How do I know if my research gap is original?

Originality in a doctorate is a lower bar than most candidates assume and a different bar than they imagine. It does not require that nobody has worked nearby. It requires a contribution to knowledge that did not exist before, which can be a new result, a new method, a new context, a new synthesis, or a well-executed test of something previously assumed.

The check that actually finds collisions is a search for your intended paper rather than your topic. Write the title your finished study would have. Extract the distinctive terms from it. Search those. A topic search returns the field, which you already know, while a title search returns the specific study, which is what you are worried about.

It also helps to know which kind of gap you are claiming, because "nobody has done this" is only one of them. Lingard sorts gaps into four types: a plain knowledge deficit, a methodological or philosophical shortcoming, an unresolved controversy, and a pervasive assumption nobody has tested. The worked examples are the absence of any study relating longitudinal integrated clerkships to medical student abuse, research read through a cognitivist lens while a humanist reading is neglected, open disagreement over how professionalism should be defined, and the untested belief that technology will fix teamwork problems (Lingard, 2018, Part I, pp. 48-49). Three of those four survive even when close work already exists, so identify which one you are making before you conclude you have been scooped.

Lingard also recommends taking notes on how each paper relates to the others, not only on what each one contributes, because a gap usually appears as a pattern across studies rather than as a missing keyword (Lingard, 2018, Part I, p. 49). The review you write should carry only the material needed to position your own contribution as the next turn in that conversation (Lingard, 2018, Part I, pp. 47-48).

Cover the places where finished work sits before it is visible. Dissertation and thesis repositories hold years of research that never becomes a journal article. Preprint servers hold the work that is currently in review, which is precisely the work most likely to scoop you. Conference programmes hold the work that is currently being done.

Then ask people. Two conversations with researchers active in the area will surface things no database will, including the study someone abandoned for reasons you would rather learn now.

What do I do if someone publishes my idea while I am working on it?

First, read it properly rather than panicking. Overlap is almost never total. They used a different sample, measure, setting, or theoretical lens, and the difference is now your contribution, stated explicitly. A published paper on your question also proves the question was worth asking.

Practically, this usually means a reframing rather than a restart. Your study becomes the replication that tests whether their finding transfers, or the extension that adds the mechanism they could not test, or the contrast that shows the result depends on a design choice they did not examine. All three are publishable and all three are defensible in a viva.

Tell your supervisor the week you find out, not the month before submission. The reframing is easy early and painful late, and supervisors have seen this many times.

How thorough does an originality check need to be?

Thorough enough that an examiner naming a paper you have not read would be a surprise. In practice: a full database search, forward and backward citation chasing on the closest three works, one preprint server, one dissertation index, and a conversation with someone who has been in the field longer than you.

Do not let one search platform's silence stand as proof. Across six systematic reviews, Haddaway and colleagues found Google Scholar coverage of the included studies ranging from 68.5% at worst, meaning it missed 31.5% of the studies that review relied on, up to 100% in the review with the smallest evidence base of 37 studies, with the rest between 94.3% and 100% (Haddaway et al., 2015, pp. 11, 17). The same authors argue for searching grey literature, including theses, organisational reports, and government papers, because positive and statistically significant results are more likely to be published than negative ones (Haddaway et al., 2015, p. 3).

Then try to break your own claim with a deliberately wider search. Bramer and colleagues suggest running a more sensitive version of your strategy, then examining only what the second strategy found and the first missed, and judging whether those extra references are relevant. Add synonyms picked up from the highest ranked results, and check anything citation chasing turns up. If a paper was sitting in the database and your strategy simply did not retrieve it, put the missing term in and run it again (Bramer et al., 2018, pp. 536, 538).

Repeat the check at least twice more before submission, since three years is long enough for the field to move. Put it in the calendar rather than trusting yourself to remember, because this is exactly the task that gets postponed when writing is going badly.

Perfect coverage is not available and chasing it is a way of not starting. The obligation is diligence and honesty about scope, not omniscience.

Bramer and colleagues say plainly that exhaustiveness cannot be established, because the number of relevant articles is unknown, even though searches from 73 reviews built with their method performed comparably to searches from 258 reviews that involved information specialists at other Dutch academic hospitals, and took less development time (Bramer et al., 2018, p. 540). So write the claim to match what you actually did. State the databases, terms, citation trails, and grey literature you searched and the date you searched them, then say which of the four gap types your study addresses (Lingard, 2018, Part I, p. 49). That claim is defensible under questioning. "Nothing came up" is not.