SCOPE: SOURCE CUSTODY OBSERVABLE PUBLICATION EVIDENCE

A Structured Documentation Protocol for External Source Verification in Published Work

Version 1.1 · June 2026

Working paper released for feedback and collaboration · Released under Creative Commons open-access terms

The Source Custody Gap in Published Work

Organizations and practitioners are publishing at scale, and most publication integrity discussion still centers on whether citations are accurate, whether AI generated them, and what enforcement should follow when they are not. What receives far less attention is the practical question that arrives after a citation is challenged: what record exists of what the author verified, when the author verified it, and whether the source supported the claim at the time of use?

This question is no longer theoretical. Automated citation-verification pipelines are entering scholarly workflows, AI-detection tools flag submissions for suspected fabricated references, and platforms in some cases impose bans while ethics bodies assign blanket responsibility. When a citation is challenged, the author has no standardized evidentiary record to produce in rebuttal, because no publication system asks authors to document what they verified at the time of use and no mechanism exists to produce that documentation on challenge.

The problem is not lack of verification, because many authors verify carefully. The problem is that verification effort is invisible. An author who checked every citation against the primary record, ran status queries, and archived the source occupies the same position as an author who never opened a single reference. No framework records the difference, and no mechanism produces it. SCOPE exists to close that gap.

What SCOPE Is

SCOPE stands for Source Custody Observable Publication Evidence. It is a structured documentation protocol that records what an author verified about each cited source at the time of use, creating a private evidentiary record that can be produced when a citation is challenged or the work’s credibility is questioned.

SCOPE is human-triggered. The author creates a SCOPE record at the point of citation, when a source is first added to the manuscript’s reference list. The author decides which sources to document, at what depth, based on their own assessment of the claim’s stakes and the source’s stability. Nothing in the record is produced or shared unless the author chooses to produce it, or unless required by publisher policy, institutional process, or legal process.

SCOPE is not a full research-governance framework and does not govern how research is conducted. It documents what external sources were verified during the research process, and the distinction matters because SCOPE can apply to any publication workflow, not only workflows operating under HAIA methodology. Any author, researcher, journalist, or practitioner who cites external sources and wants a record of what they verified can use SCOPE. The field template in the appendix to this paper is deliberately designed without framework-specific terminology so it functions across any context.

Within the HAIA ecosystem, SCOPE is the external-source verification layer alongside HAIA-CARCS (Compliance Accountability Record and Case Study). CARCS documents how AI work was conducted (sessions, decisions, arbitration, platform outputs). SCOPE documents what external sources were verified at the time of use (citations, archives, status checks, claim-support assessments). They serve different functions and answer different questions. A complete governed record of published work that cites external sources includes both.

Outside the HAIA ecosystem, SCOPE applies wherever published work cites sources and the author wants a private record of what was checked. The adoption channels described in this specification (author prerogative, publisher requirement, insurance condition) operate independently of any governance framework.

How SCOPE Relates to CARCS

CARCS and SCOPE are complementary, not overlapping. The relationship is:

CARCS documents the session. How the work was produced, what AI platforms contributed, what the human arbiter decided, what dissent was preserved, and what comes next. CARCS answers: how was this work done?

SCOPE documents the sources. What external references were cited, whether each source was verified at the time of use, what the source said, whether it supported the claim, and where the preserved copy resides. SCOPE answers: what did the author verify about what they cited?

CARCS Section 10 (Raw Evidence Index) depends on external provenance data the AI synthesis platform cannot generate independently. This was identified as a limitation in CARCS v1.4. SCOPE fills that gap. When a CARCS record references external sources, the SCOPE ledger provides the verification evidence that CARCS Section 10 cannot produce on its own.

Standalone use: SCOPE can be produced without CARCS. An author writing a traditional research paper without AI collaboration may have no need for a CARCS session record but may want a SCOPE ledger for their citations. SCOPE operates independently.

Integrated use: When both CARCS and SCOPE are produced for the same project, SCOPE can attach to the CARCS record as a companion document. The CARCS record references the SCOPE ledger in Section 10. The SCOPE ledger references the CARCS record in its project metadata. Together they form a complete governed record: session governance (CARCS) and source governance (SCOPE).

