The cover letter has a reputation as a formality, and a lot of advice treats it as a template to be filled in. That is a waste of the one opportunity in a submission to address the person making the first decision in their own terms.
It is also commonly over-estimated in the other direction. A cover letter will not rescue a paper that is wrong for the journal, and editors are practised at reading past enthusiasm. What it can do is make sure the paper is understood correctly in the first few minutes, which is worth a great deal when the alternative is being understood incorrectly.
What an editor does with it
Read it, usually first, usually quickly, alongside the abstract. The two are doing different jobs and the editor is unconsciously comparing them.
The abstract tells them what the paper found. The cover letter should tell them why it matters here, in this journal, now. If the letter only restates the abstract in longer form, it has told them nothing and they move on.
A few things in a cover letter are read closely rather than skimmed, because they carry information nothing else in the submission does: a statement of what was believed before the paper, a declaration about related work from the same group, a note about competing manuscripts, and anything about reviewers.
The structure that works
Four short paragraphs is usually enough, and under a page is the target.
What the paper claims, in one or two sentences. Plainly, in the strongest form you are willing to defend, and without a windup. Do not spend the first paragraph establishing that your field is important.
What was known before, and what changes. This is the paragraph that does the real work and the one most often missing. Two sentences on the state of the question before your paper, then one on what should be believed after it. Editors are reading continuously in the field and this is the frame they are already using; supplying it yourself means they assess your advance rather than estimating it.
Why this journal. Not flattery. A real answer about readership: who needs to read this and why they are the people this journal reaches. If you cannot write that sentence honestly, that is useful information about your journal choice rather than a writing problem.
The declarations. Anything the editor needs to know and would be annoyed to discover later.
What belongs in the declarations
Be straightforward and slightly over-inclusive here. The cost of mentioning something unnecessary is a sentence. The cost of an editor finding it out independently is their trust in the rest of the submission.
Related manuscripts from your group that are submitted, in press or under review, particularly if they share data or participants. Any part of the work that has been posted as a preprint, presented, or published in another form. Conflicts of interest. Anything unusual about ethics, consent or data availability. And, where the journal asks for it, your disclosure of how any AI tools were used, which more journals now require at submission.
On suggesting and excluding reviewers
Most journals invite suggestions and many authors treat the field as a formality. Editors do look at it, partly for the names and partly for what the choice of names tells them.
Suggest people who know the subject well enough to assess it and who are not collaborators, co-authors, or in your department. Spread them across groups and countries. A list of four people from the two labs you work most closely with reads as exactly what it is.
Exclusions are legitimate and you do not have to justify them at length. A sentence giving a real reason is better than a long one, and better than a list of everyone who has ever disagreed with you. Editors are not obliged to honour exclusions, but a short, credible reason is usually respected.
Things to leave out
Claims about how the work will be received. An editor has no use for a prediction that a paper will be highly cited or of broad interest, and the assertion slightly undermines the paper by implying the case needs help.
Any statement about significance that the manuscript does not support. The letter and the paper are read together, and a gap between them is noticed.
The word novel, used as a substitute for saying what is new. Say what was believed before and what changes, and the novelty is self-evident. Assert novelty without that, and the editor has to work out whether it is true, which they will do sceptically.
Length. A two-page letter signals, accurately, that the claim could not be made briefly.
If you are resubmitting after a rejection elsewhere
You are generally not obliged to mention a previous rejection and most authors do not. But if you are submitting to another journal from the same publisher, or transferring, be straightforward about it, and if you have reviews, consider offering them. A paper that arrives with previous reviews and a clear account of what changed in response can be assessed faster and is sometimes assessed more favourably, because the editor can see the work has already survived scrutiny.
What matters more is that you actually changed something. A paper resubmitted unchanged after a rejection on framing will meet the same judgement, and the cover letter is where that becomes visible.