{
  "measurement": "citation-extraction-server-side",
  "date": "2026-08-02",
  "question": "Given only a URL, does a purpose-built citation reader beat the established one — and if not on coverage, then on what?",
  "method": {
    "endpoint": "https://provinglab.dev/mcp",
    "tool": "extract_citation",
    "runs_per_url": 2,
    "urls": 26,
    "fields_read": "Highwire citation_*, Dublin Core, PRISM, schema.org JSON-LD, OpenGraph, <title>",
    "network": "Cloudflare Workers edge (FRA)",
    "load_average_at_start": 0.25,
    "control": "Every reported block was re-fetched from an unrelated network with a browser user-agent to check whether the block was real."
  },
  "results": {
    "records_returned": 13,
    "blocks_reported": 13,
    "type_classification_correct": "13/13",
    "required_fields_complete": "13/13",
    "false_block_reports": 0,
    "seconds": {
      "median_all": 0.35,
      "median_successful": 0.74,
      "median_blocked": 0.22,
      "p95": 3.46,
      "fastest": 0.15,
      "slowest": 12.52,
      "baseline_no_fetch": 0.15
    }
  },
  "per_source": [
    {
      "field": "Medizin",
      "source": "PMC",
      "url": "https://pmc.ncbi.nlm.nih.gov/articles/PMC10418250/",
      "seconds": 0.51,
      "outcome": "record",
      "type": "Zeitschriftenaufsatz",
      "block_reason": null
    },
    {
      "field": "Medizin",
      "source": "BMJ Open",
      "url": "https://bmjopen.bmj.com/content/13/8/e072428",
      "seconds": 0.17,
      "outcome": "blocked",
      "type": null,
      "block_reason": "http-403"
    },
    {
      "field": "Medizin",
      "source": "BMC Public H.",
      "url": "https://bmcpublichealth.biomedcentral.com/articles/10.1186/s12889-023-16556-0",
      "seconds": 1.01,
      "outcome": "record",
      "type": "Zeitschriftenaufsatz",
      "block_reason": null
    },
    {
      "field": "Psychologie",
      "source": "PLOS ONE",
      "url": "https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0287153",
      "seconds": 1.18,
      "outcome": "record",
      "type": "Zeitschriftenaufsatz",
      "block_reason": null
    },
    {
      "field": "Psychologie",
      "source": "Springer ASB",
      "url": "https://link.springer.com/article/10.1007/s10508-012-0062-0",
      "seconds": 0.97,
      "outcome": "record",
      "type": "Zeitschriftenaufsatz",
      "block_reason": null
    },
    {
      "field": "Psychologie",
      "source": "Frontiers",
      "url": "https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2021.618509/full",
      "seconds": 0.42,
      "outcome": "blocked",
      "type": null,
      "block_reason": "http-404"
    },
    {
      "field": "Informatik",
      "source": "arXiv",
      "url": "https://arxiv.org/abs/1706.03762",
      "seconds": 0.2,
      "outcome": "record",
      "type": "Preprint",
      "block_reason": null
    },
    {
      "field": "Informatik",
      "source": "PeerJ CS",
      "url": "https://peerj.com/articles/cs-1234/",
      "seconds": 0.18,
      "outcome": "blocked",
      "type": null,
      "block_reason": "http-403"
    },
    {
      "field": "Biologie",
      "source": "bioRxiv",
      "url": "https://www.biorxiv.org/content/10.1101/2023.01.01.522000v1",
      "seconds": 0.31,
      "outcome": "record",
      "type": "Preprint",
      "block_reason": null
    },
    {
      "field": "Naturwiss.",
      "source": "Sci Reports",
      "url": "https://www.nature.com/articles/s41598-023-38000-4",
      "seconds": 0.93,
      "outcome": "record",
      "type": "Zeitschriftenaufsatz",
      "block_reason": null
    },
    {
      "field": "Naturwiss.",
      "source": "MDPI",
      "url": "https://www.mdpi.com/2076-328X/13/2/167",
      "seconds": 0.23,
      "outcome": "blocked",
      "type": null,
      "block_reason": "http-403"
    },
    {
      "field": "Wirtschaft",
      "source": "EconStor",
      "url": "https://www.econstor.eu/handle/10419/234477",
      "seconds": 0.28,
      "outcome": "blocked",
      "type": null,
      "block_reason": "block-page"
    },
    {
      "field": "Wirtschaft",
      "source": "OECD",
      "url": "https://www.oecd.org/en/publications/oecd-employment-outlook-2023_08785bba-en.html",
      "seconds": 0.18,
      "outcome": "blocked",
      "type": null,
      "block_reason": "http-403"
    },
    {
      "field": "Sozialwiss.",
      "source": "SSOAR",
      "url": "https://www.ssoar.info/ssoar/handle/document/72281",
      "seconds": 0.2,
      "outcome": "blocked",
      "type": null,
      "block_reason": "block-page"
    },
    {
      "field": "Sozialwiss.",
      "source": "GESIS",
      "url": "https://search.gesis.org/research_data/ZA7500",
      "seconds": 0.18,
      "outcome": "blocked",
      "type": null,
      "block_reason": "http-403"
    },
    {
      "field": "Geisteswiss.",
      "source": "Persée",
      "url": "https://www.persee.fr/doc/rfsoc_0035-2969_1971_num_12_1_1875",
      "seconds": 3.46,
      "outcome": "blocked",
      "type": null,
      "block_reason": "http-404"
    },
    {
      "field": "Recht",
      "source": "EUR-Lex DSGVO",
      "url": "https://eur-lex.europa.eu/legal-content/DE/TXT/?uri=CELEX%3A32016R0679",
      "seconds": 0.87,
      "outcome": "record",
      "type": "Internetquelle",
      "block_reason": null
    },
    {
      "field": "Recht",
