In 2025, AI Needs to Put Up or Shut Up

PCMag editors select and review products independently. If you buy through affiliate links, we may earn commissions, which help support our testing.

In late 2022, ChatGPT was the topic du jour. There was a sense of something revolutionary and that entire industries would be upended.

Fast-forward two years, and the term "AI" elicits a more mixed reaction. The novelty of asking ChatGPT a question or generating a wacky AI image is gone, replaced by a sense of exhaustion around the buzzword of all buzzwords. Last month, a study from Slack found that the percentage of American workers who are excited about AI helping them complete their tasks dropped from 45% to 36% over the previous three months.

This comes as companies like OpenAI, Google, and Anthropic see slower gains in technical progress. OpenAI co-founder and former Chief Scientist Ilya Sutskever tells Reuters we're in an "age of wonder and discovery once again." In other words, back to the drawing board.

No More Half-Baked Tech

Personally, I'll believe the 2022-era hype when I come across a generative AI tool that redefines how I use my computer, phone, or other devices. PCMag has published 725 articles on AI this year, many of which I've written, and tested the splashiest releases (Google NotebookLM was a highlight). Most, if not all, feel half-baked.

While I use ChatGPT, it doesn't meet all my needs. The risk of hallucinations, plus a desire to sift through information myself rather than read a computerized summary, keeps me coming back to Google. But when I go to Google, its AI Overviews often serve up inaccurate or not-quite-there information. (Google Gemini recently told a user to "please die.")

Apple Intelligence has also been a let-down. After using it for a month, I want to turn off the sometimes confusing notification summaries. The first Apple Intelligence update for Siri was underwhelming, and the fully revamped version is now reportedly delayed until 2026.

It's getting harder to see AI overhauling modern life, at least in the short term. The companies pushing the tech see it differently, of course.

"AI is already making the impossible feel possible,” Chris Young, Microsoft's EVP of business development, strategy, and ventures, says in the company's 2025 AI predictions, published today. “This is the start of a full-scale transformation of how this technology will change every part of our lives.”

Microsoft says its models are getting better, and hopes to hook the public on AI "agents" in its Copilot software. (The next big buzzword in 2025 is "agentic.") So far, Redmond's AI push has had limited success; it got an early start on adding ChatGPT to Bing in February 2023—well before Google added AI Overviews to search in May 2024—yet it failed to increase Bing's share of search traffic, ZDNET reports.

It has since incorporated Copilot more deeply into Windows, launched a subscription service to rival ChatGPT Plus, and partnered with PC makers on Copilot+ PCs. That has also seen a slow start, though things may pick up when Intel and AMD formally jump into the market.

"Microsoft is now the Copilot company," CEO Satya Nadella said last year.

OpenAI, which has secured billions from Microsoft, admits that its GPT-5 model, a supposed breakthrough that will bring it closer to achieving artificial general intelligence, or AGI, is not coming any time soon. CEO Sam Altman is trying to keep the dream alive in the meantime by publishing out-of-touch blog posts about "The Intelligence Age."

"We have some very good releases coming later this year! Nothing that we are going to call gpt-5, though," Altman said during an October . Instead, 2024 saw the launch of smaller, yet still significant, model releases like GPT-4o and o1. More recently, it announced a ChatGPT Pro tier for $200 per month with unlimited access to the most cutting-edge models.

The AI Slump of 2025?

It feels like generative AI is entering the "trough of disillusionment," per the Gartner hype cycle. If tech companies cannot deliver more meaningful tech in 2025, it could get stuck there.

This analyst firm's theory says emerging technologies follow a predictable pattern. They burst on the scene with a surge of excitement, and then interest declines when unreasonable promises fail to materialize quickly. The tech must then prove itself before adoption rises again, ultimately plateauing at a more stable level.

Mapping this to generative AI, you could argue interest peaked in the months following ChatGPT's Nov. 30, 2022 release. In 2024, we've seen some cracks in the foundation; hallucinations continue, harmful and bizarre deepfakes are everywhere, and AI-based consumer hardware has struggled to get off the ground.

Hallucinations are particularly challenging since they're hard to avoid in even the most advanced AI systems. Even execs like Google CEO Sundar Pichai struggle to explain it; he says large language models just aren't that great at "factuality."

In 2025, generative AI will either start digging itself out of the trough of disillusionment or fall deeper into it alongside other over-hyped tech like cryptocurrency and the metaverse.

Gut Check: Do We Like Where This is Going?

AI's path to reaching the more stable "plateau of productivity" is checkered with lawsuits, which could be settled in 2025.

A mother suing Character.AI for encouraging her son to take his own life could influence the rules around personality-based chatbots. The New York Times is also suing OpenAI and Microsoft for copyright infringement, the result of which could have big implications for the future of media. Social media platform X is also suing the state of California over a law preventing deepfakes around election time, arguing it will make "parody illegal" on social media.

It's unclear how President-elect Donald Trump will shape the future of AI. Trump plans to nominate a so-called "AI czar" to work with Elon Musk on government efficiency, among other tasks. Trump's team reportedly wants to revise federal regulations to get more self-driving cars on the road, potentially benefitting Musk's robotaxi plans. The federal AI working groups that emerged from President Biden's 2023 Executive Order may not continue under Trump.

If the Gartner hype cycle holds any weight, buckle up for a rough year that could shake up the so-called generative AI revolution. Wealthy AI execs may not like it, but it could be welcome news for skeptics who are ringing the alarm on the dangers and limitations of AI.