Thesis Flagged as AI at a Chinese University: Real Procedures, Evidence Checklist, and How to Appeal

China's Degree Law gives degree applicants an academic re-review of an academic evaluation conclusion (Art. 40, decision within 30 days) and a degree re-review (Art. 41) once a degree is refused or revoked. How a school handles an AI-rate result, who hears it, and the deadline are set by the school's own notice, so you should find your school's appeal clause and deadline.

If your graduation thesis is flagged for exceeding an institutional Artificial Intelligence Generated Content (AIGC) screening threshold, you should recognize that while there is no single national body that hears AI-rate disputes, there are legal re-review routes. Investigative reporting by official news organizations (including Xinhua News Agency, China Media Group / China National Radio, Guangming Online, and Beijing Daily) alongside published university guidelines demonstrates that Chinese institutions treat automated percentages as preliminary screening filters or warning flags. Regarding whether a high AI rate equates to academic misconduct, national legal frameworks under the PRC Degree Law and Ministry of Education regulations define academic misconduct through specific acts (such as plagiarism, data fabrication, or ghostwriting) rather than an algorithmic similarity score. However, at the institutional level, exceeding a threshold directly governs whether a candidate may enter defense or must undergo revisions. Case management is administered internally: students submit formal objections and drafting documentation within specified institutional deadlines, followed by substantive evaluations conducted by faculty supervisors and college-level academic committees.

When responding to an automated warning, degree candidates must keep two principles in mind:

1. Submitting scores from different tools does not prove innocence: Chinese universities evaluate theses based strictly on the official platform designated by the university or secondary college (such as CNKI, VIP, or Wanfang). Presenting a low score from an unapproved commercial website carries no official weight with academic administrators;
2. Preserving draft history does not guarantee an appeal victory: Documented draft files provide essential evidence for faculty evaluation, but the final outcome depends on an academic committee's substantive assessment of your independent scholarship.

The valid path forward requires stepping away from arguments about software scores: immediately check your institution's thesis notice for its formal appeal clause and submission deadline, request a human substantive review from your faculty supervisor and secondary college, and compile a 6-part process evidence packet covering your topic approval, incremental drafts, version history, literature notes and raw data, official report export, and research timeline.

Understanding how Chinese universities address AI detection requires examining journalistic investigations and official directives issued by Academic Affairs departments. It must be noted at the outset: the documented examples reflect institutional practices that vary by university and academic year.

1. Documented Challenges in Media Investigations and Appeal Precedent

In-depth reporting by mainstream news organizations highlights several operational realities across Chinese universities:

• Initial Implementation Gaps and Student Challenges: An investigative report by The Paper (澎湃新闻, July 2024, reprinted by Guangming Online) documented that during initial rollouts, some institutions introduced screening tools without dedicated appeal channels, leaving students facing situations where "they had no appeal route and could only revise according to the machine's operational logic." Subsequent commentary by Guangming Online (Thesis 'AI Rate Lowering' Has Become a Business: AI Detection Must Not Degenerate into Formalism, June 2026) emphasized that technical screening must not degenerate into procedural formalism that penalizes rigorous academic writing.
• Algorithmic False Positives and a Documented Successful Appeal: Reporting originating from CCTV News (Will Graduation Theses Be Screened for AIGC Rates: Are Detection Standards Scientifically Sound?, May 22, 2026, reprinted by Beijing Daily; also syndicated by Xinhua News Agency WeChat on May 23) stressed that statistical models produce false positives on formal, rigorous academic prose, while warning that academic reviewers must also be alert to AI-fabricated research data. Crucially, an investigation by Guangming Daily (reprinted by China National Radio on May 27, 2026, Checking AI Rates with AI: Is It Truly Reliable?) documented an institutional appeal: a student, Zhang Li (pseudonym), received a third-party thesis blind review result with a score of 0 and an AI rate of 95%, whereas the school's platform had tested at 27%. She immediately filed an appeal, submitting her thesis proposal, original outline, and revision comments across different versions. After submitting supplementary materials several times and awaiting expert panel evaluation, the AI rate was ultimately revised to 32%, narrowly passing at the final deadline.
• Commercial Exploitation and Regulatory Boundaries: An investigation by Legal Daily (reprinted by China National Radio in July 2025) exposed deceptive commercial "AI rate lowering" services that exploit student anxiety, leading to extortion and paper leakage risks. Whether editing or rewriting aids are permitted depends on each institution's specific regulations and course requirements; candidates should consult institutional guidelines rather than relying on commercial promises of guaranteed passage.

