AI Industry Updates: Summer 2025

• Critical Thinking Conundrum: A flurry of recent research papers aims to determine how the use of large language models (LLMs) shapes how people approach writing and research. Research from the Massachusetts Institute of Technology shows that using LLMs results in reduced cognitive activity in comparison with independent writing. Meanwhile, a study from the University of Pennsylvania suggests that leveraging generative AI results in a poor grasp of the research topic. There’s also evidence that people are altering their language to mirror that of AI chatbots, whether or not they use the tools themselves. As AI use increases, it’s likely this conversation will accelerate.
  
  
• Professional Support: Globally, 58% of employees say they’ve used AI to help with their work, with 70% relying on free tools versus enterprise ones, and 47% saying they’ve used AI in ways that may be considered inappropriate, such as uploading proprietary information to a free tool (Melbourne Business School, 2025).
  
  While this gap should inspire smart brands to design tools that meet the needs of both employees and employers, the reality is more complex. Hiring data for recent graduates, alongside statements from AI company executives, suggests that some organisations see current generative AI tools as a replacement for entry-level jobs.
  
  This mentality is reflected in the swathe of AI start-ups prioritising tools that streamline work for upper-level professionals. Take the influx of services promising to help email-burdened managers and executives quickly sort through missed emails and meetings following time off. Meanwhile, tech company Delphi allows public figures to create AI surrogates – who convey their expertise to paying clients – and Artisan has attracted attention for its ads claiming that its sales support AI agents can function as fully independent employees (both San Francisco-based).
  
  For more, see Life in AI: Work, Learning & Skills.

• From SEO to AEO: By 2028, about 36 million Americans are expected to rely on AI for web searches, compared with 15 million who did so in 2024 (Statista, 2025). Already in March 2025, 58% of Americans encountered an internet search page that contained an AI-generated summary (Pew, 2025). With Google introducing its new AI Mode search in June 2025 – which replaces traditional search results with a chatbot interface – the share of people using AI for finding information online is likely to grow quickly.
  
  Consequently, companies are swapping search engine optimisation (SEO – online writing intended to rank highly in search results) for answer engine optimisation (AEO – writing intended to surface in chatbot searches). American-Dutch start-up Essio gives clients a visibility score, which calculates the likelihood that a given web page would appear in a related chatbot query. Meanwhile, established social optimisation companies, like HubSpot and Semrush (both American), are integrating tools that assess a website’s chatbot friendliness and provide tips for improvement.
  
  
• Anti-Deepfake Legislation: As the privacy and security threats posed by deepfakes become more evident, there’s a push for legislation restricting their use. In June 2025, Denmark announced its intention to amend current copyright law to specify that individuals own the rights to their body, voice and facial features. Once approved, the law will give people the right to ask online platforms to remove images and videos that look or sound like them, if posted without approval. For more on deepfakes and their security implications, see Cybersecurity 2.0.

And for further insights on AI, read 10 Tech Trends to Watch 2025.