Why Source Custody Matters

Source custody matters first at the evidentiary level. When an author is asked to show that a cited source was verified before publication, the citation itself is thin evidence. It shows what was cited but not whether the author opened the source, read the relevant passage, confirmed the content supported the claim, or checked whether the source had been retracted or corrected. A SCOPE record that documents the access date, the archived snapshot, the claim-support classification, and the status check result creates a stronger evidentiary category. It is not merely a reference list. It is verification made legible.

It matters just as much at the operational level. Citation verification during manuscript preparation is real work that takes real time. When that work is invisible, it cannot be distinguished from its absence. A SCOPE record created at the point of citation preserves the verification contemporaneously, which is stronger evidence than reconstruction from memory after a challenge arrives.

It also matters at the institutional level. A publisher that requires submitting authors to maintain a SCOPE ledger producible on request can reduce the publisher’s exposure to retraction disputes, misconduct allegations, and the administrative cost of adjudicating citation challenges without documentation. An insurance carrier can condition coverage on source custody records, creating a market-driven incentive for verification without mandating the process.

How SCOPE Works

SCOPE produces a structured record at the point of citation. Each entry documents one cited source. The author selects the documentation depth based on the stakes of the claim the citation supports.

Three Tiers

The author decides which tier applies to each citation. By default, no external party makes this decision.

Tier 1: Basic Source Check. Recommended for all references regardless of claim type. Documents that the source exists, the metadata matches, and the author confirmed access. Fields: source title, author or organization, DOI or URL or persistent identifier, date accessed, existence confirmed (the URL resolved to a live page with matching title and author). Estimated time: under one minute per citation after setup.

Tier 2: Claim Support Record. Recommended for references supporting factual, statistical, medical, policy, legal, or accusatory claims. Documents what the source is cited to support and whether the source actually supports that claim. Fields: all Tier 1 fields plus the exact claim the citation supports (one sentence), the page or section where the supporting evidence appears, a claim-support classification (direct support, indirect support, background only, or does not support), and the source type (primary research, secondary source, commentary, data, opinion, news reporting). Estimated time: two to three minutes per citation.

The claim-support field addresses a risk the UK Research Integrity Office (UKRIO) identifies directly: AI tools may provide real-looking references that do not support the claim being made. Existence verification alone is insufficient. The author must confirm that the source says what the citation attributes to it.

Tier 3: Full Source Custody Record. Recommended for high-risk claims where a challenge is reasonably anticipated or the consequences of error are significant. Documents the full verification chain including archived evidence, status checks, and the author’s reasoning. Fields: all Tier 1 and Tier 2 fields plus an archive link (Internet Archive or Perma.cc where institutional access exists), retraction or correction status at the time of access (documented through Crossmark, Retraction Watch, or publisher status check), the status check method, a later review date for time-sensitive sources, any later status change discovered after publication, and a one-sentence author diligence note explaining why the source was considered reliable at the time of use. Estimated time: three to five minutes per citation.

Time Cost

For a manuscript with 60 references, of which 20 support factual claims at Tier 2 and 5 carry high-risk claims at Tier 3, the estimated total documentation time is approximately 60 to 90 minutes. This is comparable to the time researchers spend formatting reference lists and often less than the time spent writing the manuscript itself.

Verification Gate

Before including any source in a manuscript’s reference list, the following steps apply at the selected tier.

At Tier 1: Open the source URL. Confirm the page loads with content. Confirm the title and author match. Record the access date.

At Tier 2: Complete all Tier 1 steps. Read the relevant passage. Record the specific claim and its location in the source. Classify the claim support.

At Tier 3: Complete all Tier 1 and Tier 2 steps. Check the retraction and correction status through Crossmark, Retraction Watch, or the publisher’s correction page. Create an archive snapshot through the Internet Archive or Perma.cc (where institutional access exists), or capture a local PDF. Write a one-sentence diligence note. If the source cannot be opened, does not match the metadata, does not support the claim, or has been retracted or corrected, exclude it from the reference list or document the reason for its inclusion. (Retracted or corrected sources that are the object of analysis rather than supporting evidence may still be cited if the manuscript explicitly discusses the retraction or correction.)

