Forecasts for the medical writer's job
Between 2019 and 2026, forecasts of how AI would change the medical writer's job differed on whether the role would survive, which part of the work would go first, what would become of the junior writer whose training was the routine work, how the roles would be redrawn and which skills the new work would need.
Whether the role survives
Between 2019 and 2024 several forecasts held that AI would not take the medical writer's job, but that a person using AI would. In December 2019 one writer judged it very unlikely that AI would completely overtake the job, but likely that writers who embrace AI would overtake those who do not. In September 2023 two authors whose company sells an AI tool quoted the adage as a description of where things could be going in the short term. In the same month a member of EMWA's new AI working group expected writers who embrace AI to have a competitive advantage in the job market, because they would produce high-quality work in less time than their peers. In March 2024 one speaker called the adage a cliché that he held to be entirely true, and in May 2024 the commercial head of a medical writing automation vendor quoted it as a colleague's.
In September 2019 one overview of regulatory writing judged that AI-enabled automation had not yet made major inroads into the field, though technologies that automate at least part of regulatory document preparation were already available. The same overview saw sweeping regulatory change as a rich opportunity for growth in medical writing, with one market report expecting the global medical writing market to grow to US$3.6 billion by 2026.
In 2022 one writer held that no AI would ever completely replicate the skill and art of a human medical writer, and that what would not change is the writer's ability to analyse scientific data and craft language around the results, together with leadership, communication and collaboration with project teams. Another expected that it may be many decades before AI can fully replace a human medical writer, while repetitive tasks may soon be done by automation or AI, letting the writer focus on the science.
In September 2023 one writer expected language models to reshape the medical writing space and held that navigating the shift needs an open attitude, curiosity and continuous learning. She called the tools allies rather than substitutes, with the human abilities of understanding context, strategic thinking, critical evaluation and conveying nuanced emotion remaining invaluable.
In May 2024 one panellist had heard a great deal of reassuring talk that AI is a tool and that the human must be in the loop, and agreed with it for now. But what everybody was analysing, he noted, was how one model generation behaves while a later version already existed and nobody knew what would come after it, and even the human reviewer of what a model generates could become an AI. On the analogy of cars, which did not drive themselves and now do, he asked why an AI system should not be able to write an entire document from the structured data of a whole clinical research programme. There would be shifts that are hard to predict, and what AI can do now should not blind writers into saying that they will not be replaced in future, though he did not know when that would happen, and probably not in the next ten years.
A speaker from the European Medicines Agency, asked in May 2024 whether regulatory submissions might come to be reviewed by AI and regulators be out of a job, gave a personal view rather than the agency's. Regulators, he held, are in a different category, and change will happen across the industry; he did not know, but did know that they would be using AI more, and found it hard to say because only a year earlier people had expected things to move faster than they had. Another panellist expected more and more augmented writing, augmented regulatory work, augmented everything: an assisting tool rather than a replaceable one.
Another speaker that month acknowledged the fear-mongering that AI will take everyone's job and set against it the tools the profession has already absorbed: clinical study reports and common technical document submissions going to the FDA in trucks, then the electronic submission, then automation, and now AI as one more tool that helps writers be more efficient and produce better quality.
In May 2024 a librarian who had been a medical writer held that AI is not a passing trend, that AI skills will be a prerequisite for medical writers in every discipline, and that people are not going anywhere, because someone has to interact with the tools. The commercial head of a medical writing automation vendor reported the fear fading: whether AI would take writers' jobs had been the question at the vendor's conference booth all day in the previous couple of years, and that year he had heard it once. His framing was human plus machine, with the human still needed. One who ran project delivery at a medical writing company that resources its projects with contract writers held that companies would not be removing their medical writing teams from the bottom line any time soon, and that AI would augment and benefit medical writers and not replace them in any way. He expected AI to become accessible to everyone at low cost, like the internet, with the gain going to those who start using it early and build it into their processes.
A 2023 paper by OpenAI and the University of Pennsylvania, whose title calls GPTs general-purpose technologies, consolidated the skills the US occupational database lists for each job into basic skills, such as science, writing, critical thinking and programming, and analysed what applying GPT technology to the day-to-day job would do to each. Of four of those skills that one speaker showed in May 2024, science and critical thinking were the bottom two, with science going negative. His conclusion was that a job changes skill by skill rather than disappearing, since every job is a mix of fifteen or twenty skills, some needing less work and some more.
