The author guidelines are the only authority
Every journal publishes its requirements, and they differ — on word limits, abstract length and structure, reference style, figure resolution and file format, section order, and which declarations must appear. There is no universal standard, and general advice found online is no substitute for the page the journal publishes.
Read the guidelines before you write the final draft, not after. Rewriting a structured abstract into an unstructured one, or cutting eight hundred words to meet a limit you did not know about, is avoidable work done at the worst moment.
Where the journal states a limit, treat it as a limit. Editorial offices run technical checks, and over-length manuscripts, missing statements and wrong reference styles are returned to authors before an editor reads a word.
Technical checks happen before editorial assessment. A formatting return costs days and reflects nothing about the research.
Templates: IEEE and the rest
Many publishers provide templates, and where one exists it is the shortest route. IEEE publishes LaTeX and Word templates for its journals and conferences that carry the correct column layout, heading hierarchy, figure captioning and reference format. Elsevier, Springer Nature, ACM and most large publishers do the same.
Use the template from the journal you are submitting to, in the current version, rather than one saved from an earlier paper. Requirements are revised, and a conference template is not a journal template even within the same publisher.
If you write in LaTeX, check which engine and class file the journal expects and whether it accepts a PDF at submission or requires source files at acceptance. If you write in Word, check whether figures should be embedded or supplied separately — practice differs and both are common.
Abstract, references and figures
Abstract requirements vary by journal: some impose a word range, some require a structured abstract with named headings, some prohibit citations or abbreviations in it. Take the number and the structure from the journal rather than from a remembered convention, since both are commonly specified and easily checked.
Reference style is where most avoidable errors sit. IEEE uses bracketed numeric citation; Vancouver and NLM styles are numeric with their own conventions; APA and Harvard are author–date. A reference manager will apply any of them, but check the output against the journal’s examples — style files drift, and a mismatched style is what a technical check catches first.
Figures are usually rejected on resolution and file type rather than content. Check the minimum resolution for line art and for photographs, the accepted formats, whether colour is charged for in print, and whether captions belong in the manuscript file or with the image.
Format-free submission changes the order
Several large publishers now accept format-free initial submission: you submit in any reasonable format, and journal-specific formatting is only required if the paper is accepted in principle or invited for revision.
Where it is offered, take it. Formatting a manuscript to one journal’s style before the first decision is work that is discarded the moment the paper is rejected and resubmitted elsewhere, which is the normal outcome for most first submissions.
What is never waived, format-free or not, is the substance: the declarations, ethical approvals, data availability statement, author contributions and a manuscript within a sensible length. Those travel with the paper whatever the layout.