How to Choose Keywords for Your Paper

The keyword field sits near the bottom of the submission form, below the title, the abstract and the author list. It is the box most authors fill in last and fastest, usually with terms lifted straight out of the title they typed two fields above.

How to choose keywords for a research paper is a smaller problem than it looks and a more consequential one. The field holds a handful of terms. Each one is a slot, and a slot spent on a word your title already carries has bought nothing at all.

The rest of the journal submission process, from the cover letter through to an editor's screening pass, is covered separately. This piece stays inside one box on the form.

How to Choose Keywords for a Research Paper: Start With the Journal's Number

The journal sets how many keywords it wants, and the figure sits on the submission form or in the author guidelines beside the abstract limit. Draft to that number rather than to a habit. A field taking five terms and a field taking ten are different exercises, and a list built for one does not compress or stretch cleanly into the other.

Some journals don't ask for any author keywords. The PLOS ONE submission guidelines state that "the entire title should be 250 characters or less," and the abstract is limited to 300 words. There's no mention of author keywords on the page. Check the form before you spend an evening on a list the journal has no box for.

Why Repeating Your Title Words Wastes the Slot

A database record already stores your title in full and searches it. A keyword repeating a word from the title puts that word into the record a second time, which changes nothing about whether the record is found. The slot has been spent and no new search reaches your paper because of it.

What earns a slot is a term a reader would plausibly type that your title does not contain: the synonym you decided against, the alternative spelling, the method rather than the topic, the population rather than the finding. Each of those is a search that currently misses your paper altogether.

Keyword type	What it adds to the record	Example
Synonym for a title word	Reaches searchers using the other name for the same thing	myocardial infarction, where the title says heart attack
Alternative spelling	Catches the variant the title does not carry	randomised alongside randomized
Method or instrument	Reaches readers searching by technique rather than by topic	semi-structured interviews; thematic analysis
Population or setting	Reaches readers who search by who was studied	adolescents; primary care; rural clinics
Broader category	Places a narrow paper inside a field people browse	health services research
A word already in the title	Nothing, since the record already stores the title	spend the slot elsewhere

Which terms end up in the title is a separate decision, made earlier and under its own constraints, and the guide to how to write a good research title covers it. The keyword list is drafted after the title is settled, and its job is to work around what the title already took.

Controlled Vocabulary, and How to Check a Term in MeSH

The Medical Subject Headings thesaurus is the one biomedical authors meet first. A controlled vocabulary is a fixed list of subject terms an index uses. Each concept has one preferred term, and the other variants are mapped on to this. The National Library of Medicine describes MeSH as "a controlled and hierarchically-organized vocabulary" that "is used for indexing, cataloging, and searching of biomedical and health-related information". MeSH is a controlled vocabulary used to organize the indexing of articles.

Checking a candidate takes about a minute in the MeSH Browser on the NLM site. Search your term and read what comes back. If it's a preferred heading, use it as written. If it is an entry term mapping onto a different heading, the heading it maps to is what readers of that database search with, and the version in your head is the variant nobody types.

Fields outside biomedicine often have no equivalent thesaurus, and the working substitute is your own reference list. Take the ten papers you cite most closely and read their keyword lines. Terms recurring across several of them are the vocabulary your subfield already uses, whether or not anyone has formalised it into a list.

How to Test a Candidate Keyword Before You Submit

Testing a keyword means running it as a search and reading what comes back. Type the candidate into the database your field actually uses and look at the first page of results. If your paper would sit naturally among them, the term is doing its job. If the results come from a different literature entirely, the term means something to that database other than what it means to you.

This indicates two failures. First, if a search returns almost nothing, that means you've used a term that's either too specific or so new that nobody is searching for it yet. If a search returns tens of thousands of records, that means you've used a term that's too general to get a reader to one specific paper, and all it'll do is put your work at the back of a very long queue.

Finding the terms that sit between those two extremes is easier with a list to react to than with an empty field. A keywords generator proposes candidates from your abstract, and the useful part is the terms it suggests that you would never have reached, since your own vocabulary is the thing being tested here.

How the Title, Abstract and Keywords Divide the Work

Why a Clever Title Costs Citations

A title built on a pun, an allusion or a borrowed quotation spends its most valuable position on words nobody searches for. MEDLINE's automatic indexing relies on both of these fields. The title is the first field a reader reads and one of the two that MEDLINE's automatic indexing works from. A quotation-and-colon opening pushes the actual subject into the second half, where it arrives after the reader has already decided.

ICMJE asks authors to avoid abbreviations in the title. PLOS ONE asks for titles that are "specific, descriptive, concise, and comprehensible to readers outside the subject field", capping the full title at 250 characters and the short title at 100. Both describe the same requirement from different directions: the title has to work for a reader who does not already know your subfield.

