Fluency Is Free. Judgment Is Not.

October 2026 · Lee Seaman, Seaman Medical, Inc.

For decades, Japanese pharmaceutical professionals writing in English faced a two-part challenge:

• Fluency: Writing correct, natural English, which until recently was hard and time-consuming
• Judgment: Deciding which information is important, explaining how it relates to other information, and making sure that explanation is easy to understand

The first difficulty has largely disappeared. Any employee can now produce grammatically perfect English in seconds. Large language models (LLMs) have given everyone native-level fluency.

They have not given everyone clear judgment. Fluency was never the same as clarity. Many native speakers write unclear medical English, and LLMs were trained on all of it.

The coming flood

The cost of producing fluent text has fallen almost to zero, which will cause an immense increase in volume. Meetings that once left no record now generate AI transcripts, often checked lightly or not at all. Reports that took a day now take minutes. Slide decks will multiply: well designed, professionally formatted, and in many cases without a clear point.

The flood will be uneven. Formal regulatory submissions, such as those to PMDA, are likely to remain much as they are for several years. The rise will be steepest in internal communication, literature monitoring, partner updates, and competitive intelligence, which is where many decisions are actually shaped.

Most of this material will be polished, confident, superficially plausible, and flat, with every point given equal weight. Here is an illustration (the numbers are invented):

AI English: 'The study demonstrated a generally favorable safety profile, although certain adverse events, including hepatic enzyme elevations, were observed in a subset of patients and may warrant further consideration.'

Clear English: 'ALT elevations above 3× ULN occurred in 4% of patients. This is the main safety signal and should be monitored in the next study.'

Both versions are grammatically correct. Only the second tells the reader what matters. Writing the second version requires knowing when to go back to the source data and add specific information. That is judgment, not fluency.

Few people will read the flood directly. AI will read it for them and deliver summaries, and often summaries of summaries. That creates three risks:

• Layered distortion. Each summary compresses the one before it. Each compression is a chance to lose a nuance or turn a possibility into a fact.
• Misplaced trust. Fluent summaries sound authoritative, so they are checked less than they should be.
• Invisible omissions. A bad sentence can be spotted. A safety signal that was never included cannot.

The result will be more words with less meaning. The challenge is to know what a good summary should contain, and to take action when something is missing.

Two readers, not one

Pharmaceutical English now has two audiences.

The first is the international community of physicians, regulators, and colleagues, many of whom read English as a second language. They need clear, unambiguous sentences that show which points matter most.

The second is AI agents: the systems that increasingly summarize, translate, classify, and route documents before a human sees them. These AI agents cannot ask what you meant. When a sentence is vague, each step in an automated chain can make it worse. A hedge becomes a claim, or a safety signal disappears into a summary. In a regulated industry, that is more than a language problem. It is a safety and compliance risk.

The 3Cs, revisited

To manage that risk, the basic techniques of medical writing become more important than ever:

• Clear: Each sentence has one accurate meaning.
• Concise: The reader does not have to dig through extra words to find the point.
• Coherent: Ideas follow a logical order, priorities are visible, and the relationships between ideas are obvious.

In the past, medical writing was difficult for second-language writers because producing correct English consumed their time. Today the challenge is to recognize when fluent English is unclear, bloated, or incoherent, and then to fix it.

Coherence deserves special attention. It is what AI English most often lacks, and it is what machine readers most need to transmit meaning accurately.

What this means for organizations

Training employees to use AI tools is a real but short-term need. Most will become competent on their own fairly quickly.

The lasting skills relate to judgment, in two directions:

1. Incoming: Using AI to summarize, organize, and prioritize a rising flood of information, and recognizing when a summary has distorted, overstated, or omitted what matters.
2. Outgoing: Producing English that is clear, concise, and coherent for both human and machine readers.

For employees, these core professional skills will mean more demand for their work, not less: when information is everywhere, people who can make sense of it become more valuable. Employers who invest in these skills will make faster, safer global decisions.

(End of essay.)