Evidentiary Strength Hierarchy

Not all documentation methods carry equal weight under challenge. The following hierarchy ranks SCOPE methods from strongest to weakest based on the independence and verifiability of the evidence they produce.

Internet Archive snapshot (strongest). Third-party timestamp. The Wayback Machine records the page content and assigns a permanent URL with a date that the author cannot alter. An adjudicator can independently verify the snapshot exists and when it was created.

Perma.cc link (strong, institution-gated). Third-party timestamp from Harvard Library Innovation Lab. Stronger than a local copy but access to the creation tool requires institutional affiliation at participating libraries.

Browser “Save as PDF” (moderate). Creates a local file with an operating system timestamp. The timestamp is locally generated and therefore weaker as independent evidence, but the full content is preserved and readable.

Screenshot with browser address bar (moderate-low). Shows the URL, title, and visible content at time of capture. Does not independently prove when the screenshot was taken and can be edited. Stronger than nothing. Weaker than any timestamped method.

Copy-paste into a text document (weakest). Preserves content only. No layout, no independent third-party timestamp, no visual proof of the source page. Acceptable as a last resort when other methods are unavailable.

For high-stakes work (Tier 3), the strongest practice combines an Internet Archive snapshot with a saved PDF. For routine work (Tier 1), a screenshot with the address bar visible is sufficient. The author should select the method that matches the stakes of the claim.

How to Create a SCOPE Record

A SCOPE record does not require specialized software, institutional subscriptions, or technical infrastructure. It requires, at minimum, a browser, a screenshot tool, and a folder. The implementation scales from a basic visual record that any author can create in minutes to a full-text custody archive suitable for high-stakes publications.

Basic Implementation (Tier 1)

The minimum viable SCOPE record is a collection of screenshots.

Step 1: Open each cited source in a browser. Navigate to the URL, DOI, or persistent identifier listed in the reference. The browser’s address bar must be visible in the screenshot.

Step 2: Take a screenshot of each source page. The screenshot must show the browser address bar (showing the URL), the page title, the author or organization name, and enough visible content to confirm the page is live and matches the citation metadata. Any system screenshot tool works: Snipping Tool on Windows, Screenshot on Mac, or a browser extension such as GoFullPage for long pages.

Step 3: Name each screenshot file. Use the citation’s author and year: Topaz_2026.png, Hsiao_Schneider_2021.png, COPE_Position_2023.png. Consistent naming makes the record navigable.

Step 4: Store the screenshots. Two options, choose one or both.

Option A: File folder. Create a folder named SCOPE followed by the project name and date using the format SCOPE_ProjectName_YYYY_MM_DD (for example, SCOPE_Fault_Based_Ethics_2026_06_02). Place all screenshots in the folder. This is the author’s private archive.

Option B: Compiled PDF. Combine all screenshots into a single PDF document (for example, six per page), with the source citation labeled beneath each screenshot. This produces a single portable file that can be stored, shared, or attached as a supplementary document if the author chooses to produce it.

The compiled PDF is what Appendix D of the companion SSRN working paper, Abstract ID 6872038, shows in practice. That appendix is a working example of a basic SCOPE record created using this method.

Time cost for basic implementation: Under one minute per source for the screenshot. For a manuscript with 20 to 25 references, the entire basic SCOPE record takes 20 to 30 minutes including file organization.

Standard Implementation (Tier 2)

The standard implementation adds claim-support documentation to the basic visual record.

Step 5: For each claim-bearing citation, annotate the screenshot or create a companion note. Identify the specific passage in the source that supports the claim the citation is intended to support. This can be done by highlighting the passage in the screenshot (using any image markup tool), by adding a text annotation to the screenshot, or by creating a separate text document that lists each citation with its exact claim, the page or section where the supporting evidence appears, and the claim-support classification (direct, indirect, background only, or does not support).

The companion note can be as simple as a spreadsheet or text file with four columns: citation, claim it supports, where in the source the evidence appears, and the claim-support classification.

Time cost for standard implementation: Two to three minutes per claim-bearing citation beyond the basic screenshot. For a manuscript with 20 claim-bearing citations, the standard SCOPE record adds 40 to 60 minutes.