      "source": "RIS StGB",
      "url": "https://www.ris.bka.gv.at/GeltendeFassung.wxe?Abfrage=Bundesnormen&Gesetzesnummer=10002296",
      "seconds": 11.29,
      "outcome": "record",
      "type": "Internetquelle",
      "block_reason": null
    },
    {
      "field": "Statistik",
      "source": "Statistik AT",
      "url": "https://www.statistik.at/statistiken/bevoelkerung-und-soziales/bevoelkerung/bevoelkerungsstand",
      "seconds": 0.59,
      "outcome": "record",
      "type": "Internetquelle",
      "block_reason": null
    },
    {
      "field": "Statistik",
      "source": "Destatis",
      "url": "https://www.destatis.de/DE/Themen/Gesellschaft-Umwelt/Bevoelkerung/_inhalt.html",
      "seconds": 0.62,
      "outcome": "record",
      "type": "Internetquelle",
      "block_reason": null
    },
    {
      "field": "Repositorium",
      "source": "Zenodo",
      "url": "https://zenodo.org/records/10005281",
      "seconds": 0.54,
      "outcome": "record",
      "type": "Internetquelle",
      "block_reason": null
    },
    {
      "field": "Repositorium",
      "source": "OSF",
      "url": "https://osf.io/preprints/psyarxiv/n7sxr",
      "seconds": 0.29,
      "outcome": "blocked",
      "type": null,
      "block_reason": "block-page"
    },
    {
      "field": "Presse",
      "source": "Guardian",
      "url": "https://www.theguardian.com/science/2023/aug/01/superconductor-lk-99",
      "seconds": 0.23,
      "outcome": "blocked",
      "type": null,
      "block_reason": "http-404"
    },
    {
      "field": "Enzyklopädie",
      "source": "Wikipedia DE",
      "url": "https://de.wikipedia.org/wiki/Sexualtherapie",
      "seconds": 0.37,
      "outcome": "record",
      "type": "Internetquelle",
      "block_reason": null
    },
    {
      "field": "Verlag/Wall",
      "source": "ScienceDirect",
      "url": "https://www.sciencedirect.com/science/article/pii/S0747563223000626",
      "seconds": 0.21,
      "outcome": "blocked",
      "type": null,
      "block_reason": "http-403"
    },
    {
      "field": "Verlag/Wall",
      "source": "Wiley",
      "url": "https://onlinelibrary.wiley.com/doi/10.1111/jan.15234",
      "seconds": 0.22,
      "outcome": "blocked",
      "type": null,
      "block_reason": "http-403"
    }
  ],
  "limits": [
    "Read server-side without JavaScript. Details a page adds after loading are invisible.",
    "Three of the reported blocks were HTTP 404 from stale test addresses, not publisher policy.",
    "Timing is dominated by the target site, not the endpoint: two runs of the same address differed by up to 176 %.",
    "A first run measured 26 HTTP 403s in 0.13 s each. That was the measuring setup, not the pipeline: urllib sends 'Python-urllib/3.x' and Cloudflare's bot protection rejects it."
  ],
  "license": "CC-BY-4.0",
  "comparison": {
    "against": "Citoid (Wikimedia), which uses the Zotero translators",
    "why": "It solves the same problem, is publicly reachable, and is the de-facto standard for citation extraction. Without it, a hit rate is a number without a yardstick.",
    "urls": 18,
    "complete_citations": {
      "ours": 11,
      "citoid": 11
    },
    "any_record": {
      "ours": 11,
      "citoid": 14
    },
    "citations_built_from_a_block_page": {
      "ours": 0,
      "citoid": 2,
      "detail": "Citoid returned 'Making sure you're not a bot!' as the title for EconStor and SSOAR — a formatted reference to a bot wall."
    },
    "median_seconds": {
      "ours": 0.68,
      "citoid": 1.92
    },
    "slowest_seconds": {
      "ours": 10.0,
      "citoid": 29.2
    },
    "reached_only_by_citoid": [
      "MDPI",
      "PeerJ",
      "OECD"
    ],
    "reached_only_by_us": [
      "Zenodo",
      "Wikipedia"
    ],
    "measurement_note": "Citoid rate-limits anonymous callers. A first run hit HTTP 429 from the twelfth call on; those results measured our own impatience and were discarded.",
    "sampling": "18 works drawn at random from OpenAlex (is_oa, has_doi). The earlier 26-source list was assembled by hand and showed parity; random sampling showed 4 correct against Citoid's 7. A list the measurer assembles measures the list.",
    "wrong_records": {
      "ours": 0,
      "citoid": 3,
      "detail": "Citoid returned 'Redirect URL' as the title of a work, an empty title, and one title prefixed 'Article: '."
    },
    "accuracy_percent": {
      "ours": 100,
      "citoid": 79,
      "crossref": 100
    },
    "coverage_percent": {
      "ours": 61,
      "citoid": 61,
      "crossref": 33
    },
    "reference": "OpenAlex, independent of all three tools"
  },
  "thematic_check": {
    "topic": "Electronic health records and interoperability in Austria",
    "why": "A real research scenario mixes statutes, agency pages, standards and studies — the combination a term paper actually needs, and the one publisher-only sampling never covers.",
    "sources": 12,
    "outcome": "Legal sources were the gap: EU regulations and Austrian consolidated law declare neither author nor year. Both are now recognised as a source type of their own, with the issuing body as author and the year read out of the title of the act."
  },
  "test_suite": {
    "file": "tests/zitation.test.mjs",
    "cases": 16,
    "principle": "Every defect found in operation becomes a case here before it is fixed."
  }
}