Crafting a Title That Gets Found and Cited
Oct 7, 2026
A postdoc I've known since her thesis defense emailed me at 11pm last March with a subject line that just said "title help, sorry." Below it were six candidate titles for a paper she'd spent two years on, numbered, each one slightly worse than the last. The last line of her email: "My PI likes number 4 but I think it's boring. Which one would you pick?"
I picked number 4.
Then I told her why, because that's the part she actually needed. Number 4 was boring in the way a well-labeled drawer is boring. It said what the study was, in whom, with what design, and what it found. Someone scanning 200 search results at 7am with a coffee would have known within two seconds whether to open it. Her favorite — a clever one with a colon and a metaphor about tipping points — would have been invisible to everyone who wasn't already looking for her.
I've handled manuscripts for about twenty years now, and I've watched a lot of good work get buried under bad titles. So here's what I actually believe about titles, including the places where I disagree with the standard advice.
Your title is a retrieval device first, a rhetorical device second
Almost nobody finds your paper by browsing a journal's table of contents anymore. They find it through a database query, a reference list, a citation alert, or someone else's tweet-length summary. In every one of those contexts, the title is doing the same job: helping a stranger decide, in under three seconds, whether this paper is relevant to the question in their head right now.
That has two mechanical consequences.
First, search engines and bibliographic databases weight title words heavily. PubMed's relevance ranking, Google Scholar's indexing, your institution's discovery layer — all of them treat title terms as strong signals. If the phrase a reader would type isn't in your title or abstract, you're relying on luck.
Second, the first 60 to 70 characters do most of the work. That's roughly what shows in a search result line, a citation alert email, a reference manager list, or a conference program. Everything after the colon is a bonus, not a foundation.
I once asked a group of clinicians in a journal club how they decided what to open from a PubMed search of 140 results. One of them said, without irony, "I read the first half of the title and the journal name." That's your audience.
The catchy-versus-descriptive trade-off, honestly
You'll read a lot of advice that says clever titles get shared more. Sometimes they do. I've also sat in editorial meetings where a witty title helped a paper get picked up by a news desk, and the paper went on to accumulate citations partly because of that attention.
But I've seen the failure mode far more often than the success. The failure mode looks like this: a good paper with a title that assumes the reader already knows the field's in-jokes, published in a solid specialty journal, cited eleven times in five years, while a duller paper covering similar ground gets cited eighty times because it's findable.
My honest position: wit is a luxury good. You can afford it when you already have distribution — a high-profile journal, a big name on the author list, a topic with press interest. If you don't have that, every character you spend on cleverness is a character you're not spending on being found.
There's a middle path I like, and it's the one I recommend most: be specific enough to be interesting. Specificity is its own kind of hook. "Nurses skipped handwashing most often during shift change" is more compelling than any metaphor you could build, and it's also perfectly searchable.
Before and after
Before: "Breaking the chain: a quality improvement initiative in acute care"
After: "Hand hygiene compliance during shift handover: a stepped-wedge quality improvement study in three acute medical wards"
The second is longer and far less charming. It will also be found by anyone searching hand hygiene, handover, stepped wedge, or acute medical ward, which is to say by the exact people who should cite it.
Another one, from a manuscript I handled last year:
Before: "A novel approach to an old problem in pediatric asthma"
After: "Text message reminders and inhaled corticosteroid adherence in children aged 6–11: a randomized controlled trial"
"Novel" tells the reader nothing. I've started treating it as a small red flag — it's often what authors write when they haven't settled on what the paper actually claims.
A formula that works, and when to break it
Here's the skeleton I give students. It isn't elegant, and it's not meant to be used rigidly, but it forces you to make decisions.
[Exposure or intervention] + [outcome] + [population or setting] + [design]
Worked examples:
"Early mobilization and 90-day functional outcomes after ischemic stroke: a multicenter prospective cohort study"
"Metformin use and risk of hepatocellular carcinoma in patients with type 2 diabetes: a systematic review and meta-analysis"
"Perceived discrimination and treatment delay among Black women with breast cancer: a qualitative interview study"
Boring? Yes. Each of them is also a complete sentence of information, front-loads the searchable terms, and tells an editor instantly whether it fits the journal's scope.
Break the formula when the finding itself is the headline and it's unambiguous — "Dexamethasone reduces 28-day mortality in patients receiving respiratory support" is a declarative title that earns its confidence. Use declarative titles sparingly and only when your design supports a causal claim. I've desk-rejected observational papers whose titles asserted effects their methods couldn't deliver, and reviewers notice this immediately. A title that overclaims poisons the whole read.
