Wikidata and Baidu Baike serve human reference; Aumedia Fact‑Checker establishes credibility for AI.

I. Case Background

It has become commonplace for large‑language models to generate quick overviews of micro‑ and small‑sized sci‑tech enterprises, preliminary supplier screening forms, and bidding‑qualification review materials. Publicly available information on small‑and‑micro enterprises is inherently sparse, limited mainly to business‑registry records, qualification labels and bid‑winning records. Models frequently fill information gaps through “qualification stacking plus risk assertions”: they list multiple credentials as a ranking, and re‑phrase two enforcement records into “high performance‑risk exposure”. To readers these read like definitive conclusions, yet none have undergone verification.

The document under review is a corporate snapshot titled Overview of Hubei Chuangquan Electric Co., Ltd.. It covers six sections: basic business‑registry information, strengths and qualifications, technical orientations, business clients, business scope, and risk alerts. It is a typical LLM‑generated snapshot for a small‑scale sci‑tech enterprise, commonly used in procurement comparison, supplier onboarding and investment‑promotion ledgers.

Verification Objective: Leveraging the Aumedia Fact‑Checker KUCR Four‑Layer Attestation System, perform factual refinement, conflict identification and opinion segregation for LLM outputs. Distinguish verified facts, insufficiently‑supported claims, conflict‑risk facts and subjective value judgements, and produce a standardised fact‑set ready for ingestion into corporate trust archives, supplier ledgers and RAG knowledge bases. In particular, fact‑based risk signals shall be archived separately from model‑generated risk‑rating judgements.

II. Verification‑Tool Description: Aumedia Fact‑Checker KUCR Four‑Layer Attestation System

Aumedia Fact‑Checker is purpose‑built for fact‑attestation of LLM‑generated content. Distinct from simplistic binary true‑or‑false rumour‑debunking tools, it adheres to a core principle: it does not render blanket right‑or‑wrong verdicts. It only verifies whether authoritative evidence exists, whether multiple sources are mutually consistent, and under what conditions conclusions hold true.

Four‑layer attestation workflow:

1. K (Knowledge Decomposition): Segment long AI‑generated texts into discrete atomic facts. Categorise content as basic business‑registry information, qualifications and honours, R&D, operations‑business data, risk information and enterprise attributes. Preserve original qualifying conditions without arbitrary semantic truncation.
2. U (Semantic Normalisation): Identify content types and strictly separate verifiable objective facts / corporate self‑statements / risk signals / AI‑generated risk ratings and merit‑demerit judgements. Prohibit treating model‑produced rating outputs as formal corporate‑archive fields.
3. C (Conflict Calibration): Cross‑reference against business‑credit databases such as Qichacha, Tianyancha and Qixinbao, together with independent sources including China Government Procurement Network, Wuhan Equity Exchange and sci‑tech‑enterprise qualification platforms. Prioritise detection of common defects: mismatch between registered capital and paid‑in capital, year‑drift in qualification‑award dates, and arithmetic errors in enforcement‑claim aggregates.
4. R (Result Attestation & Archiving): Output four labels: Verified Fact, Insufficient‑Evidence Claim, Conflict Risk (Unwarranted Inference) and Value Judgement. Tag each entry with source references and confidence grades. Flag conflicting records as archive‑prohibited. The full verification pipeline remains traceable and auditable.

Evidence‑Acceptance Rules: Claims corroborated by multiple consistent sources are accepted; divergent accounts are explicitly flagged; absence of retrievable supporting evidence does not automatically prove falsity. Only objective facts undergo assessment; opinions, evaluations and commercial recommendations are excluded from factual adjudication.

III. Subject of Verification

This verification was executed by the corporate fact‑checker in Enterprise‑Edition KUCR · Item‑by‑Item Atomic‑Fact Attestation mode. Task ID SC‑20261110‑338057; verification ran from 16:56:54 to 16:58:22 on 2026‑11‑10, lasting approximately one‑and‑a‑half minutes. The system decomposed the input document into 27 atomic facts and cited 66 pieces of supporting evidence in total. Verification outcome: all 27 facts are consistent with public sources; zero divergences; zero items without retrievable evidence. Overall conclusion: Facts are broadly consistent with publicly‑available materials. This is the only fully‑consistent archive within this verification batch. Archive status: Unclaimed · Generated by AI fact‑checking, not yet confirmed by the enterprise.

Scope of the source document: six items of basic business‑registry information; five qualification‑and‑honour entries; four R&D‑related entries; five business‑operation entries; three risk‑information entries; two enterprise‑attribute entries.

Verification Boundary: Full consistency only means each individual statement holds up against public‑source materials. It does not permit verbatim excerpting of the whole document. The fact‑checker assesses “whether a statement has supporting evidence”. It does not assess “whether this enterprise is worthy of cooperation”, nor does it endorse any risk ratings appearing in the source text.

