How to Cite Your Own Work Without Self-Promotion

Key Takeaways

• Citing your own prior work is legitimate, and sometimes required, whenever a reader needs it to follow your methods, data, or line of results.
• The risk sits in patterns: tangential references, padded introductions, preprints cited when published versions exist, and your own review cited where primary sources belong.
• A reader test settles most individual cases. If the sentence loses meaning without the citation, keep it; if the citation only attaches your name to a claim, cut it.
• Institutions police this behavior. Clarivate has suppressed journals from its Impact Factor calculations over self-citation problems, and the Committee on Publication Ethics treats manipulated citation as misconduct.
• Before submission, count your own papers in the reference list. Many experienced authors treat one in five references as a working ceiling for a single paper, with field norms varying on either side.

1. Introduction

Most research papers are follow-ups to earlier work by the same authors. That creates a recurring moment of doubt at the reference list: which of your own papers belong there, and how many is too many? The stakes have risen because citation counts feed hiring, funding, and national evaluation systems, so editors and reviewers scan for inflation. Journals have lost their Impact Factor over citation behavior, and individual researchers have seen metrics questioned once self-citations were excluded.

Colleagues hand out conflicting advice. Some say never cite yourself; others say cite anything remotely relevant. Neither extreme matches what editors actually check. This article gives a decision rule for individual citations, the patterns that trigger objections, phrasing that keeps a self-citation informative, and a short audit to run before submission.

2. When citing your own work is the right call

The reader test covers most cases: if deleting the citation leaves a sentence unable to do its job, the citation belongs. Four situations consistently call for it.

Method and instrument continuity. When your sampling protocol, code, apparatus, or survey instrument was described in an earlier paper, cite it at the point of use and add a one-line summary so the current paper stands alone. A team publishing a 2023 evaluation of a model first specified in their 2021 methods paper should cite the 2021 paper in the methods section, next to a sentence summarizing the architecture, so readers can follow the analysis without leaving the page.

Direct extension. A new result that tests, refines, or contradicts your prior finding requires the citation. Omitting it misleads readers about the state of evidence and invites a reviewer to ask why the baseline is missing.

Data reuse. When the analysis runs on a dataset you published before, cite the original publication and the dataset's DOI. Both matter: the paper explains the design, the deposit lets readers verify the files.

Corrections. If your new paper corrects an error in an earlier one, the citation is mandatory. Readers tracing the record need the link between the flawed claim and its repair.

Cite at the point where the dependence occurs, with the specific contribution named, for example: "using the adherence thresholds established in our 2019 cohort (Author et al., 2019)." That sentence tells the reader what the older paper supplies and why it appears.

3. The patterns that read as self-promotion

Reviewers rarely object to a single self-citation. Objections target repeated shapes across a manuscript or a career.

• Tangential overlap. You cite your own paper because the topic overlaps, though it contributes no method, data, or result to the argument. A reviewer checking the reference will find nothing the sentence needs.
• Introductory padding. Your opening paragraph cites three of your papers where a recent review or another group's primary study carries the claim better. Openings get the closest reading, so padding there is the easiest to spot.
• Version stacking. The reference list includes both your preprint and its peer-reviewed version. Cite the published version; cite the preprint only when its content genuinely differs, as with conference proceedings that later gained appendices or data.
• Review substitution. A claim cites your own narrative review while the primary studies behind it sit uncited. Primary sources carry the weight; reviews are signposts.
• Revision stacking. Several of your earlier papers enter the reference list after a reviewer or editor hints at "coverage." The Committee on Publication Ethics, whose guidance most journals adopt, classifies citation requests unrelated to a manuscript's content as citation manipulation, and authors can decline such requests in their response letter.

Indexers watch aggregate behavior as well. Clarivate, which produces the Journal Citation Reports, has suppressed journals from its Impact Factor calculations when self-citations or coordinated citation exchanges distorted the record, and that damage reaches a journal's standing well before any sanction touches an author. Author-level metrics face the same scrutiny: hiring and tenure panels that use Scopus or Web of Science can exclude self-citations from citation counts, so inflating the record this way survives only until someone runs the exclusion.

4. Phrasing that keeps a self-citation informative

A self-citation earns its place by stating what the earlier paper contributes at that exact point. Compare two versions of the same reference list entry.

Weak: "Prior work has examined this relationship (Author, 2018; Author, 2020; Author, 2022)."

Stronger: "Our 2018 cohort study established the baseline infection rates that the present analysis re-examines (Author et al., 2018)."

The second version gives the finding in words, because readers of the current paper may never open the older one. One sentence of substance beats a bare pointer.

Three techniques cover most phrasing decisions. Lead with the finding and follow with the citation, using natural self-reference such as "we previously showed" or "in our earlier trial." Summarize borrowed methods in one line before citing, so a reader can keep going without hunting for a PDF. Vary placement across papers: if every manuscript you write opens by citing your own last paper, the pattern itself becomes the signal, whatever your intent.

Treat the older paper as evidence for a specific claim at a specific point, then move on. A citation that needs a paragraph of justification is a citation to cut.

5. Decision table and pre-submission audit

The table compresses the decisions above into one block you can keep beside your manuscript.

After the table, run a short audit:

1. Count the references authored or co-authored by you. If more than about one in five entries is yours in a typical empirical paper, examine each one against the reader test. This is a working heuristic rather than a published threshold.
2. Delete each self-citation mentally. Sentences that still work without it lose the citation.
3. Confirm the version cited. The peer-reviewed version comes first; the preprint appears only where its content genuinely differs.
4. Compare with two or three recent papers in the target journal to gauge field tolerance. Clinical journals expect current external literature; mathematics and the humanities tolerate longer arcs and heavier reliance on books.

6. FAQ

Q1. How much self-citation is too much?

No single number applies across fields. A 2019 analysis in PNAS by John Ioannidis and colleagues put self-citations at roughly one in eight of the average researcher's total citations, with wide variation across disciplines and career stages. Per-paper judgment matters more than the aggregate: a reference list dominated by your own work, or the same cluster of your papers appearing in every new manuscript, draws scrutiny before any percentage threshold does.

Q2. Can I cite my own thesis or dissertation?

Yes, when it contains material published nowhere else, such as pilot data or an extended appendix. If the thesis content has since appeared in peer-reviewed papers, cite the papers instead, and check the target journal's policy; some journals discourage thesis citations when published equivalents exist.

Q3. A reviewer asked me to add citations to my earlier papers. Do I comply?

Comply where the papers genuinely support the manuscript, and decline the rest in your response letter with a specific reason, such as no shared method or data. COPE's guidance on citation manipulation covers requests unrelated to a manuscript's content, and editors arbitrate when authors and reviewers disagree about coverage.

Q4. Do the rules differ in the humanities?

The arithmetic changes. A historian or philosopher may cite their own monograph to point at a fully developed argument, and books carry more weight than in clinical or laboratory fields, so one monograph citation can stand in for several. The reader test stays the same: the citation must give the reader something the sentence needs.

7. Conclusion

Self-citation is a citation with your name on it, judged by the same standard as every other citation: does the reader need it at this point in the argument? Use the reader test for individual cases, the decision table for edge cases, and the audit before submission. The next manuscript offers the test run: count your entries, apply the reader test to each, and keep only the citations a stranger would need to follow the paper.