Citation Anomaly Check

A private self-check for researchers: is your citation profile unusual relative to your field? Compare annual growth, top-work concentration, coauthor counts and self-citation against measured reference distributions — nothing is stored, nothing is shared.

// How this works
  • Enter an OpenAlex author (search by name, or paste an Author ID like A5108093963). The tool fetches public OpenAlex data through our proxy and computes 4 pattern signals
  • Each signal is compared against the measured distribution of the same signal across thousands of researchers in the same field — our citation pattern reference dataset, drawn from the same seeded OpenAlex sample as our h-index benchmarks
  • There are no absolute thresholds: a value is only ever flagged as "unusual relative to field" when it falls outside the middle 80% (p10–p90) of the field reference
  • Privacy: everything runs in your browser. No results are stored, no result URL is generated, this page is excluded from search engines (noindex), and lookups are rate-limited server-side
// Read this first

An unusual signal is not evidence of manipulation. Citation bursts follow viral papers, large collaborations are normal in particle physics and clinical trials, and concentrated citations often just mean one landmark work. This tool reports where a profile sits inside its field's measured range — interpretation is yours. For context on why citation-integrity tooling matters, see the reporting at Retraction Watch.

Run the check

Try:

The five signals

Each signal is a number computed from public OpenAlex records — then placed inside the measured distribution of the same number across thousands of researchers in the same field. "Unusual" means outside the field's middle 80% (below p10 or above p90). Nothing more, nothing less.

1 · Peak-year citation share

The share of lifetime citations that arrived in the single most-cited year. A very high share means citations cluster in one burst — typical for a single landmark paper, but also consistent with short-lived coordinated citing.

2 · Largest annual growth jump

The largest year-over-year multiplication of yearly citations (only years with ≥5 citations count, to avoid noise). Organic careers jump too — a widely read review or award can double citations overnight — so only extreme outliers versus the field are flagged.

3 · Median coauthors per work

Median author count across a random sample of up to 200 of the researcher's works, compared against the per-work distribution of the field. Very large coauthor counts inflate citation totals in every metric — legitimate in big-science fields, unusual elsewhere.

4 · Top-10 citation concentration

The share of total citations held by the 10 most-cited works. High concentration means impact rests on few works — a single influential paper is the commonest innocent explanation.

5 · Self-citation share (sampled)

Across the researcher's 10 most-cited works, we sample up to 200 citing works each and count how many were authored by the same researcher. Reported as a plain number with the sample size — no verdict, because published thresholds vary and we do not use cutoffs without a field-specific data basis.

Methodology & limits

  • Data source: OpenAlex API (Oct 2026 snapshot for references). Author lookup, top-10 works, a ≤200-work random sample, and up to 200 citing works per top work are fetched through the Calcuja OpenAlex proxy (24 h cache, per-IP and global daily rate limits — under load you may see "try again later").
  • Field assignment: the researcher's primary field is taken from their top OpenAlex topic and mapped to the 11 reference groups of our benchmark dataset. If no field is found, the all-fields reference is used.
  • Reference basis: growth signals come from counts_by_year of 63,665 sampled authors; coauthor counts from a seeded sample of 1,000 works per field; concentration from ≈300 authors per field from the same sample. Full methodology and downloadable CSV: /research/citation-pattern-references-2026/.
  • Samples, not censuses: the self-citation figure covers at most 2,000 citing works (top-10 papers × 200). For very highly cited researchers this is a small share of total citing works — the exact sample size is disclosed on the result card.
  • What this is not: no plagiarism check, no fraud detection, no ranking of researchers. Unusual values have many innocent explanations — field conventions, career stage, a single landmark paper, database artifacts. OpenAlex records can also contain misassigned works, which distorts every signal.
  • Known database limits: OpenAlex coverage before ~1990 is fragmentary; author disambiguation is imperfect (our Marie Curie profile documents a contaminated record); Google Scholar counts differ systematically.