In August 2026 an experienced regulatory writer at a medical writing company judged AI not yet ready for prime time and overhyped, saw it beginning to change the field, and expected change over the next year or so and a great deal of it over the next decade, in a direction nobody could predict.
The routine work and the writer's judgement
In December 2019 one writer expected that improving AI tools for regulatory writing would inevitably redefine the role, with writers no longer responsible for extracting information from different sources and integrating it into a readable document. Her image of the future writer was a specialised craftsman using the tools in a complementary way to enhance the writing of complex documents, with AI in the long run lessening repetitive tasks such as extracting information from sources and preparing drafts, and giving the writer more opportunity to apply their know-how to complex scientific work. In an optimistic and, she suggested, perhaps realistic view, the interaction between writers and machines would be a friendly one, with major advantages for the craft.
In September 2023 one journal editor set the robots' old three Ds, the dirty, the dangerous and the dull, beside the dear and the difficult more recently added, and asked where that leaves writers and whether they are ready to delegate their deliverable documents to a digital do-it-all. Her answer was that clinical research requires skill sets that AI can never fully provide, and that in a RACI matrix the responsibility and the accountability remain the writer's, because AI, like human intelligence, has limits, including hallucination, ethical questions and a lack of context and creativity, and she could not see an artificial system fully understanding the principles of Good Clinical Practice soon. Writers, on her account, can delegate the driving but determine the direction and the destination. She held that in medical writing the old trade-off of good, fast and cheap should never be a pick of any two, since good has always been the standard and there is no trade-off on quality, but that AI lets writers pair quality with speed, developing good regulatory documents quicker and getting treatments to patients faster. Computer programmes now did the in-text tables and patient narratives she once built by hand, and she held quality control, from data checks to readability metrics, to be crucial and the detection of errors to be a very useful AI capability.
In the same month the guest editors of a journal issue on automation in medical writing framed AI and automation as opportunities rather than threats. On one side they set the grunt work the tools can reduce, such as repetitive text, formatting, reference checking and consistency, and on the other the challenges the tools bring: the quality, accuracy and reliability of machine text, the human touch and creativity of the writing, and ethical and legal issues of health information privacy and intellectual property. Writers, on their account, need new skills and collaboration with programmers, data managers, statisticians, reviewers, regulators and patients, so that the tools serve the best interests of all parties, and those who embrace the technologies with curiosity, creativity and critical thinking will be able to harness their potential.
In September 2023 two authors writing on narratives for clinical study reports, answering a news article's estimate that generative AI could affect some 300 million jobs, expected the role to change rather than go. Writers would need a new skill set to understand and use AI, and would be relieved of manually inserting data into each narrative and of full quality-control checks, so they could focus on the story of events for each participant and on their scientific expertise. With each narrative project less labour-intensive, writers could work on more projects at once, which calls for skill in prioritising and managing them. The medical review of narratives would stay human, and any AI output would remain, at least for the foreseeable future, an initial draft. The authors speculated about AI building narrative templates from the structure of the clinical database and, once narratives are final, redacting or anonymising them.
In September 2023 two authors whose company sells an AI tool put critical thinking beyond AI's reach until artificial general intelligence arrives. They expected most writers to be delighted to pass routine tasks to a computer, which could free them to contextualise and interpret clinical data and to have data discussions with clinical teams much earlier, and named the limiting factor as how the tools are implemented and how ethically checks and balances are introduced.
In May 2024 one writer whose medical writing company had built its own AI drafting tool held that the hype about such tools is real: they will dramatically change how medical writers work and let them focus on critical thinking and the higher-level tasks humans should bring. Her analogy was autopilot: nobody would board a plane with no human pilot, yet autopilot has been in use for decades, with the human there for the critical parts, the takeoff and the landing, while the computer takes the dull middle of keeping the aircraft level and away from obstacles, so that the pilot can rest and be at full attention when needed. An AI tool, she held, does the same for medical writers, raising the writer's role to a more strategic and critical-thinking level, and in her company's experience that works. Her company's tool was not meant to replace the medical writer, since it would be daft for a company of medical writers to build something that did them out of their own jobs, but to remove the mundane routine tasks and free writers for creative, strategic and critical-thinking work. It produces a draft that the writer and the clinical authoring team develop into the drafts that are discussed and the final document, so every downstream task is still performed. AI cannot, and in her opinion should never, interpret new data, and it cannot contextualise, think critically or emote: its job is to pull information together, and the human's is to look at the signals and decide what they mean.