One piece of folklore is worth clearing away first. Neither NLM nor PubMed publishes a recommended title length, and searches for one return studies indexed in PubMed that analysed title length against citation counts, which is a different thing from a policy. The verified anchors are ICMJE's note that some journals require a short title of no more than 40 characters including letters and spaces, and whatever limit your own target journal states.

How a title should be built, declarative or descriptive, where a colon earns its place, is covered in the guide to how to write a good research title. What matters for discoverability is narrower: which search terms the title takes, and therefore which ones it leaves for the other two fields.

Divide the Search Terms, Do Not Repeat Them

The three fields work when each carries search terms the other two don't. Write out every term a reader might plausibly use to reach your paper, then assign each one to exactly one field: the two or three strongest to the title, the near-synonyms and method terms to the abstract, and whatever is still unclaimed to the keyword list.

Take a study of burnout among junior doctors in emergency departments. The title carries burnout, junior doctors and emergency department. The abstract has room for the terms the title had to leave out: resident physicians, emergency medicine, occupational stress, cross-sectional survey. The keyword list then holds what neither field reached, which might be the instrument used, the country, or the older name the same concept went by ten years ago.

Field	Read by	Should carry	Should not carry
Title	Every reader, every result list, and MEDLINE's automatic subject indexing	The two or three terms a reader is most likely to type, in plain words	Puns, quotations, abbreviations, or the subject buried after a colon
Abstract	Result-list previews, database search, and MEDLINE's automatic subject indexing	Synonyms and method terms the title had no room for, early in the paragraph	The title's own terms repeated three more times
Keywords	The journal's site search, the publisher's metadata feed, and databases that read author keywords	Whatever the other two missed: alternative spellings, population, instrument, superseded names	Any word already sitting in the title

The common failure runs the other way. The same five terms appear in the title, in the abstract's opening line and in the keyword field, assuming that repetition helps. Repetition inside one record buys far less than a term the record doesn't contain at all: a search for a synonym you never wrote returns nothing.

A title generator earns its place here for a slightly different reason than the obvious one. Producing five candidate titles shows which terms survive every version and which keep getting cut, and the ones that keep getting cut are the beginning of your keyword list rather than a set of rejects.

What a Reader Sees Before Deciding to Click

A result list shows four things: the title in full, the journal and year, the authors, and the first line or two of the abstract before it truncates. That is the entire basis for the decision to open a paper or scroll past it. Everything else in the record stays invisible until someone clicks.

This gives the abstract's first sentence a job the rest of the paragraph does not have. A first sentence establishing that the topic matters tells a scrolling reader nothing, since they searched for that topic thirty seconds ago. A first sentence naming what was done and what was found is the one that survives truncation, and how to write an abstract for a research paper covers building the rest of the paragraph behind it.

Author keywords are almost never displayed in a result list at all. They work upstream of it, helping decide whether the record appears in the list in the first place, which is what makes a keyword copied from the title a wasted slot twice over. It widens no search, and no human reads it.

The Ten-Minute Check Before You Submit

Put the three fields on one screen, side by side, before the submission form is open. Underline every search term in each of them. A term underlined twice is a decision waiting to be made: keep it in the field where it works hardest and replace it in the other.

Then run your own title as a search in the database your field uses. If the first page of results is full of papers doing roughly what your paper does, the title carries the right terms. If it returns a different literature, the title is pointing somewhere you did not intend, and the fix belongs in the title rather than in the keyword line.

Earlier in the process, the cover letter is covered separately, along with a full journal submission checklist for the day itself.

Three years from now a stranger will find this paper by typing a phrase into a box, or will fail to, and what that phrase has to match is fixed the moment the form is submitted. The free academic writing tools for each of the three fields are available at exactly that moment. Which leaves the separate question of what the paper does with a reader once they have clicked.

What the Keyword Field Does Not Control

Your keyword list does not produce the subject terms on a PubMed record. The National Library of Medicine's own training material states that since spring 2022, "MEDLINE records have MeSH terminology applied completely automatically using the title and abstract of the publication as supplied by the publisher." An author keyword field is no part of that input.

It has no bearing on whether the paper is indexed at all either. NLM decides whether a journal's scientific and editorial character merit its inclusion in MEDLINE, and Scopus and Web of Science run their own journal-level reviews. Each of those decisions is made about a whole title, usually years before your manuscript existed.

What the field does control is real enough: the journal's own site search, the publisher's metadata feed, and every database that does read author keywords. Those are the searches where a well-chosen term earns its slot, and they are the reason the box deserves more than the forty seconds it usually gets.

Keywords are one input to discoverability, and getting a paper indexed and found covers the rest. On the submission itself, the cover letter is the document editors read first.

The keyword box is one of three fields a submission form collects for discoverability, and TextPulse's free academic writing tools cover each of them. Filling all three in one pass down the form is how the same five terms end up in the title, the abstract and the keyword line, which is three copies of one search and none of the other two.