A systematic review is judged as a study, not as a summary. If the method section cannot be followed and repeated, the findings will not be trusted regardless of how much reading went into them.
Know what a systematic review actually is
A systematic literature review is a study. It has a question, a method, a dataset made of published work, and findings. That is different from a narrative review, which is an expert's informed reading of a field, and different again from a background chapter, which explains what is already known so the reader can follow what comes next.
Most manuscripts described as systematic reviews are narrative reviews with a search string added. Reviewers detect this quickly, usually from the absence of a protocol, the absence of screening counts, or a synthesis that describes studies one after another without comparing them.
If your work is a narrative review, call it one. Well-executed narrative reviews are publishable. Mislabelled ones are not.
Write the protocol before you search
The protocol states the question, the databases, the search strings, the inclusion and exclusion criteria, the screening procedure, the appraisal instrument and the synthesis approach, all before screening begins. Writing it afterwards defeats the purpose, because the point is to prevent decisions being shaped by what you have already seen.
Registration is expected in health research and increasingly common elsewhere. Even where no registry applies to your field, a dated protocol file you can show a reviewer is worth having.
Make the question answerable
“What is known about X?” cannot be answered systematically, because nothing decides when the answer is complete. A structured question, covering population, intervention or exposure, comparator, outcome, and study designs eligible, tells you what to search for and when to stop.
Frameworks such as PICO, SPIDER or PEO exist to make that structure explicit. Use whichever fits your field, and state which one you used.
The search strategy is the method
Report it fully enough that a reader could re-run it: every database, the exact strings including Boolean operators and truncation, field tags, filters, date ranges and the date each search was run. Put the full strategy for at least one database in an appendix.
Search more than one database. Add reference-list checking and, where your field expects it, grey literature. Reviewers routinely ask why a particular database was omitted, and “we only had access to one” is a legitimate limitation if you state it.
Screen against written criteria, in two stages
Title and abstract first, then full text. Two screeners working independently, with disagreements resolved by discussion or a third reader. Report the agreement statistic if your field expects one.
Count everything: records identified, duplicates removed, records screened, full texts assessed, exclusions with reasons, studies included. This is what the flow diagram shows, and a review without one is difficult to trust.
Appraise quality, and report what you found
Use an established instrument appropriate to the designs you included, and report the results study by study rather than as a single average. If most of your included studies are weak, that is a finding about the field and should shape your conclusions rather than be buried.
Synthesise rather than summarise
The most common weakness in submitted reviews is a results section that walks through studies in turn. That is a summary, not a synthesis.
Group studies by construct, method or finding. Show where results agree, where they conflict, and what explains the conflict: population, measurement, design or context. Where meta-analysis is appropriate and the studies are comparable enough, do it and report heterogeneity honestly. Where it is not appropriate, say why, and use a structured narrative synthesis method rather than an unstructured one.
Interpret, and be specific about limits
The discussion should answer the question you asked, say what the evidence can support, and say plainly what it cannot. Reviewers are more forgiving of acknowledged limitations than of claims that outrun the evidence.
Be concrete: language restrictions, date ranges, databases not searched, designs excluded, quality of the included studies. A specific limitations paragraph reads as competence. A generic one reads as a formality.
Report to the standard your field uses
Follow the reporting guideline that applies to your discipline and study type, and include the checklist as supplementary material. Check the target journal's author guidelines separately. Journals frequently add their own requirements on top.
Finally, before submitting, give the manuscript to someone who has not been involved. A reader who cannot follow your screening process from the text alone is telling you something a reviewer would tell you four months later.
Where reviews most often fail
Across the reviews we are asked to look at, three problems recur: no protocol written before screening, a search strategy reported too vaguely to repeat, and a results section that summarises rather than synthesises. All three are fixable, and all three are much cheaper to fix before submission than after a four-month review.