A search string is built, not typed
Start from the claim rather than the title. Pull out the population, the intervention or concept, and the outcome, then find the words the literature actually uses for each — which are frequently not the words you would use. Those become the blocks of a search string, joined so that synonyms widen it and the concepts narrow it.
That string is worth keeping. It can be re-run when a reviewer asks for something more recent, adjusted when a concept turns out to be too broad, and pasted into a methods section when a marker asks how the search was conducted.
Retrieval is the test, not relevance
A source that looks perfect and cannot be opened through your library is not a source. Everything supplied here is checked for retrieval on your side before it reaches the list, because the person marking your work will take the same route you would and will draw the obvious conclusion when it fails.
This has become more important, not less. Reference lists produced by a language model contain entries with real journals, believable authors and page ranges that lead nowhere, formatted immaculately. No tool detects that. Opening each item does.
The list is annotated, so you know what you have
Each source comes back with what it actually says, what it can support, and where it is weak. A list of twenty titles you have not read is not an advantage; it is twenty decisions still to make under time pressure.
That annotation is also what stops the commonest failure in a draft, which is citing a source for a claim adjacent to the one it makes. If you know what a paper concluded, you cannot accidentally attach it to something it did not.
Three steps, then the record keeps itself.
Send the claim, not the title
What you are actually trying to argue, plus your school and level so the search runs against the databases you can reach.
Get the string and the list
The search as a re-runnable string, and sources annotated with what each one supports and how strongly.
Check them yourself
Every item retrievable through your own library. If one does not open for you, it is replaced.
Queries on record.
What if my library does not carry a journal?
Then it is not a usable source for that paper and it gets replaced, however good it looks. Interlibrary loan exists and is worth using when there is time, but a reference you cannot produce on request is a liability in the work rather than an asset. Everything supplied here is checked against what you can actually reach.
How do I know a source really says what I claim?
By reading the part you are citing rather than the abstract, which is where this goes wrong most often. Abstracts state findings without their conditions, so a claim that holds for one population under one set of circumstances gets cited as though it held generally. The annotation returned with each source names those conditions explicitly.
Can you find sources for a topic with almost no literature?
Sometimes the honest answer is that the literature is thin and the paper should be framed to acknowledge that, which is a legitimate finding rather than a failure. What is not acceptable is padding a list with tangential work to reach a number. If your topic is genuinely under-researched, saying so with evidence is stronger than pretending otherwise.
Is grey literature acceptable?
It depends on your rubric and on what you are using it for. Government reports, professional body guidance and registry data are often exactly right for context or for practice recommendations, and wrong as evidence of effect. Where it is used, it is labelled as what it is rather than presented alongside peer-reviewed work as though equivalent.
Do I get the search itself or just the results?
Both, and the search is the more valuable half. A documented string can be re-run when a committee asks for more recent work, narrowed when a concept proves too broad, and dropped into a methods section when somebody asks how you searched. The results are this paper; the string is every paper after it.
Elsewhere in the hall.
Send the assignment and the deadline.
A specialist credentialed in your own field reads what you actually have and replies within two hours with a plan and a price. The reading costs nothing, and if the honest answer is that we are not the right hall for it, you will get that instead.