Say the design out loud
This isn't just good practice, it's a reporting-guideline requirement in several cases. CONSORT asks you to identify a randomized trial in the title. PRISMA 2020 asks you to identify the report as a systematic review. STROBE asks you to indicate the study design with a commonly used term in the title or abstract. You can find all of these through the EQUATOR Network library of reporting guidelines, and the full PRISMA materials at the PRISMA statement site.
Editors check this. I check this. When a trial arrives with no "randomized" anywhere in the title, my first thought isn't "sloppy title," it's "what else did they skip?" Fair or not, that's how the judgment forms.
Finding your keywords without guessing
Most people pick title words based on how they talk in lab meeting. That's the wrong reference class. You want the words your readers type.
Three things I do, and they take about fifteen minutes:
Check controlled vocabulary. Look up your key concepts in MeSH. If the indexed term is "Medication Adherence" and you've written "treatment compliance" throughout, use both — the MeSH term somewhere in title or abstract, your preferred phrasing where it reads naturally.
Run the search your reader would run. Literally type it. A query like medication adherence text message reminders randomized on PubMed will show you how the best-cited papers in your space phrase things, and whether your title would appear at all.
Read the first 60 characters aloud. If a colleague outside your subfield can't tell what the paper is about from that fragment alone, rewrite.
Keyword stuffing, by the way, doesn't work and reads badly. I've seen titles that crammed in four synonyms for the same construct. Reviewers find it grating, and databases aren't fooled.
The mistakes I see most
Question titles that don't need to be questions. "Does vitamin D supplementation prevent falls in older adults?" is fine for a review of uncertain evidence. It's weak for a trial that answered the question. Answer it.
Acronym-first titles. If your trial is known as PROTECT-HF among forty people, leading with the acronym costs you everyone else. Put the acronym after the colon.
Vague scope words. "Factors associated with," "aspects of," "insights into," "a study of." These are the title equivalent of clearing your throat. Cut them and see if the sentence still works. It usually does.
Omitting the population. Age range, condition, country, setting. These are among the most common search filters, and they're how someone decides your paper is about their patients.
Titles longer than about 15 words. Not a law, but past that point readers stop reading and some journals impose character limits anyway. Check the author instructions before you fall in love with 22 words.
The hidden-finding title. You tested four outcomes and the title mentions none of them. If a single result drove the paper, let it show.
A ten-minute title check
Run this before you submit. It catches most problems.
Do the first 60 characters contain the two words a reader would search?
Is the study design named, in the words a database would index?
Are population and setting specified?
Does the strength of the claim match the design — association language for observational work, effect language only for trials?
Is every acronym either universally known or placed after the colon?
Have you removed "novel," "a study of," "insights into," and "exploring"?
Does it fall within the journal's stated title length?
Would a competent colleague in an adjacent field understand it without reading the abstract?
Does it match the abstract and the primary outcome, exactly?
That last point matters more than people expect. Title-abstract mismatch is a documented source of reviewer irritation and, in trials, sometimes evidence of outcome switching. The ICMJE recommendations and COPE guidance both circle this territory from different directions — the first on honest reporting, the second on what happens when reporting goes wrong. Your title is a promise. Keep it.
Where to spend the extra hour
Write six titles. Not two. Six forces you past the obvious ones, and the fifth is often where something good shows up. Then send them to one person inside your field and one person outside it. The outsider's confusion is the more useful data.
I also tell people to draft the title twice: once before analysis, as a working hypothesis, and once after the results are in, when you know what the paper actually says. The gap between those two versions is often the most interesting thing about the project.
If you're doing the keyword work I described above, it helps to have your searching, reading, and drafting in one place rather than scattered across tabs and documents. That's the thinking behind PACR, an AI-native research workspace — you can run a literature search, keep the papers you're benchmarking your phrasing against, and draft alongside them without losing the thread between steps. A workflow I've seen work well: pull the twenty most-cited papers on your exact question, list their titles in one column, and look at what they all do that yours doesn't.
Boring titles with precise words get read. Clever titles with vague words get admired for a week and then forgotten. If you've got two years of your life in this paper, spend the extra hour and set up a workspace where the drafting and the searching sit side by side.
And if your PI likes the boring one, your PI is probably right.