Advanced Implementation (Tier 3)

The advanced implementation adds full-text preservation and status verification to the visual and claim-support record.

Step 6: Download or copy the full text of each cited source. For each source, save a local copy of the full content as it appeared at the time of access. Three methods, ranked by evidentiary strength.

Method A (strongest): Internet Archive snapshot. Go to web.archive.org/save/ and paste the source URL. The Internet Archive creates a timestamped copy with a permanent URL. The timestamp is third-party verified and cannot easily be recreated retroactively by the author. This method is free and requires no institutional affiliation.

Method B: Save as PDF. Use the browser’s “Print to PDF” or “Save as PDF” function to save the full page. This creates a local copy with a file creation timestamp from the operating system. The timestamp is locally generated and therefore weaker as independent evidence.

Method C: Copy and paste into a document. Copy the full text of the source page into a Word document or text file. Label it with the source citation and access date. This preserves the content but loses the visual layout and creates no independent third-party timestamp.

For the strongest record, use Method A (Internet Archive) for the third-party timestamp and Method B (PDF) for the local copy. Together they provide both external verification and a readable local archive.

Step 7: Run status checks. For each source, check whether it has been retracted, corrected, or flagged by a correction, retraction, expression of concern, or publisher notice.

Retraction Watch (retractionwatch.com): Search by author name or paper title. Free, usually no account required.

Crossmark: Look for the Crossmark button on the publisher’s page. Click it to see whether corrections, retractions, or expressions of concern have been issued. No account required.

Publisher correction page: Check the journal’s or publisher’s page for the specific article to see if any notices have been posted.

Record the result: “Retraction Watch: no entry found. Crossmark: no updates found. Checked [date].”

Step 8: Write the diligence note. One sentence per source explaining why the source was considered reliable at the time of use. Examples:

“Peer-reviewed article in The Lancet, accessed directly via DOI, claim verified against the abstract and results section.”

“Official COPE guidance page, accessed at the published URL, content matched the quoted language.”

“arXiv preprint, not peer-reviewed, labeled as preprint throughout the manuscript, accessed directly.”

Step 9: Compile the full record. The advanced SCOPE record for each source consists of: the screenshot (from Step 2), the full-text copy where legally permissible (from Step 6), the claim-support annotation (from Step 5), the status check result (from Step 7), and the diligence note (from Step 8). Store in the project folder alongside the basic screenshots. If compiling into a PDF, the full-text copies can be included as additional pages following each source’s screenshot.

Time cost for advanced implementation: Three to five minutes per citation including the standard implementation steps. For a manuscript with 5 to 10 high-risk citations at Tier 3, the advanced layer adds 15 to 50 minutes.

Total Time by Publication Type

The author decides. These estimates are guidelines, not requirements. A blog post with a single controversial claim may warrant Tier 3 for that one citation and nothing else. A book manuscript may warrant Tier 1 only if the author judges the risk as low. The tiers are tools, not mandates.

Version Control for the SCOPE Record

Manuscripts change across revisions. Citations are added, removed, and updated. The SCOPE record should track these changes without destroying evidence from earlier drafts.

Minimum practice: Use dated subfolders within the project SCOPE folder. Each major revision gets its own subfolder using the format SCOPE_ProjectName_YYYY_MM_DD. When a citation is removed from the manuscript during revision, keep the SCOPE entry in the record. Do not delete verification evidence for sources that were once part of the project. The removal itself is information.

Stronger practice: If the author uses Git or another version control system for manuscript management, commit the SCOPE folder alongside the manuscript at each revision. The commit history provides a timestamped, auditable trail of what was documented and when. This also prevents accidental loss of earlier SCOPE entries.

When a source changes status between revisions (for example, a source is retracted after the first draft but before final submission), update the SCOPE entry with the new status, the date the change was discovered, and any action taken (source removed, source retained with retraction noted, or claim revised). Do not overwrite the original entry. Add the update as a new field or a new line in the diligence note.

The SCOPE Record Is Private

The SCOPE ledger is not submitted with the manuscript. It is maintained by the author as part of their research governance practice. The author controls the record the way a researcher controls raw data and a journalist owns source files.

