The release of DeepSeek’s R1 model created quite a stir in the tech world, causing the largest stock fall in US market history for microchip maker Nvidia, which lost nearly $600 billion (£479 billion) in a day.

The AI model was founded by Chinese businessman Liang Wenfeng and was reportedly cheaper to develop and run than its competitors such as ChatGPT and Gemini AI. Estimates stated that the Chinese AI model cost around $6 million (£4.8 million) and 2 months to develop.  Though independent research firm, SemiAnalysis, puts the total cost of developing DeepSeek at a much steeper $1.3 billion (£1.04 billion).

The app is open-source and therefore free which has caused many in the marketing industry (and those outside it) to rush to download the model.

But is it truly a better option?

Senior vice president of analytics at creative agency Rapp, James Heimers explains: “DeepSeek vs ChatGPT isn’t quite comparing apples with apples. DeepSeek is better understood by comparing it to both proprietary models (e.g. Gemini and Claude) and other open-source models (e.g. LLaMA and BLOOM).

“Where DeepSeek differs from most large language models is that it leverages a Mixture-of-Experts design, which optimises efficiency by activating only the specialised parts of the model needed for a specific task.”

B2B agency Park and Battery’s president and chief creative officer Michael Ruby says: “We’ve put DeepSeek through its paces and the big difference is that it’s fast. Crazy fast. Faster than ChatGPT and other AI platforms we’ve tested, and seemingly without sacrificing accuracy.”

Ruby added: “But, to me, the speed isn’t enough to meaningfully separate DeepSeek from the pack. It’s largely the same in terms of what it can generate, and it requires actual coding to build AI agents – which makes democratising and scaling AI a real challenge without serious [developer] resources.

“We can spin-up really effective agents in literally minutes with ChatGPT! Right now, DeepSeek feels better suited to developers, not marketers.”

However, Esme Robinson, director of Platform Solutions at Epsilon, believes DeepSeek may still be useful to marketers.

She explains: “Right now, it’s all about testing. Marketers will experiment with DeepSeek for copy creation and soon, image generation.”

“The real shift will come with ‘agentive AI’, where generative and predictive AI combine to create entire ads in real-time, from copy and visuals to calls-to-action and placements, reacting to live triggers like breaking news, sports results, or weather changes.”

She adds: “This level of automation could be a game-changer, but human oversight is critical. Without it, AI risks producing content that feels off-brand or irrelevant. DeepSeek’s true potential lies in how well it fits into this evolving AI ecosystem, helping marketers balance efficiency with creative control.”

Dave Porter, global director of technology integration and innovation at agency Oliver agrees.

“Marketers can leverage DeepSeek for multiple functions, including automated content creation, SEO optimisation, sentiment analysis from customer reviews, personalised ad campaign creation, and 24/7 chatbot customer support, he said.

“Its strength lies in technical analysis and automation of repetitive marketing tasks, making it valuable for data-driven marketing strategies.”

Furthermore, he adds: “Most users will use open-source versions that have been hosted in “safer” locations. In a response to DeepSeek, Open AI released GPT3o mini, with a powerful reasoning model and an agent that uses reasoning to synthesise large amounts of online information and complete multi-step research tasks for you, called Deep research.”

Subscribe to Marketing Beat for free

Sign up here to get the latest agency-related news sent straight to your inbox each morning

Data Risks

Despite being cheaper, DeepSeek’s Chinese origins are causing many in the West to cite security concerns.

Ruby says: “One significant concern with DeepSeek is data privacy. DeepSeek stores data on servers inside China, which means it’s subject to China’s strict data laws, laws that could allow government access.

“Even more concerning, it’s been reported that DeepSeek is linked to China Mobile, a company the US has flagged for its ties to the Chinese military. Marketers and agencies – especially those handling sensitive data, should be intimately aware of privacy and usage terms before using any AI platform, but especially DeepSeek.”

And it’s not just marketers that are concerned over data privacy. The Italian Data Protection Authority has already issued a nationwide ban on the AI bot, due to security concerns with the model’s parent companies, Hangzhou DeepSeek Artificial Intelligence and Beijing DeepSeek Artificial Intelligence-.

It has also reportedly been banned from all government devices in Australia. And they aren’t alone, many other countries including the US are considering banning the AI model altogether due to its Chinese links.

Future

But this doesn’t mean the app should be counted out altogether.

Ruby explains: “It’s far too early to tell if DeepSeek will be a go-to generative AI platform for marketing. But GenAI in marketing is inevitable. We’re already aggressively leveraging ChatGPT and other platforms at Park & Battery, to streamline workflows, cut down on work and free us up to focus on what actually moves the needle – big ideas, bold strategies and breakout creativity.

“If DeepSeek continues to prove itself out and can address privacy concerns, it could certainly make its way into our AI tech stack.”

Porter adds: “Ultimately, the future belongs to companies that not only use AI, but also have control over it in their marketing.

“So, the battle of the models continues…”