AI in 2025: Five trends for marketing, media, enterprise and e-commerce

While there are plenty of topics to follow in the coming year, here are five things industry experts think will happen with AI in 2025 — not to mention all the news from Las Vegas this week during CES. (Read more in our 2024 AI news timeline and how platforms are evolving with AI content and ads.)

Generative AI will continue re-shaping search

It’s no surprise that many expect generative AI to continue disrupting and reshaping the future of search. What’s less certain is how fast, to what extent, and how it’ll impact competitors, advertisers and users. It’ll also be worth watching what happens with the ongoing search and ad-tech trials — especially if Google’s forced to sell off its Chrome browser.

Agency executives and search experts expect search to rely less on keywords and more on multimodal capabilities for semantic text, image and video search. Others expect to see more personalization with chatbots and autonomous agents that browse the internet. Changes also could lead to brands relying less on clicks and more on impression-based metrics or micro-conversions — especially with AI search platforms featuring unordered lists, detailed call-outs or deeper chat-based conversations. 

Recent research from BrightEdge found ChatGPT had a 44% month-over-month surge in traffic referrals, but OpenAI still captures less than 1% of overall market share. Meanwhile, Google has been rapidly increasing the volume of its AI Overviews responses. 

Jim Yu, CEO of BrightEdge, said advertisers, publishers and creators will all need to rethink their approach to content by optimizing enough to show up in AI results while also finding ways to grow and maintain audiences.

“It’s going to be a delicate dance next year for most,” Yu said. “And it is going to happen industry by industry. Certain industries are a little more isolated and going to come a little later. But you can kind of see e-commerce, publishing and travel starting to get hit.”

The risks and rewards of AI content will attract more attention — and more resources

There are still plenty of unresolved risks and unanswered questions that come with AI adoption: copyright litigation, accuracy issues, privacy and bias concerns, and other ethical dilemmas that affect people and practitioners.

Companies that adopt AI tools for content and ads also need to make sure they maintain consumer trust. According to Gartner analyst Nicole Greene, that requires being transparent with AI-generated content, listening to audience feedback, and engaging with customers about both the benefits and concerns related to AI. One Gartner survey found 80% of consumers think generative AI makes it difficult to identify what’s real online while 78% said it’s important for AI-generated content to be properly labeled.

“Regulators expect that all organizations and their leaders will comply with responsible use, even as regulations shift,” Greene said. “For this reason, it is crucial to establish a strong AI governance foundation based on common AI principles now, rather than in the future. The technology is outpacing our ability to regulate it, so organizations must put safeguards in place.”

Maturing LLMs will help drive enterprise growth

After years of hype around general purpose large language models, enterprise adoption could depend on smaller, more specialized LLMs to help with industry-specific tasks. One example is the enterprise-focused AI startup Writer, which has developed industry-specific LLMs. While its Palmyra Med model has medicine-specific knowledge and tasks, the Palmyra Fin model is more focused on math, reasoning and calculations. The most recent rollout is Palmyra Creative, which debuted in December with tools for creative writing — and creative thinking.

Waseem AlShikh, Writer’s co-founder and CTO, thinks enterprise adoption will rise as industry-specific models help improve accuracy and performance based on what companies need for their particular kind of business. He pointed out that accuracy rates of 80% or 90% aren’t good enough for industries like finance and healthcare that require utmost precision. To him, productivity with AI means paying less money or the same amount of money for better results. However, he said some AI models are still costing companies more on an hourly basis than they’d pay for a full-time employee. 

“You need something cheaper, you need high accuracy. And the way we think we’re going to achieve it for those who adopt it is by having [AI models] that self-evolve by simply not making the same mistake again,” AlShikh said. “Because companies working on specific flaws don’t change that much if you adopt it without having those domain-specific models that can actually learn and have a deeper knowledge on those specific areas.”

More momentum for open-source adoption? 

Joel Burke, a policy analyst at Mozilla, thinks open-source AI models will pick up more momentum as they become easier, cheaper, more secure and more effective. That could be especially helpful for companies that want to protect their IP and user data since open-source models often run locally on-device without sending data to the cloud or third-party providers. 

Burke thinks open source will foster innovation, reduce barriers to entry and give users more choice when choosing an internet browser or LLM for various tasks. Mozilla is also working on projects like Llamafile, a Mozilla Innovation Project initiative to make full-stack LLM chatbots less complex and more transparent.

“When you’re using open source, it just makes it easier for you to build something,” Burke said. “I think you’ll see a lot of startups basically take something open source, build on top of it and use it as their foundation or backbone. Not every startup is going to have the money that [other major companies] have to train their own huge models.”

AI will change branding priorities — and the branding process

Shifts in consumer behaviors driven by AI adoption could also affect how companies think about branding, said David Placek, founder and CEO of the naming agency Lexicon. For example, increased adoption of voice AI search —  with ChatGPT, Apple’s Siri, and Google’s Gemini — could increase the importance of how brand names sound.

“If ChatGPT gives you five brands to look at, you’re probably not going to look at all five of those,” said Placek.

Lexicon also is incorporating AI into its processes, such as using an LLM to analyze potential brand names based on five years of sound symbolism research and developing tools to organize and search its project archive.

“Understanding what was said and the memorability of it amid all the clutter,” Placek said. “Did something say ‘Siri’ or “Sarah,” and things like that. Is [a name] too soft? We’re going to be more and more sensitive as marketers to both the sound and the noise a word emits so it is both understood and more memorable.”