Felix's agency wrote confident copy. Every landing page had a strong verb in the headline and a bold claim in the first line. Then his clinical advisor read the draft and asked where the second sentence came from.
That one question killed the piece, not because the writing was weak, but because it was built for a reader who doesn't exist in digital health: someone who decides fast, trusts a confident tone, and doesn't check the citation.
Generalist agencies are good at what they do. The skill isn't the problem. It was built for a different reader than the one Felix is actually trying to reach, and that mismatch shows up in almost every piece before it ever gets to a reader at all.
The persuasion playbook assumes a buyer who knows less than the writer
Most content frameworks (AIDA, PAS, before-after-bridge) come from consumer and general B2B marketing, where the reader usually knows less than the writer and needs to be walked from a felt problem to a confident answer. In that world, confidence reads as authority. Urgency reads as relevance.
A clinician or a compliance-minded reviewer flips both of those. They often know more about the mechanism than the person who wrote the page. Confidence without evidence doesn't read as authority to them. It reads as a reason to stop trusting the rest of the page. Urgency doesn't read as relevance either. It reads as pressure on a decision that shouldn't be rushed.
The tell shows up fast: a headline promising a result the underlying data doesn't support, a statistic with no source, a claim that overreaches by a single word nobody flagged before it published. None of these look like errors to a generalist editor. They look like good copy. That's exactly the problem.
A confident claim is the first thing a clinical reader tests, not the thing that convinces them
A generalist writer treats "shown to reduce" and "associated with a reduction in" as interchangeable. To a marketing ear, they sound the same. To a clinician, they're different claims about different kinds of evidence, and swapping one for the other changes what the sentence is actually asserting.
This is where a lot of health tech copy quietly loses its reader. Usually not in the big claims, which get reviewed before anything ships, but in the small ones: the adjective that got upgraded during an edit pass, the hedge that got cut because it "read weak" in a Google Doc comment. A clinical reader catches that swap in about one sentence, and once they catch it, they read the rest of the page differently. Every claim after that gets the skeptical read, including the true ones.
That's the expensive part. One sentence gets flagged, and the whole piece loses credibility on the strength of that single word choice. No amount of good writing around it earns the trust back on the same page.
The patient and the clinician are reading the same page for different reasons
A generalist agency coming from consumer content tends to write one persona and multiply it across a site. That works fine when every reader wants roughly the same thing. It doesn't work in digital health, where a patient looking for reassurance and a clinician looking for a mechanism can land on the exact same page minutes apart.
Outcome language ("feel better within weeks") answers the patient's question. It does almost nothing for the clinician, who wants to know the mechanism, the population it was studied in, and what "better" was actually measured as. Write only in outcome language and the clinician bounces. Write only in mechanism language and the patient bounces, because nobody reassured them first. A generalist agency, used to writing for one buyer at a time, usually picks one and loses the other, then wonders why engagement looks fine but referrals from clinicians never show up.
Regulatory caution isn't a line you add at the end
The generalist pattern: write for persuasion first, then send the draft to legal or a clinical reviewer for cleanup. The cleanup often guts the piece, because the strongest lines are exactly the ones a reviewer flags. Or the founder catches it first and rewrites the draft himself the night before launch, which is its own kind of expensive. It costs founder time that was supposed to go somewhere else, and it costs the agency relationship, one late-night rewrite at a time.
A specialist writes inside the constraint from the first draft. Claims get sourced as they're written, not chased down afterward. Anything that needs a named reviewer gets flagged in the brief, not discovered in review. The argument gets built out of language that can survive a compliance pass, instead of language that has to be defended after the fact in a comment thread nobody has time for.
Voice inconsistency is a symptom, not the root problem
Founders often describe the issue as inconsistent brand voice: the blog sounds one way, the LinkedIn posts sound another, and the fundraising deck sounds like neither. Most of the time, it's what happens when nobody on the writing team actually owns the clinical register, so every piece gets written from scratch by whoever's available, at whatever confidence level feels right that day.
A generalist writer without health tech experience has no fixed sense of how confident a given claim is allowed to sound. So the confidence level drifts, piece to piece, reviewer to reviewer. Fix that one variable and a surprising amount of the "voice inconsistency" complaint disappears on its own.
The fastest way to spot the mismatch: count the rewrite
Here's a diagnostic that takes less time than most content audits. Pull the last 4 pieces your current agency delivered. Count how many sentences per piece needed correcting for clinical or regulatory accuracy after your own review, not just typos or style notes.
One or 2 sentences an article is normal editing. Every article needing a paragraph rewritten, every quarter, isn't a phase you'll grow out of. It's a structural mismatch between the frameworks the agency was trained on and the reader you're actually writing for. That gap doesn't close with a better brief or a longer style guide. It closes by working with someone who never needed the correction in the first place.
Generalist agencies aren't wrong for everything a health tech company writes
Worth being precise here, because the honest answer isn't "never use a generalist." Plenty of what a digital health company publishes doesn't touch a clinical claim at all: culture posts, hiring content, general commentary on the market, most of what goes out around a funding announcement. A generalist agency can write that well, sometimes better, because none of it needs the domain fluency.
The mismatch shows up specifically where a claim about the body, a condition, or an outcome enters the page. That's a narrower problem than "our content isn't working." It's worth sorting your own content calendar into the pieces that touch clinical claims and the ones that don't, before deciding the whole engine needs replacing.
Ask to see the piece that failed, not just the ones that worked
Every agency will show you a portfolio of finished, approved work. That tells you almost nothing about how they handle the moment a claim gets rejected, because you're only seeing what survived.
Ask instead for an example of a draft that got flagged by a clinical or compliance reviewer, and what changed between that draft and the version that shipped. An agency with real health tech experience will have one ready, and it'll usually be a small, specific fix: a claim narrowed, a source added, a comparison removed. An agency without that experience will either not have an example, or the story will be about a much bigger problem than a single sentence. That gap tells you more in 5 minutes than a case study will in an hour.
What actually changes with a specialist
Fewer cycles between a finished draft and clinical sign-off, because sourcing happens while the piece is being written, not after. The same research written twice on purpose, once for the clinician and once for the patient, instead of written once and hoped to work for both. A working sense, built into the first draft, of how confident a sentence is allowed to sound given the evidence actually sitting behind it.
It also changes speed, in a way that looks counterintuitive at first. A generalist agency that needs 3 rounds of clinical correction per article is slower in practice than a specialist who gets it close to right the first time, even if the specialist's first draft takes a day longer to deliver. The clock that matters is time to a publishable page, not time to a first draft.
That's most of what PulseCopy's content engine is built around for founders like Felix: a monthly cycle where the brief already accounts for who's reading and what will need sourcing, so the clinical review catches almost nothing that wasn't already handled.
The question worth asking, whether you're screening a new agency or deciding whether to renew with the one you have, isn't whether they write well. Plenty of agencies write well. It's whether they've ever written for a reader who already knows more than they do about the subject, and built the habit of proving it on every claim instead of just the big ones.