In May 2024 a regulatory medical writer of more than twenty-five years, who had spent the last eight on a team of writers developing a rules-based AI tool, answered common misconceptions about how AI is used in medical writing. One is that AI will strip out the strategic part of the work and leave writers as pure editors. The role grew out of editing, certainly in Europe, where many writers supported non-native English speakers, and became one of handling content, with writers working as subject-matter experts in their own right to find the story behind the data. Half of writing an initial draft is still collating information, the copying and pasting, and the strategic part begins after it: looking at the patterns in the data, recognising the story, and mediating the debates between functional areas that tease out the storyline. A tool, generative or rules-based, removes the collating and does it better, faster and more accurately than a tired human who makes copy-and-paste errors, but it cannot think. So the writer becomes a guide and adviser to the people helping to write the document, and an architect of the output, who has to understand its purpose and its data well enough to tell the tool what to do, since without that knowledge what the tool produces is not useful.
Another misconception concerns bias: whether writers will be biased by what the system selects from the data. Bias cannot be eliminated, but writers will be no more and no less biased than before, because the tool selects what it is directed to. Told to find every case in the adverse event tables where the treatment groups differ by 3%, a rules-based tool finds every one, where a human sent to do the same may well miss some, and clinicians have been seen to wave away a 5% difference as not that important. The tool does not interpret, since it makes no judgement calls: it is like being handed the pages with the highlighting already done, after which the writer, as the story builder, decides with the clinical, regulatory and pharmacovigilance colleagues what the highlighted findings mean.
The same writer's image for the tool was the electric mixer. It saves the chopping, and nobody imagines a kitchen without one or thinks that the chef has lost control of the dish or of the creative decision whether to make a soup or an omelette: the tool puts the chopped vegetables on the counter and the cook decides what to do with them. The forecast was that writing documents without AI will become unthinkable to later writers, and that documents will be written faster and more accurately, giving writers more time to think about the story. Only the preparatory phase is removed, and every round of review with the team remains, so the iterative discussion is untouched, which the writer expected to mean better stories about the data for the assessors and medicines reaching patients sooner.
In 2026 a writer of forty years, founder adviser at a company whose platform builds regulated clinical documents with symbolic AI and deterministic automation, estimated that about ten of those years had gone on going through outputs looking for signals, and predicted that this will be a thing of the past. If he could choose any time to start as a medical writer, he would choose now, because writers will have these tools. At a company where he had worked, a brand-new writer with no experience used to spend a couple of months taking text from the protocol and pasting it into a clinical study report template, which he called totally brainless and no training in critical thinking or in anything else, and which an AI can do at once. The AI might point to a set of laboratory events and ask what the writer makes of them, and the writer is the one who says that this looks like an effect on the liver and takes it to the team, so he expected critical thinking to be accented more in the writer's future. His analogy was looking up the capital of Azerbaijan: getting it instantly rather than through a library's file cards does not make him less of a critical thinker, but gets him the information he needs in order to think critically. In his view AI does really well at going through a 180-page laboratory safety table and saying what the four clinically relevant findings are, so that the writer can sit with the team and ask what they say about the drug and how it should be prescribed. It does not do well at taking a pile of tables and producing a final clinical study report fit to send to the FDA, so both are needed. The writer still needs to know what the tool is doing, how and why, what its rules are and whether they are appropriate in the case at hand, and so has to be involved in the tool's design and functioning, not only programmers. Another speaker in the same session, while not knowing the profession's future, expected the role to change quite a lot and perhaps to move towards examining what a tool has generated and looking for what needs updating.
In June 2026 one writer found it becoming clearer that AI can do the basic manual work, such as formatting a few hundred tables, but has not shown much value in the high-value work medical writers bring: mediating conflicts within a team, determining clinical relevance, inspection readiness and strategic positioning for regulatory defence. She described a move from static text to a data-centric world in which the underlying principles have not changed and human critical thinking remains a primary hallmark of the profession. In the same month medical writers were described as shifting their skill set towards AI prompt engineering and critical review of AI output, which to do well still requires deep foundational knowledge and a wider strategic appreciation of where medicines development is heading.
The junior writer's training
In May 2024 a panel chair asked whether extensive dissemination of AI for writing tasks would lower the skill level of medical writers, and whether people would stop knowing how to read tables, noting that there had always been something to be said for doing the grunt work by hand. One panellist maintained that the skills remain essential, because what the medical writer brings is cognitive and analytical: signal detection, pattern detection, and the interpretive work of aggregating information within a subset of data and across it to give a representative picture of a drug's safety and efficacy profile, not the grammar or whether to use the serial comma. Whether or not AI does the grunt work, on his account, that tool set still has to be applied, and he did not think AI would affect that dimension at all.