Production on challenge. The record is produced at the author’s discretion when a citation is challenged, unless an existing agreement, institutional process, or legal process requires production. Typical triggers include when a citation is challenged, when the work’s credibility is questioned, when a publisher or institution requests it, or when legal process compels it.

Right to decline. If the author chooses not to produce the record, that remains the author’s choice unless production is required by agreement or process. The absence of a produced record is discretionary unless a publisher, institution, or legal process requires production.

Confidentiality protections. If a challenge proceeds before a private institution (journal, publisher, university review board), the author can decide whether to share the ledger, what portions to share, and under what conditions. The author may request a non-disclosure agreement or request that the ledger remain confidential beyond the review board to protect intellectual property, methodology, or source relationships. A reviewing body can request the full ledger for all disputed citations. The author can decline, but declining after being asked is information the adjudicator may weigh.

Legal production. If allegations are brought before a legal entity, production may be compelled through legal process, subject to the same protections that govern any evidentiary production.

Protected content. Parts of the record may be unproducible regardless of the author’s willingness, because of HIPAA protections (where applicable), privacy regulations, anonymous data-collection protocols, or other legal constraints on the underlying source material. The diligence note can document that a source was verified under conditions that prevent full disclosure without specifying the protected content.

Right not to create. The author can choose not to produce a SCOPE record for any piece of work, for any reason, without that choice carrying automatic negative inference. The reason can be as simple as “this was an opinion editorial” or “this was a blog post” or “I did not expect this work to attract that level of scrutiny.” The decision belongs to the author.

Three Channels of Adoption

SCOPE adoption is voluntary for authors. Three independent channels create incentives for its use.

Author prerogative. The author decides which outputs warrant governed documentation based on their own assessment of stakes. A personal blog carries no expectation of a SCOPE record. A book manuscript approaching final publication warrants thorough documentation. The cost of maintaining the record is borne by the person who benefits from having it. The author can choose not to maintain a SCOPE ledger for any piece of work, for any reason.

Publisher requirement. Journals and platforms can require that submitting authors maintain a SCOPE ledger for any work they submit, producible on request if a citation is challenged. This is analogous to existing submission requirements for data availability statements, ethics declarations, and conflict-of-interest disclosures. The journal does not review the ledger at submission. It requires the author to attest that one exists and can be produced. A journal can calibrate: require attestation for research articles, recommend it for reviews, waive it for editorials. This requirement can reduce risk to the publishing entity by ensuring documentation exists before a dispute arises.

Insurance condition. Errors-and-omissions or directors-and-officers insurance policies covering publication-related claims can condition coverage on the maintenance of source custody records. If a citation dispute produces a claim and the policyholder maintained a SCOPE ledger, the claim could be covered. If the policyholder did not maintain one, coverage could be denied or limited, depending on policy language. The insurer does not mandate verification. The insurer prices the risk. This pattern is established in medical malpractice, legal malpractice, and financial services errors-and-omissions coverage, where record-keeping affects both insurability and claim outcomes.

What SCOPE Produces at Each Fault Level

SCOPE integrates with the five-level fault ladder proposed in the companion SSRN paper (Puglisi, 2026, SSRN Abstract ID 6872038).

Level 1 (Fabrication) and Level 2 (Failure to Verify). The inability to produce any SCOPE entry is consistent with the absence of verification, but is not conclusive proof of absence. An author who never checked cannot produce evidence of checking. The mechanism is asymmetric: only the author who maintained a record can produce one. A careless author cannot easily recreate contemporaneous third-party timestamps retroactively.

Level 3 (Negligent Verification). A SCOPE entry showing the author accessed the source and recorded the passage they relied on, but misinterpreted its meaning, is evidence of good-faith engagement. Correction is warranted. The record distinguishes misinterpretation from non-engagement.

Level 4 (Downstream Contamination). A SCOPE entry showing the source was not retracted or flagged at the time of access, documented through a Crossmark or Retraction Watch check, is evidence that the author verified and the source status changed after verification. The fault lies upstream.

Level 5 (Source Decay). A SCOPE entry with an archive link or local copy is evidence that the source existed and supported the claim. The medium failed. The author did not.

