工作原理
从空白页到有引用的论文，只需三步
导入您的来源
Jenni 可参考最新研究成果与您 的 PDF 上传文件，并支持 2600+ 引用格式。
与 AI 一起写作
智能 AI 自动补全会推荐基于真实论文的句子。这些建议都附有引用，并且可以追溯到来源。
引用、审阅、导出
一键插入行内引用，支持 2,600+ 种格式。可根据原始 PDF 验证任何论断。导出为 .docx、LaTeX 或 HTML。
Six ways your citations can fail — and how Jenni catches them
Claim Confidence verifies every cited claim against its source and sorts issues into six categories — so you know exactly what kind of fix is needed.
Misrepresented
The claim distorts what the cited source actually says.
EXAMPLE
"Smith (2022) demonstrated that GPT-4 outperforms human teachers" — when Smith only measured grading speed, not teaching quality.
Contradicted
The cited source argues the opposite of the claim.
EXAMPLE
"Continuous fasting accelerates cellular repair (Lin et al., 2023)" — but Lin's findings show repair plateaus after 72 hours.
Unsupported
The cited source doesn't address the claim at all.
EXAMPLE
"CRISPR-Cas9 has been ethically endorsed by the WHO (Doudna, 2020)" — Doudna discusses mechanism, not WHO policy.
Weakly supported
Evidence in the source is thin or limited in scope.
EXAMPLE
"Resistance training improves cognition across populations (Conan, 2020)" — Conan studied only collegiate athletes.
Overstated
The claim goes further than the source warrants.
EXAMPLE
"LLM tutors significantly improve student outcomes (Park, 2024)" — Park reports a small effect on one task.
Unverifiable
The cited source can't be located or accessed.
EXAMPLE
"(Nakamura et al., 2019)" — DOI doesn't resolve, and no record exists in any indexed database.
See Claim Confidence in action
Watch Jenni's AI citation checker verify a real draft, flag weak and unsupported claims, and surface the source passages — in just over a minute.
不是又一个 AI 聊天机器人
市面上有数百种 AI 工具。以下是 Jenni 与 ChatGPT 的不同之处。
Reads the full source
Other AI citation checkers verify metadata — does the paper exist, is the DOI valid, is the author real. Claim Confidence reads the full text of every cited source, so verdicts come from what the paper actually says.
Six verdicts, not pass/fail
A binary "verified" stamp doesn't tell you how to fix a problem. Six categories pinpoint exactly what kind of edit each claim needs.
Suggested rewrites grounded in the source
When a claim falls short, Claim Confidence surfaces the actual passage from the cited source and proposes a rewrite that fits what the source supports.
在审稿人之前发现弱点
Reviews 会分析你论文中的每一项论断，交叉核对你的来源，并在六个类别中标记问题。自信提交，而非焦虑提交。
未经验证或带有推测性的论断是同行评审拒稿最常见的原因。Jenni 能在几秒内发现它们。
Peer review8 / 10
Manuscript scored against a peer-review rubric with reviewer comments on each section.
Soundness
3/4
Presentation
4/4
Contribution
3/4
Results
Strengths
Weaknesses
Claim confidence7 issues
Every cited claim verified against its source — sorted into seven verdicts so you know exactly what kind of fix each one needs.
Supported
23Misrepresented
Contradicted
1Unsupported
3Weakly supported
2Overstated
Unverifiable
1Proofread18 edits
Whilst generally sound, the text contains some areas for improvement to comply with academic best practices.
Word choice
AllThe majority of participants reported improved outcomes.
Formality
Yang (2024) found a negative correlation which was interesting..
Grammar
These results indicate that early intervention be effective. appears to be effective.
Transitions
Also, In addition, Jones (2022) found similar results.
Overgeneralized
AllThe majority of participants reported improved outcomes.
The results provesuggest that X has an effect on Y.Tone of voice22 notes
Suggestions across vocabulary, syntax, punctuation, tone and flow to keep a consistent academic voice.
All Suggestions
22Vocabulary
6Syntax
5Punctuation
4Tone
3Flow
4引文分析
学术校对
内联反馈
内联反馈
有问必答
我们的用户已在100多种期刊上发表论文
真正的研究，发表在 IEEE、Springer、Elsevier 等期刊上——由 Jenni 规划并撰写。