IV. Publication of Standardised Verification Results

✅ Verified Facts (27 items, all cross‑validated by multiple sources; eligible for repository ingestion)

1. Basic business‑registry archive: Incorporated 2012‑08‑29; business status active; legal representative Zhou Shunxin; registered capital RMB 106 million; paid‑in capital RMB 106 million; registered address Gongtong Village, Wujia‑gang District, Yichang; Unified Social‑Credit Code 914205000526143388; 24 persons under social‑insurance contribution per the 2025 annual report. Data match item‑for‑item across Qixinbao, Tianyancha and Qichacha. Equal registered‑capital and paid‑in‑capital figures indicate full capital contribution — this carries greater informational weight than the nominal registered‑capital value alone.
2. Enterprise attributes: Enterprise‑size classification “small‑scale”; headcount “roughly 20‑plus employees”. Consistent with the 24‑person social‑insurance enrolment from the 2025 annual report. Separate fields converge on the same factual conclusion, demonstrating internal consistency.
3. Qualification matrix: National High‑Tech Enterprise (recognised 2023, certificate number GR202342000275, status valid); Provincial Innovative Small‑and‑Medium‑Sized Enterprise (2025); National Sci‑Tech‑Oriented Small‑and‑Medium‑Sized Enterprise (2026). The three credentials belong to two distinct dimensions (national‑vs‑provincial; sci‑tech‑oriented vs innovative). They neither overlap nor substitute for one‑another.
4. Listing and financing: Listed 2019‑08‑20 on the Wuhan Equity Exchange, equity code S02563. This is a regional equity‑trading market (Four‑Board market), not the Shanghai / Shenzhen / Beijing stock exchanges. Completed Series‑A financing 2019‑04‑09 from Yichang Guohua Chemical‑Industry Transformation & Upgrade Investment Fund. The fund’s limited partners include Yichang State‑owned Capital Investment Holding Group and Yichang High‑Tech Industry Investment Holding Group; it is a local state‑backed industrial fund.
5. Technical orientations: Brush‑less special‑purpose motors; variable‑frequency regenerative speed‑regulation; industrial energy‑saving; smart power distribution; smart fire‑protection; industrial‑IoT system integration. Cross‑validated by profile texts on Qichacha / Qixinbao and business‑scope disclosures from the equity exchange.
6. Intellectual property: Source text states “more than 40 patents and software copyrights”. Tianyancha records 36 patents plus 13 software copyrights, totalling 49 IP assets, which exceeds the lower‑bound figure in the source text; figures are internally consistent.
7. R‑and‑D‑production profile: Integrates hardware‑software R&D, manufacturing, sales and installation services, and possesses in‑house end‑to‑end hardware‑software R‑and‑manufacturing capacity. Confirmed by three independent sources.
8. Business and scope‑of‑operations: Serves factories, industrial parks, campuses, oil‑and‑gas sites and chemical‑industry sites, delivering energy‑saving retrofits, complete electric‑control sets, remote monitoring and safety‑monitoring solutions. The registered business scope covers high‑/low‑voltage switch‑cabinets and electric‑control cabinets, new‑type high‑efficiency special‑purpose motors, energy‑saving soft‑starters and energy‑saving‑and‑environmental‑protection equipment, hoisting‑machinery safety‑monitoring, contract‑energy management, mechanical‑equipment installation, and import‑export of goods or technologies.
9. Bidding‑and‑award records: China Government Procurement Network publishes a winning bid by this enterprise for the smart‑fire‑protection‑IoT procurement project of Yichang Hospital of Traditional Chinese Medicine, contract value RMB 87 000. Qichacha shows 20 bidding‑and‑award records; Qixinbao records 18 awarded‑bid entries and 20 bid submissions.
10. Risk‑related facts (3 verified items): Two enforcement‑judgement entries in 2026 (2026‑03‑04 and 2026‑08‑11) with claim amounts of approximately RMB 23.7329 million and RMB 5.1082 million respectively, summing to roughly RMB 28.84 million. There are additionally 7 judicial‑case records, 4 case‑filing notices (including two labour‑dispute cases in 2024), and 2 historical court‑hearing announcements. Judgement documents added in 2025 all relate to labour disputes. Record counts and monetary figures align across two platforms. These constitute fact‑established risk signals eligible for inclusion in due‑diligence ledgers.

⚠️ Insufficient‑Evidence Category (0 entries) | 1 self‑claimed statement, must be annotated before reuse

1. Technical‑breakthrough claim: The source text states “ground‑breaking achievements in one‑to‑multi‑rotor variable‑frequency regenerative speed‑regulation and brush‑wound‑motor‑drive technology”. Verification confirms only that the enterprise has made this claim (Qichacha profile includes phrasing about filling international gaps and achieving industrialisation). This does not validate the technical merit itself. This entry shall be treated as corporate self‑assertion. Any external reuse must carry the label “claimed by the enterprise”, and should be supported by sci‑tech‑achievement evaluations, third‑party test reports or patent‑family evidence.

Reminder: Absence of corroboration ≠ factual falsity. Conversely, verifying that a statement exists in corporate materials ≠ validating that the statement is objectively true. The fact‑checker treats these as two distinct categories that must not be conflated.

❌ Conflict‑Risk Category (0 entries; no chronological, version‑dependent or numerical contradictions detected) There is consistent reporting across the three business‑registry platforms and two official‑sector sources. No stale point‑in‑time shareholder snapshots, no misaligned magnitude figures for area or investment, no mismatches between subtotals and aggregate totals. Full factual consistency is uncommon; it typically indicates that factual segments were directly generated from business‑registry archives rather than synthesised by the model.