Artificial intelligence assistants can now survey a field rapidly and with increasing accuracy, but impactful Review and Perspective articles have never been just surveys. Their value lies in perspective provided by the author(s).
Virtual assistants built on transformer-based large language models (LLMs) have entered the cultural and professional zeitgeist at an astounding pace. Over the first half of 2025, the global number of active ChatGPT users doubled to around 800 million1, and by September of that year, more than 20% of all US workers reported using artificial intelligence (AI) tools to perform at least some of their work2. Billboards in major cities across the world, advertising popular services such as Claude (Anthropic) and ChatGPT (OpenAI), now feature cinematically shot scenes of people living their lives, with barely a hint of the product itself3, signaling a shift from publicizing the general concept of an LLM to selling the choice of which AI assistant ecosystem to use.
The alternative text for this image may have been generated using AI.
Credit: Stefano Guidi / Stringer / Getty Images Entertainment / getty
Basic science and engineering are no exception to the broadening reach of AI assistants and other emerging methods across a rapidly evolving AI landscape. Consumer‑facing tools built on these advances, covering everything from literature reviews to figure generation, now seem to appear almost daily.
Agentic assistants can now identify consensus findings, synthesize common methods and point to key citations with a level of accuracy that brings them over the threshold from liability to asset. These skills overlap with what many Review and Perspective articles do — or at least what they seem to do on the surface. And so we arrive at a natural question: if these tools can perform the same nominal function as a Review article, what purpose does a Review now serve?
Well, transformative Review articles have never been just a survey of their field4. The utility of a Review article lies not in collation but in reflection, contextualizing why a field matters and pointing ahead to where it is going. In some fields, the need for the perspective of the author(s) is self-evident because defining the ‘big questions’ is an intrinsically human task. Think of areas where the answer to ‘why do we care?’ is often simply ‘because we want to know’.
But in applied science and engineering, it can be easy to convince oneself that the importance of the big questions is fundamentally objective. Why do we care about carbon-capture technologies? Because we want to mitigate climate change, not necessarily (or, more accurately, not only) because we care about the science of carbon capture. But if we take the question of ‘why’ one step further, the human element reappears.
Sustainable chemical-process design is a common theme in the journal’s Reviews and Perspectives articles: see, for example, Thomas Epps, III and colleagues’ Review on analyses of circular solutions for plastic waste recycling5, and Ryan Lively and colleagues’ Perspective on water management and heat integration in direct-air-capture systems6. Why give priority to environmental stewardship? To preserve a livable world. This is a value judgement as much as it is a technical one. Hypothetically, a society could decide that current consumption outweighs future risk, that climate impacts on its neighbors are not its concern, or that leaving a healthy ecosystem for future generations is unimportant.
Similarly, Reviews in the journal have covered technologies that can be efficiently deployed in low- and middle-income countries: see, for example, Shihong Lin and colleagues’ Review on separation technologies for direct lithium extraction7 and Erwin Reisner and colleagues’ Perspective on floating solar technologies for chemical synthesis on open water8. Why give priority to equitable access to technology? Because we, as a global society, have decided that equality in standard of living is a problem worth addressing.
And so the question is not whether AI assistants can or will be able to write an accurate and comprehensive summary of a field. It has become increasingly clear that these frameworks are well on their way to being able to do so, and that substantive strides are being made to shed their distinctive ‘LLM-like’ voice in favor of an adaptive style that can reflect that of individual writers.
But summary does not a transformative Review make. Importance, urgency and direction arise from human context and priorities — elements shaped by experts in the field. And it is precisely these experts who write Reviews; their perspectives cannot be replaced because it is exactly their opinions that give a Review the ability to steer the field rather than merely describe it.
The situation is a bit like the diverging fates of departmental libraries and journal clubs from the pre‑Internet era. Before every paper was available at the click of a button, departments and research institutes maintained substantial libraries to surface the latest findings in their fields. Those same spaces often hosted regular journal clubs to discuss and debate those findings.
Today, many institutional libraries have been cleared out and repurposed; however, journal clubs have not only endured but expanded into digital formats. Then, as now, the value of perspective has persisted, even as the work of information gathering has been transformed.
From day one, the perspective of the author(s) has been the central factor in our editorial team’s evaluation of potential Review articles. Writing and review assistants have already become important tools in a Review author’s toolkit, and as they become more capable, this emphasis on field‑shaping guidance will become even more prominent.
In an era when summary and synthesis have become commodities, it is the expert’s voice that gives a Review its value.