the new face of Hollywood, triggering gigabytes of online outrage. Beneath the scandalised headlines, though, the truth of Tilly was much less impressive, fuelling the mismatch between expectations and reality.
Ethics, AI and advertising is a writhing rat king of issues. Environmentally, the vast volumes of water and energyconsumed by the data centres that power gen AIhave put the technology on the agenda for Ad Net Zero. There’s the feedback loop of stereotypes and unrealistic beauty standards, the erosion of a sense of shared truth as models like the new Google Nano Banana Pro are said to “erase what's left of the line between reality and AI”.
Legal Smeagols
One of the biggest constraints on brands fully embracing AI, particularly in terms of tools trained on the wild, wide world, are the ethical and legal grey areas. Harvesting and regurgitating copyrighted creative material tends to get the legal department on edge.
Holding companies like WPP and Publicis Groupe have addressed this by creating walled gardens for brands, trained only on approved materials. And insiders have admitted that these sorts of restrictions in input have a knock on effect on output. But for other brands, there’s been a real divergence. For some, particularly mid-sized national brands without huge legal departments, there’s been a devil-may-care approach, while at those with a culture of uncertainty, AI’s been avoided completely.
But recently there’s been a shift -- in the legal side if not the ethical side. Getty Images lost a recent lawsuit against Stability AI in the UK, which may indicate a direction of travel in favour of the AI companies. And just last week, Disney agreed to license its own precious characters to Open AI’s Sora 2 (with restrictions) and have invested a billion dollars for the privilege. There’s a lot of very modern things to be said about ceding control and co-creating with fans, but at the same time, for a company with a history of ferociously defending its IP, it’s a sharp change of direction and huge risk for the family-friendly Disney brand. Since the announcement of the deal, Sora users have been using Disney IP to fuel problematic shorts about everything from 9/11 to Jeffrey Epstein.
National governments vary on their stances, with Australia proposing to clamp down on big tech’ scraping creatives' IP, while the UK government's approach is more ‘don’t ask, don’t tell’.
Bubbling or Booming?
With an opaque pattern of investment between AI companies and outlandish valuations with little resembling an actual profit in sight, some observers have been talking about the possibility of an AI bubble. But just as the dotcom bust didn’t stop digital advertising becoming a dominant part of the industry, an AI collapse might devastate markets and livelihoods but it’s unlikely to prevent the technology’s continued infiltration of advertising.
Looking to 2026, it seems unlikely that AI controversies will stem the flow of gen AI advertising as the holding companies have invested so much in it. If the past couple of years is anything to go by, the quality of the visuals and audio output will continue to improve, and we may see more intuitive means of input beyond the laborious prompts as AI businesses explore ‘multi-modal’ AI.
Wherever the AI journey takes the industry, though, we’ll be keeping up with the lovers, haters and something-in-betweeners, platforming debates and discussions, and doing our best to keep the LBB community up to date.