The chair then challenged that, at the snapshot of 2024, anybody with ten years' experience had reached their strategic judgement by doing a great deal of grunt work first, and asked what happens to the invaluable learning of junior writers, whose menial tasks are what solidify their knowledge. The panellist conceded the point from his own history: boxes of paper case report forms delivered to his office, hand tabulation, Student's t and chi-square tests on a hand calculator, and the understanding of what every single data point means in characterising a population that came out of it, while allowing that it may have been too much pain for too little gain. Another panellist of the same generation held that not all skills are equal: by her account she learnt nothing useful from Friday nights sobbing over hundreds of pages of tables, figures and listings beyond her own alcohol tolerance, and she did not think junior writers would miss it. She also held that some elements of the craft have to be learnt and that automating too much of them away will be an issue, but that this is not a reason not to automate; it is a reason to change how medical writers are trained.
In 2026 webinar participants asked how junior writers are still to develop scientific expertise and learn the documents, and noted that the change also affects how many junior writer positions are available. One experienced writer answered that AI will make it faster and easier for junior writers to learn critical thinking, because they will be doing more of it rather than searching through documents trying not to miss the one signal on page 48. Responsibility cannot be completely handed over to the AI, on his account; what it can do is gather the information together and lay it out so that thinking critically about it is easier.
How the roles are redrawn
AI is often associated with efficiency gains, cost reductions and job displacement, but in 2025 two authors placed its primary value in pharmaceuticals in augmenting capability, industrialising complex domain knowledge and accelerating delivery to patients. In 2024 other authors answered the fear of job displacement by treating generative AI as augmentation that frees writers for strategic analysis, creative writing and interpretation of complex data. Leaders of medical writing departments at AMWA's 2023 Executives Forum held that advances in technology will not make the writer's role redundant, and gave the autopilot analogy: the technology is standard in aviation and has not removed the need for a human pilot, and in the same way advanced technology could help with mundane tasks such as pulling information together for a first draft. Current tools cannot understand nuances across documents, on the forum's account, so a human in the loop, a content strategist for instance, has to keep monitoring and controlling the data sources and tools. The forum held that the role is evolving and that each process will need to be refined by individual working teams.
The 2025 authors traced the regulatory writer's role from circulating paper drafts through leading collaborative authoring and enforcing eCTD standards to its next form, configuration lead and editor. They saw demand growing for people expert in both life sciences and AI or data science. New roles include creating and implementing AI tools, and working as product owners, product managers, prompt engineers and business analysts on implementation teams; validating an adverse-event analysis prompt, for instance, needs medical and scientific knowledge. For users the role moves from crafting every line to configuring settings, reviewing output and aligning the tool with team and project needs. Automation, on their account, raises the expertise needed to evaluate and control generated content, a seemingly paradoxical shift that elevates the seniority of writing roles, with junior writers using AI for repetitive tasks. Pharmaceutical companies are pairing domain experts with technologists in blended "squads", and writers fluent in the tools' strengths and limits are well placed to become AI-enabled subject-matter experts. The 2024 authors foresaw a writer who keeps quality control, insight and a cohesive document set, hands over the tedium of combing through data sets, and spends the time on strategic storyboarding, sourcing input from subject matter experts and sharing key insights.
The forum's leaders expected every role, writers, managers, leaders, formatting specialists and quality functions, to adapt within one to five years. New technical roles would set up tools and integrate them with clinical and safety databases and trial management systems, and content strategist and content architect roles could manage adaptable content and would monitor system performance and content quality, with clear, consistent rules for content creation needed so that content is reproducible across documents. Respondents to AMWA's 2023 survey of people who lead, train or develop medical writing teams likewise saw a need for trainers and implementation specialists expert in the technology, called quality control of all AI-generated content critical, and thought interested writers could be trained into the new roles. The 2025 authors and the forum both expected super-users: writer AI leads who build AI into daily work and train others, in the authors' account, and senior writers who may serve as superusers and digital champions training junior writers and cross-functional users, in the forum's.