The SCOPE record does not immunize an author. It creates reviewable evidence of what the author checked, when the author checked it, and whether the author’s reliance was reasonable under the conditions available at the time.

Limitations and Current Status

SCOPE v1.1 has been tested in practice through one documented use. Appendix D of the companion SSRN working paper, Abstract ID 6872038, constitutes the first documented SCOPE record produced under this specification: 26 source verification screenshots captured on the day of submission, compiled into a five-page contact sheet, and published as part of the working paper. The specification has not yet been tested by external users, adopted by any publisher, or produced during a post-publication citation challenge. No publisher has adopted the tiered field template. No insurance carrier has conditioned coverage on source custody records.

The specification is derived from the evidence base documented in the companion SSRN paper, which draws on peer-reviewed retraction-persistence studies, reference-rot measurements, AI-fabricated citation audits, and established precedents for private production-on-challenge in law, medicine, journalism, and financial services.

The fault ladder referenced in this specification has not been adopted by any publisher, ethics body, or repository. It is a proposed framework, not an established standard.

SCOPE v1.1 is released for feedback, testing, and challenge. The validation roadmap in the companion paper proposes three stages: pilot testing with a single journal or preprint server, retrospective testing against known retraction cases, and integration testing with citation manager plugins.

Appendix: SCOPE Field Template

Tier 1: Basic Source Check

Tier 2: Claim Support Record

Tier 3: Full Source Custody Record

Sources

The following sources are cited in this specification and its companion SSRN paper. All URLs were documented as live during the June 2026 research session. Source verification screenshots for cited external sources appear in Appendix D of SSRN 6872038 (26 sources documented). Author works and supplementary sources are not included in Appendix D.

Primary Empirical Sources

Ansari, S. (2026). Compound deception in elite peer review: A failure mode taxonomy of 100 fabricated citations at NeurIPS 2025. arXiv preprint arXiv:2602.05930. https://arxiv.org/abs/2602.05930

Hsiao, T.-K., & Schneider, J. (2021). Continued use of retracted papers: Temporal trends in citations and (lack of) awareness of retractions shown in citation contexts in biomedicine. Quantitative Science Studies, 2(4), 1144–1169. https://doi.org/10.1162/qss_a_00155

Topaz, M., Roguin, N., Gupta, P., Zhang, Z., & Peltonen, L.-M. (2026). Fabricated citations: An audit across 2.5 million biomedical papers. The Lancet, 407(10541), 1779–1781. https://doi.org/10.1016/S0140-6736(26)00603-3

Xu, Z., et al. (2026). GhostCite: A large-scale analysis of citation validity in the age of large language models. arXiv preprint arXiv:2602.06718. https://arxiv.org/abs/2602.06718

Retraction, Reference Rot, and Citation Integrity

Avenell, A., Stewart, F., Grey, A., Gamble, G., & Bolland, M. (2019). An investigation into the impact and implications of published papers from retracted research: Systematic search of affected literature. BMJ Open, 9, e031909. https://doi.org/10.1136/bmjopen-2019-031909

Bakker, C., Boughton, S., Faggion, C. M., Fanelli, D., Kaiser, K., & Schneider, J. (2024). Reducing the residue of retractions in evidence synthesis: Ways to minimise inappropriate citation and use of retracted data. BMJ Evidence-Based Medicine, 29(2), 121–126. https://doi.org/10.1136/bmjebm-2022-111921

De Oliveira Andrade, R. (2022, December). Retracted scientific articles continue to be cited by other scientists as reliable sources. Pesquisa FAPESP, (322). https://revistapesquisa.fapesp.br/en/retracted-scientific-articles-continue-to-be-cited-by-other-scientists-as-reliable-sources/

Jones, S. M., Van de Sompel, H., Shankar, H., Klein, M., Tobin, R., & Grover, C. (2016). Scholarly context adrift: Three out of four URI references lead to changed content. PLOS ONE, 11(12), e0167475. https://doi.org/10.1371/journal.pone.0167475