1. Operational Procedures in Documented Institutional Notices

University directives frame automated screening figures as administrative benchmarks, and some institutional notices explicitly set out specific appeal and review procedures:

• Sichuan University: Directives from the Academic Affairs Office on academic misconduct detection for the 2025 graduating cohort (April 9, 2025) set distinct guidelines for humanities (in principle not exceeding 20%) and STEM/medicine (in principle not exceeding 15%). The notice explicitly specified that "secondary colleges may independently select their AIGC detection tools... recommending students within the same college use the same system," and instructed colleges to notify supervisors and students to handle the situation.
• Tianjin University of Science and Technology: The Academic Affairs Office notice on undergraduate thesis screening (April 9, 2024) specified that AIGC detection results must in principle remain under 40% (strictly < 40%). Crucially, the notice explicitly contains an appeal clause: "Students who have objections to the similarity check or AIGC detection results may file an appeal with their college within the prescribed timeframe, and the college will review and handle it in accordance with relevant procedures," with determinations made by the college academic committee.
• Aba Teachers University: The Academic Affairs Office notice on regulating AI tool use in undergraduate theses (March 21, 2025) set a clear cutoff for the 2025 cohort: "An AIGC detection rate $\le 35\%$ is deemed a pass"; theses exceeding 35% are ordered to be revised within a set timeframe for re-testing. The notice explicitly provides a formal appeal route: "Students who have objections to the AIGC detection results may file an appeal within the prescribed timeframe for procedural review." Furthermore, the directive instructs thesis supervisors to "periodically inspect original working files of phased research deliverables" and "require students to provide human revision records when necessary."
• Hebei Engineering University: Reporting by Beijing Daily (July 19, 2026) noted that institutions including Hebei Engineering University and Nanjing University of Aeronautics and Astronautics adopted a 40% threshold for undergraduate theses.

Across these institutions, handling an AI flag follows an internal administrative cycle: automated screening flag → checking institutional notice for appeal clauses and deadlines → submitting process evidence and written explanation → supervisor and college academic committee substantive review.

Students flagged by screening software often fall into the trap of trying to "prove the algorithm wrong." Understanding the distinction between arguing over software numbers and requesting an authentic human review is essential:

Two Paths When Responding to an AI Warning:

• Ineffective Path (Arguing Over Percentages): Purchasing third-party checks to show a lower score, then arguing with faculty that the university software is inaccurate. Academic administrations evaluate only official institutional systems; debating probabilistic algorithms stalls communication and risks missing formal appeal deadlines.
• Effective Path (Requesting Human Substantive Review): Identifying the school's appeal deadline, notifying the faculty supervisor with the complete official report, submitting a 6-part process evidence packet with a chronological timeline, and requesting an evaluation by the college expert committee based on actual scholarship.

1. Why Arguing Over Percentages Fails
2. Third-Party Reports Carry No Institutional Standing: Universities and secondary colleges contract with specific licensed platforms (such as CNKI, VIP, or Wanfang). Presenting an external commercial report showing 0% or 4% carries no institutional standing, because screening baselines and text corpora differ completely across vendors.
3. Algorithmic Debates Lead to Dead Ends: Detection tools output probabilistic estimates based on text predictability. Arguing to faculty that the software's probability metric is flawed rarely succeeds, because professors are subject-matter specialists, not detection algorithm engineers.

1. What Requesting a Human Substantive Review Means
2. Focusing on Actual Scholarship: A human substantive review redirects the inquiry from "how closely does the prose match an algorithmic pattern?" to "did the candidate personally conduct the investigation and write the thesis?"
3. Exercising Faculty Academic Judgment: Departmental committees and supervisors possess academic authority to verify independent scholarship. By reviewing raw research datasets, incremental drafts, and oral explanations, professors evaluate whether the research represents the student's authentic intellectual work.

When requesting a substantive review from your supervisor or academic committee, verbal assertions of independent writing are insufficient. Candidates must compile a comprehensive, verifiable evidence packet.

[!IMPORTANT] A realistic caution is essential: preserving and submitting these materials does not offer an absolute guarantee of an overturned flag. Their genuine function is to provide faculty reviewers with an unbroken, verifiable factual record, enabling them to evaluate independent authorship objectively.

