Manuscript Submission Pressures: What the Evidence Can Show

The original Publons/Clarivate Global State of Peer Review 2018 report used ScholarOne administrative records. Figure 28 reports a median of 16.4 days from agreeing to review to completing a report during 2013-2017. That excludes recruiting the reviewer and the author's revision time. It is historical platform evidence, not a current global review-cycle estimate.

The report also shows why different clocks should not be combined: invitations per completed review increased during the studied years while the median report-completion duration decreased. A story about recruitment pressure therefore needs evidence about recruitment; it cannot be inferred from one handling-time median.

Likewise, comparing selected journal receipt-to-acceptance medians over a decade can reveal variation and sensitivity to date completeness. It cannot prove that reviewer availability, AI drafting or ownership caused the change. A discontinuity in deposited dates can have several explanations, including a real editorial change.

1. Check the current journal rules. Confirm article type, scope, reporting standards and AI-use requirements on the publisher's own pages.
2. Match the clock to the deadline. A first-decision figure does not promise acceptance; an acceptance figure does not promise publication by a particular date.
3. Make the contribution assessable. Align the abstract, figures and methods with claims the evidence actually supports. Preparation can expose weaknesses, but it is not evidence that desk screening has become stricter everywhere.
4. Verify references. Check the source, relevance and correction or retraction status of important citations. Do not treat fluent prose as proof of accuracy or fabrication.
5. Leave room for revision and resubmission. The duration depends on the decision and work required. There is no defensible universal three-to-six-month rejection-cycle estimate in these sources.

A manuscript readiness review can help prioritise problems in framing, methods, references and journal fit. It cannot create missing evidence, guarantee a favourable decision or shorten a journal's queue.

The earlier version of this page quoted approximately 15% review-time growth since 2020, 15% reviewer decline, 30% AI screening adoption and 8% annual publication-charge growth without substantiating original quantitative sources. Those estimates have been removed. The associated claims of a universal recent rise in desk rejection and increased acceptance odds from pre-submission review are also withdrawn. The practical checklist above does not rely on those claims.

A useful comparison starts with the journal, the events defining the interval and the manuscripts that enter the calculation. Record the sample size alongside the median. A median from a small set of dated papers can change when a few different papers are sampled. A large sample reduces some sampling uncertainty, but it does not repair inconsistent date conventions or make the selected journals representative of all publishing.

The middle half of the observations, often reported as the interquartile range, gives more information than a median alone. A broad interval means papers in the same journal followed substantially different timelines. It does not tell an author where their own submission will fall. Article types, revision needs and editorial decisions can differ within a journal, and a table may not distinguish all those circumstances.

Missing dates matter too. If only some published papers carry both endpoints, the measured papers can differ from those without usable dates. Reporting the proportion with dates makes that limitation visible. It does not prove that the available dates refer to the original submission or that missingness is random. Similarly, a high ratio of PubMed records to Crossref records is a count comparison, not proof that the same individual papers are present in both sources.

Treat a comparison over time as a defined panel comparison. Check whether the same journals are included, whether the date and article-type rules changed, and whether flagged records were removed consistently. The median of percentage changes within journals can differ from the percentage change between the two period medians. Neither calculation should be described as a change for every manuscript or every publisher.

For a fellowship, graduation or project deadline, write down which event must occur by the deadline: submission, an editorial decision, acceptance, online publication or indexing. Those events are distinct. Confirm the requirement with the organisation setting the deadline instead of assuming that an acceptance letter and a published record are interchangeable.

Then identify which journal source addresses that event. Preserve its reporting period and whether the number covers all submissions, externally reviewed papers or accepted papers. If a source gives only receipt-to-acceptance histories, leave the unobserved publication component explicit. Adding separately reported medians cannot recover the median of the total because the components may concern different papers and their joint distribution is unknown.

Keep a dated submission log for your own manuscript: journal, submission acknowledgement, status changes, decisions and requested revisions. This can support a precise account of your experience without implying that it measures a journal-wide queue. When asking an editor about a delay, use the journal's communication guidance and the actual submission history. A population median cannot establish that an individual case is overdue.

• Publons/Clarivate Global State of Peer Review 2018: historical ScholarOne clocks, Figures 28 and 32; not a current inventory.
• PubMed search help: query dates and publication-type definitions.
• Manusights journal timeline research and corrections: selected-panel histories, versioned data and limitations.
• How to choose a journal.
• Journal AI policies: check the controlling publisher source for current policy.