Klein, M., Van de Sompel, H., Sanderson, R., Shankar, H., Balakireva, L., Zhou, K., & Tobin, R. (2014). Scholarly context not found: One in five articles suffers from reference rot. PLOS ONE, 9(12), e115253. https://doi.org/10.1371/journal.pone.0115253

Van der Vet, P. E., & Nijveen, H. (2016). Propagation of errors in citation networks: A study involving the entire citation network of a widely cited paper published in, and later retracted from, the journal Nature. Research Integrity and Peer Review, 1(3). https://doi.org/10.1186/s41073-016-0008-5

Policy, Governance, and Ethics Sources

Committee on Publication Ethics. (2023, updated 2025). Authorship and AI tools: COPE position statement. https://publicationethics.org/guidance/cope-position/authorship-and-ai-tools

Committee on Publication Ethics. (2026). Glossary. https://publicationethics.org/glossary

Hosseini, M., & Resnik, D. B. (2026). Hallucinated citations produced by generative artificial intelligence may constitute research misconduct when citations function as data in scholarly papers. Accountability in Research: Ethics, Integrity and Policy. https://doi.org/10.1080/08989621.2026.2645390

Parker, L., Boughton, S., Bero, L., & Byrne, J. A. (2024). Paper mill challenges: Past, present, and future. Journal of Clinical Epidemiology, 176, Article 111549. https://doi.org/10.1016/j.jclinepi.2024.111549

Sousa, S., Rowcliffe, N., Iorio, B., & Perchyk, T. (2026). Managing the risks of generative AI in academic publishing. Global Institute for Economy and Finance. https://www.giefoundation.net/responsible-ai-in-science/managing-the-risks-of-generative-ai-in-academic-publishing

UK Research Integrity Office. (n.d.). AI in research. https://ukrio.org/ukrio-resources/ai-in-research/

Proposals and Commentary

Glynn, A. (2025). Guarding against artificial intelligence-hallucinated citations: The case for full-text reference deposit. European Science Editing, 51. https://doi.org/10.3897/ese.2025.e153973

Dulin, K., & Ziegler, A. (2017). Scaling up Perma.cc: Ensuring the integrity of the digital scholarly record. D-Lib Magazine, 23(5/6). https://www.dlib.org/dlib/may17/dulin/05dulin.html

Lin, J., Shan, R., Zhu, J., Xi, Y., Yu, Y., & Zhang, W. (2025). Stop DDoS attacking the research community with AI-generated survey papers. arXiv preprint arXiv:2510.09686. https://arxiv.org/abs/2510.09686

Tramèr, F. (2025, August 3). Trends in LLM-generated citations on arXiv. SPY Lab Blog, ETH Zurich. https://spylab.ai/blog/hallucinations/

Zittrain, J., Albert, K., & Lessig, L. (2014). Perma: Scoping and addressing the problem of link and reference rot in legal citations. Harvard Law Review Forum, 127, 176–199. https://harvardlawreview.org/forum/vol-127/perma-scoping-and-addressing-the-problem-of-link-and-reference-rot-in-legal-citations/

Legal Sources

Federal Rules of Civil Procedure, Rule 11(b). https://www.law.cornell.edu/rules/frcp/rule_11

Mata v. Avianca, Inc., 678 F. Supp. 3d 443 (S.D.N.Y. 2023). https://law.justia.com/cases/federal/district-courts/new-york/nysdce/1:2022cv01461/575368/54/

Jibril, A. B. (2025). False authorship: An explorative case study. Research Integrity and Peer Review. https://pmc.ncbi.nlm.nih.gov/articles/PMC12107892/

News and Secondary Coverage

Chawla, D. S. (2026, May 19). Researchers who use hallucinated references to face arXiv ban. Nature. https://www.nature.com/articles/d41586-026-01595-5

Naddaf, M. (2026). Surge in fake citations uncovered by audit of 2.5 million biomedical-science papers. Nature. https://www.nature.com/articles/d41586-026-00748-w

Research Information. (2026, May). arXiv imposes one-year ban for unchecked AI-generated content. https://www.researchinformation.info/news/arxiv-imposes-one-year-ban-for-unchecked-ai-generated-content/

The Scientist. (2026, May). One in 277 biomedical papers carry fake references. https://www.the-scientist.com/one-in-277-biomedical-papers-carry-fake-references-74480