In 2022 one writer expected the writer to be recognised as an expert in their own right, guiding a team through the limits and advantages of automated document generation. Another expected the writer to focus on message and content while the tool handles repetitive and routine tasks, and writers to co-develop the AI tools. A third thought future writers' skill sets likely to include elements of data science and content management as clinical research moves from a document-driven to a content-driven industry, with writers working on interlinked content developed in parallel rather than on one document in isolation. Two others named as one huge change a move by regulators from the document-centric Common Technical Document to a real-time flow of study data, and speculated that advanced computation in the coming decades might let regulators interpret trial data automatically, with applications built on a rolling basis; documentation would then be reduced to introductory and contextual statements, leaving interpretation to the agencies, and writers would draw on technical and data science skills.
In 2024 the team behind CORE Reference, a manual for writing clinical study reports, held that as AI tools are increasingly used to generate clinical study report data and text, tool developers rely on knowledgeable medical writers who understand the content requirements of disclosure-ready reports, and the writers must use that knowledge to prompt the tools to output the correct content. The writer's role, on the team's account, will evolve from de novo content creator to preparing expert-led prompts and critically reviewing the AI-generated text, and writers who own these parts of the process will reduce the potential for AI hallucination and ensure continued trust in regulatory documentation. The team described the role as undoubtedly evolving from pure de novo content creator to encompass curious and robust content interrogator, both of which need a sharp regulatory eye.
The 2025 authors concluded that writers become collaborators in software development, stewards of quality content and architects of intelligent content ecosystems, moving from hands on the keyboard to strategic content designers, and that those who adapt to AI expand their influence across the regulatory lifecycle. The survey's authors concluded that writers should be part of the discussion on what technology to use, how and when, and included in developing, deploying and implementing AI, and that they can raise their value by facilitating adoption, taking on the new roles and improving the quality and pace of AI-assisted output. Asked by the survey's authors what AI's role in medical writing is, ChatGPT ended its answer with the caution that human expertise and oversight remain crucial. The forum concluded that writers, often underused on copying from sources and formatting, can move towards document strategy and message formulation through the careful adoption of appropriate tools, that integrating the tools will need leaders to take a positive approach, and that senior writers will need to hone their critical thinking and appraisal skills and be expert users and even contributors to the tools.
The skills the new work needs
The 2025 authors named data interpretation, prompt engineering, structuring document creation, modular thinking and cross-functional collaboration as the valued skills, while manual formatting, repetitive drafting and versioning matter less as generative AI and automation take them over. The survey's respondents wanted training in the tools and especially in writing prompts that produce relevant responses, and an understanding of how the tools work so that accuracy and security can be judged. In 2022 two writers held that technical proficiency will be increasingly important as the industry adopts automation and content reuse, and that diversity and inclusion, missing from many document templates, should become part of writers' training. In March 2024 one speaker held that prompt engineering is one of the skills all medical writers are expected to learn: the models may change, but writers will go on creating the prompts, and human intelligence will still be needed to make sense of what the tools produce.
The regulatory writer who had spent eight years developing a rules-based tool also answered the misconception that writers will have to become tech-savvy, learning some programming, understanding how the systems work and ceding ground to younger, more technical colleagues. The answer was no, on the analogy of a car, which most people cannot fix and can still drive. The writer learns to use the tool as with any other software, PerfectIt for instance; in a rules-based system that means answering the questions that steer it rather than writing prompts, and once it is learned it runs the way driving does, without thinking about the clutch.
The forum named leadership and communication; return on investment and risk management for leaders; change management and continuous learning; digital literacy, including writing prompts with clear instructions; and critical thinking and content logic, to plan standardised content reuse from the protocol through the statistical analysis plan and clinical study report to the summary documents, in collaboration with subject matter experts on standard content libraries. Writers may need regular refresher training and quick reference cards as tools change, and technologically savvy writers may carve out roles developing systems and supporting change. The forum saw AMWA as able to help by defining skills and roles, a lexicon and a toolkit; by working with quality assurance professionals on audits of technology-generated content; and by training in programming and engineering, business cases, measuring value and persuading stakeholders.
In March 2026 one of the CORE Reference team saw AI not as a replacement for human intelligence but as a complement to writers' work: at that stage AI tools did not represent a stand-alone and self-sufficient path to a compliant clinical study report, and the technology still depended heavily on the expertise of medical writers, particularly their understanding of the evolving regulatory landscape and of the increasing emphasis on data transparency and reporting standards. In 2024 the team held that, as AI-generated clinical study report texts arrive, it is more important than ever for regulatory writers to fully understand the content requirements of the report and to keep up to date with evolving requirements.