A structured process evidence packet includes the following 6 categories:

1. Assignment Brief, Approved Proposal, and Topic Registration
2. Document Content: Official project task sheets, thesis proposals, and topic approval forms signed by your supervisor at the beginning of the academic year.
3. Core Value: Demonstrates that the research trajectory, theoretical framework, and milestones were established months before writing, precluding assumptions of last-minute generative AI ghostwriting.

1. Dated Incremental Drafts
2. Document Content: Preserved files representing distinct phases of work: initial topic outlines, chapter drafts, mid-term inspection submissions, and subsequent revisions.
3. Core Value: Sequential multi-stage files illustrate the developmental progression from outline to complete study. While operating system timestamps provide helpful supporting context, they can be altered; the primary weight rests on the coherent intellectual evolution visible across successive drafts.

1. Software Version History and Tracked Changes
2. Document Content: Built-in revision logs from Microsoft Word, WPS Office, or cloud collaboration platforms. For practical instructions on configuring version archives, see preserving version history in WPS and Tencent Docs.
3. Core Value: Authentic human scholarship involves extensive restructuring, sentence rewriting, phrasing adjustments, and typo corrections. Detailed software revision logs and comments demonstrate deliberate prose drafting rather than bulk external pasting.

1. Literature Notes, Annotations, and Raw Research Data
2. Document Content: Humanities and Social Sciences: Detailed reading notes, annotated academic PDFs, and citation management records (e.g., Zotero, EndNote) reflecting historical collection dates. STEM and Medicine: Laboratory notebooks with dated entries, instrument output files, raw survey spreadsheets, interview audio transcripts, or statistical code scripts.
3. Core Value: Raw empirical records and personalized source annotations demonstrate authentic scholarly work. Mainstream news reports have noted that AI can also generate synthetic data, making verifiable raw logs with real experimental parameters and methodology notes particularly vital. This reflects directives such as Aba Teachers University's policy empowering advisors to inspect phased working files and revision records.

1. Complete Official Detection Report Export
2. Document Content: The complete, unabridged PDF report downloaded directly from the screening portal designated by the university or college.
3. Core Value: Summary pages only display aggregate percentages. The full report identifies every highlighted sentence, allowing you and your supervisor to determine whether flags stem from specialized academic jargon, standard citations, or repetitive phrasing patterns.

1. Chronological Research Timeline Statement
2. Document Content: A concise, 1- to 2-page objective memorandum documenting your research and writing milestones: Date: Topic proposal submitted and approved; Date: Primary literature review assembled from library and electronic databases; Date: Experimental measurements taken or empirical survey administered; Date: Working draft shared with supervisor, noting initial guidance; Date: Chapter revisions executed based on supervisor comments.
3. Core Value: Connects separate pieces of evidence into a clear chronology, helping faculty reviewers follow how the research evolved stage by stage without having to piece together disorganized files.

If your thesis triggers an institutional AI detection warning, work through the following 5 steps systematically:

• [ ] Step 1: Preserve original drafts and system reports without frantic overwriting Do not panic or seek commercial rewriting tools to scramble text. Permissibility of editing aids depends on your institution and assignment guidelines. Avoid repeatedly saving over your master file, which risks overwriting version timestamps. Immediately back up your submission, drafts, and the official report by compressing them into a secure archive (such as a ZIP file) on an external drive.
• [ ] Step 2: Check institutional notices for appeal clauses and deadlines Search your university Academic Affairs portal and college notice board for the current-year thesis detection directive (such as notices issued by Tianjin Tech or Aba Teachers University). Carefully identify the specific clause governing detection disputes (异议申诉), the responsible committee, and the prescribed submission deadline.
• [ ] Step 3: Contact your faculty supervisor proactively Your supervisor is the primary academic mentor responsible for your graduation project. Schedule an appointment, share the complete detection report, state clearly that the work was conducted independently, and request guidance on initiating the institutional review or appeal procedure.
• [ ] Step 4: Assemble the 6-part evidence packet and draft the timeline Organize your task sheets, incremental drafts, version history, literature notes, and empirical data using the checklist above. Attach a clear, factual chronological timeline statement.
• [ ] Step 5: Submit the formal appeal and prepare for oral defense Submit your appeal dossier to the secondary college academic committee or teaching office within the prescribed deadline. In human reviews, the definitive proof of authorship is your intimate command of the manuscript. Review your hypotheses, methodology, and data analysis so that you can address academic inquiries with precision and confidence.