Inside Higher Ed. (2026, May 22). Ban on authors who submit AI content “welcome but unenforceable.” https://www.insidehighered.com/news/faculty/books-publishing/2026/05/22/ban-authors-who-submit-ai-content-welcome-unenforceable

TechCrunch. (2026, May 16). Research repository ArXiv will ban authors for a year if they let AI do all the work. https://techcrunch.com/2026/05/16/research-repository-arxiv-will-ban-authors-for-a-year-if-they-let-ai-do-all-the-work/

Author’s Published Works

Puglisi, B. C. (2025). Governing AI: When capability exceeds control. ISBN 9798349677687.

Puglisi, B. C. (2026). Fault-based publication ethics: The case for source custody in an era of AI citation contamination. SSRN Working Paper. https://papers.ssrn.com/abstract=6872038

Puglisi, B. C. (2026). The AI risk economy: Insurance governance as the missing enforcement layer for responsible AI. SSRN Working Paper. https://papers.ssrn.com/abstract=6823580

Puglisi, B. C. (2026). The human enhancement quotient: Measuring cognitive amplification through AI collaboration. SSRN Working Paper. https://papers.ssrn.com/abstract=6583419

Puglisi, B. C. (2026). AI provider plurality and the Verified AI Inference Standards Act (VAISA): Legislative framework for the 119th Congress. SSRN Working Paper. https://papers.ssrn.com/abstract=6195238

Puglisi, B. C. (2026). HAIA-CARCS: Compliance accountability record and case study (v1.4). basilpuglisi.com.

Puglisi, B. C. (2026). HAIA: Human artificial intelligence assistant ecosystem. GitHub (Creative Commons). https://github.com/basilpuglisi/HAIA

SCOPE v1.1 · Basil C. Puglisi, MPA · basilpuglisi.com · June 2026

#AIassisted using HAIA Ecosystem

Open-source governance work: github.com/basilpuglisi/HAIA

FAQ

What is HAIA-SCOPE?

HAIA-SCOPE (Source Custody Observable Publication Evidence) is a structured documentation protocol that records what an author verified about each cited source at the time of use. It creates a private evidentiary record that can be produced when a citation is challenged, comparable to attorney work product or physician clinical notes.

How does SCOPE differ from HAIA-CARCS?

CARCS documents how AI-assisted work was conducted: sessions, decisions, arbitration, and platform outputs. SCOPE documents what external sources were verified at the time of use: citations, archives, status checks, and claim-support assessments. CARCS governs the session. SCOPE governs the sources. They can operate independently or together.

How do I create a basic SCOPE record?

Open each cited source in a browser with the address bar visible. Take a screenshot of each page. Name each file by author and year. Store screenshots in a folder or compile into a single PDF with six per page. The entire basic record for a 25-reference manuscript takes 20 to 30 minutes.

Is the SCOPE record submitted with the manuscript?

No. The SCOPE record is private. The author maintains it and produces it at their discretion when a citation is challenged. The author can decline to produce it, can require a non-disclosure agreement, and can choose not to create one for any piece of work without automatic negative inference.

What are the three tiers of a SCOPE record?

Tier 1 is a basic source check: title, author, URL, access date, existence confirmed. Tier 2 adds claim-support documentation: what the citation supports and where the evidence appears in the source. Tier 3 adds full custody: archive snapshot, retraction status check, and a diligence note.

What is the evidentiary strength hierarchy?

Internet Archive snapshots are strongest because they carry a third-party timestamp the author cannot alter. Perma.cc links are strong but institution-gated. Saved PDFs carry a local timestamp that is weaker as independent evidence. Screenshots show the URL but carry no independent timestamp. Copy-paste text notes are weakest.

Basil C. Puglisi, MPA | Human-AI Collaboration Strategist | basilpuglisi.com

Published works: Governing AI: When Capability Exceeds Control (ISBN 9798349677687) | Digital Factics X (Kirkus reviewed)

Listen to The Other AI: Audio Briefings on Augmented Intelligence and AI Governance

Spotify | Apple Podcasts | Amazon Music | YouTube

#AIassisted using HAIA Ecosystem