The journalistic investigations, institutional notices, and administrative guidelines referenced in this article include the following records (Note: certain institutional directives may be archived within university networks or historical bulletin boards; readers can verify records via university portals using the notice titles):

1. The Paper / Guangming Online (澎湃新闻 / 光明网): "自己写的毕业论文，被确诊AI" (My Personally Written Graduation Thesis Was 'Diagnosed' as AI). Investigative report by The Paper (Published July 22, 2024; reprinted by Guangming Online), documenting initial implementation challenges where students lacked formal appeal channels and faced pressure to rewrite text to satisfy algorithms. URL: `https://m.gmw.cn/2024-07/22/content_1303798841.htm`
2. Guangming Online (光明网): "论文‘降AI率’已成生意，AI检测不能沦为形式主义" (Thesis 'AI Rate Lowering' Has Become a Business: AI Detection Must Not Degenerate into Formalism). Commentator article (Published June 11, 2026) analyzing the commercialization of AI lowering services and procedural formalism. URL: `https://guancha.gmw.cn/2026-06/11/content_38823723.htm`
3. CCTV News / Beijing Daily (央视新闻 / 北京日报): "毕业论文将检测AIGC率，检测标准科学吗？" (Will Graduation Theses Be Screened for AIGC Rates: Are Detection Standards Scientifically Sound?). Published May 22, 2026 by Beijing Daily (originating from CCTV News WeChat; also syndicated by Xinhua News Agency WeChat on May 23), examining algorithmic false positives on academic prose and urging vigilance against AI-fabricated research data. URL: `https://news.bjd.com.cn/2026/05/22/11759420.shtml`
4. Guangming Daily / China National Radio (光明日报 / 央广网): "用AI查AI率，真的靠谱吗" (Checking AI Rates with AI: Is It Truly Reliable?). Investigative report by Guangming Daily (Published May 27, 2026; reprinted by CNR), documenting an appeal where a student with a 95% third-party blind review AI rate (versus 27% on the university platform) submitted her proposal, outline, and revision comments to have the rate corrected to 32%. URL: `https://news.cnr.cn/rebang/20260527/t20260527_527637099.shtml`
5. Legal Daily / China National Radio (法治日报 / 央广网): "代降论文AI率服务乱象调查：学生遭遇欺诈还被威胁" (Investigation into Deceptive 'AI Rate Lowering' Services: Students Defrauded and Threatened). Investigative report by Legal Daily (Published July 14, 2025; reprinted by CNR) on commercial intermediary fraud and extortion. URL: `https://news.cnr.cn/native/gd/20250714/t20250714_527260647.shtml`
6. Beijing Daily (北京日报客户端): "论文AI检测，正在‘逼疯’这届大学生" (Thesis AI Detection Is Pushing College Students to the Brink). Published July 19, 2026 at 21:05. Documenting Sichuan University's 20% and 15% discipline benchmarks, Nanjing Tech's departmental delegation, and 40% thresholds at Hebei Engineering University and Nanjing University of Aeronautics and Astronautics. URL: `https://xinwen.bjd.com.cn/content/s6a5ccb7de4b03fa51a81b89b.html`
7. Academic Affairs Office, Tianjin University of Science and Technology (天津科技大学教务处): Official Notice on Undergraduate Thesis Similarity and AIGC Detection (Published April 9, 2024; establishing a threshold strictly under 40% and providing formal appeal procedures through secondary college academic committees). URL: `https://jw.tust.edu.cn/ggtz/65b256ce743a41fa82c3533e16009807.html`
8. Academic Affairs Office, Aba Teachers University (阿坝师范学院教务处): Notice on Regulating the Use of AI Tools in Undergraduate Theses (Published March 21, 2025; setting a $\le 35\%$ pass standard for the 2025 cohort, establishing formal appeal and re-test procedures, and directing supervisors to inspect phased working files and revision records). URL: `https://jw.abtu.edu.cn/info/1131/4590.htm`
9. Academic Affairs Office, Sichuan University (四川大学教务处): Notice on Conducting Academic Misconduct Detection for the 2025 Graduating Cohort Undergraduate Theses (Published April 9, 2025; specifying guidelines of $\le 20\%$ for humanities and $\le 15\%$ for STEM/medicine, permitting secondary colleges to select detection tools). URL: `https://jwc.scu.edu.cn/info/1069/9801.htm`
10. National People's Congress (全国人民代表大会): Degree Law of the People's Republic of China (《中华人民共和国学位法》). URL: `http://www.npc.gov.cn/npc/c2/c30834/202404/t20240426_436840.html`