免费 AI 引用生成器 一个生成器，六套格式规则
此 AI 引用生成器可按 APA 7、MLA 9、Chicago 18、Harvard、IEEE 和 Vancouver 为 ChatGPT、Claude、Gemini、Copilot 或其他 AI 工具生成规范引用。选择工具，输入版本、日期和提示词，即可返回参考文献条目及所有文内引用形式，依据各格式手册对生成式 AI 的既有规定。十个模型家族已预设其详细信息，其他模型只需多输入几个字符。
APA 现在会引用对话，因此它需要聊天本身的标题和该聊天的分享链接。模型名称在产品的模型选择器中。提示词是 MLA、Chicago 和 Harvard 构建条目的基础。
84,000+
国际研究人员
平均评分
深受在以下期刊发表论文的研究人员信赖
三步引用 AI 对话
记录模型的准确版本
查看 AI 工具中的模型选择器或设置面板，记下系统提供给你的版本字符串。这些系统通常按周更新，因此版本号能让读者定位这次交互，正如版次能定位一本书。
记录日期和提问内容
条目的形式由两个细节决定, 文本生成日期和你的提示词措辞。MLA 和 Chicago 以提示词为引用核心，六种格式都会在条目中的某处记录日期。
选择一种格式并获取条目
APA、MLA、Chicago、Harvard、IEEE 和 Vancouver 均可用。每一种都会保留其斜体格式，并附带文内引用形式，以及该格式所要求的披露或注释措辞。
此 AI 引用生成器的不同之处
格式机构实际发布的指南
APA、MLA 和 Chicago 都已发布关于引用生成式 AI 的正式指南，而此生成器严格遵循这三者, APA 的公司作为作者模式, MLA 的提示词作为标题结构, Chicago 的仅注释规则。
十个模型家族即选即用
ChatGPT、Claude、Gemini、Meta AI、Copilot、Grok、DeepSeek、Le Chat、Perplexity 和 Qwen 已预填公司名称和网址，因此只需输入版本、日期和提示词。
保留各格式自身的结构
Chicago 不含参考文献条目，MLA 不含作者，Harvard 要求附录，IEEE 要求披露行。输出会按格式改变结构，而不是用同一模板套用六套规则。
可直接粘贴的格式
复制会同时带上富文本版本和纯文本版本，因此模型的斜体标题会在 Microsoft Word 或 Google Docs 中正确到位。
APA、MLA 和 Chicago 实际如何处理 AI 来源
在生成式 AI 工具进入学生与研究者之后的第一年里，还没有任何风格手册规定如何引用它，因此写作者只能临时发挥。随后有三份手册给出了各自的立场，但这些立场并不一致。APA 将输出归到模型背后的公司，开发者占据作者位置，模型名称以斜体作为标题，版本信息放在括号中。MLA 则完全不把 AI 系统称为作者，因此其条目围绕提示词构建，并把工具命名为提示词出现所在的容器。Chicago 将整个引用归入个人通信：脚注或文内提及承载引文，参考文献列表保持为空，这种拆分方式与如何在 APA、MLA 和 Chicago 中引用 ChatGPT所走的步骤一致，逐步讲解。
这三种模式中的每一种都依赖于两个写作者很容易忽略的细节，具体是哪一个版本回答的，以及日期是什么。今天运行一次提示词，下一次模型更新之后再用同一个提示词运行，可能会返回不同的文本，因此版本字段的权重就像一本书的版本信息那样，往往是书籍编辑才需要考虑的。也正因为如此，指导会把读者引导到工具的通用网页地址，而不是指向某一次对话的链接，另外三份手册都指向的后备方案，是在附录中保留完整的转录记录。
有一点学术出版已经停止争论了，没有任何 AI 系统被列为作者。ICMJE 和 COPE 都认为，署名需要对工作的责任承担，而这是任何模型都无法承担的，出版方包括 Nature 和 Science 也适用同样的规则。接下来要做的是两项独立义务，而不是一项义务：在文本中任何出现 AI 输出的地方进行引用，并在期刊或院系要求作出该声明时披露 AI 协助，通常写在方法部分或致谢部分。
与引文并列的还有一种相关做法。模型推荐的任何来源, 在进入参考文献表之前都需要先在真实数据库中核查, 因为这些系统会生成看似可信、实则不存在的参考文献, 我们对 AI 生成参考文献的审计 将这种虚构比例估为请求参考文献表的 15.2 percent。AI 引文检查器 会一次性核查整份粘贴的列表, 而 主引文生成器 则直接从 Crossref 生成经过验证的条目。一旦来源确定, TextPulse humanizer 就负责让周围的行文不至于也像是模型写的。
Every field in an AI citation, labeled
A single ChatGPT exchange broken into its parts across six styles. Point at any colored segment of the reference entry or an in-text form to see what that piece is and which rule put it there.
OpenAI. (2026, July 14). Correlation and causation explained [Generative AI chat]. ChatGPT-5. chatgpt.com/share/...
APA's current pattern, published 9 September 2025, which cites the conversation rather than the tool. The earlier format, with the product as the title and a version in parentheses, is now marked archival on APA's own site. Alphabetize under the company's name.
(OpenAI, 2026)
OpenAI (2026)...
When prompted, ChatGPT generated... (OpenAI, 2026)
APA recommends describing the prompt in the text or method section; a full transcript can go in an appendix.
(OpenAI, 2026)
AI output has no page numbers, so no p. locator is possible; make clear in the sentence that the words are the model's.
Hover any coloured part for what it is and the rule behind it.
Six style manuals, six different answers
Four distinct positions produce seven finished formats. Here is the reasoning behind each one, and what the generator does with it.
APA hands authorship to the company
OpenAI, Anthropic or Google sits in the author position, credited with the output the algorithm produced. The model name is italicized as the title, the version sits in parentheses, and a bracketed label states the source type.
MLA builds the entry around the prompt
MLA will not name an AI system as an author, so the prompt you typed, in quotation marks, becomes the title that opens the entry, with the tool as the italicized container and a shortened prompt carrying the in-text citation.
Chicago keeps it to a note
Chicago classes an AI exchange as personal communication: a numbered footnote records the model, the prompt, the company and the full date, and neither Chicago system sends any of it to the bibliography.
Harvard treats it as correspondence
Cite Them Right files generative AI under personal communication, on the reasoning that a reader cannot retrieve the exchange. The developer and tool serve as originator, and the full transcript is recommended for an appendix.
IEEE asks for a disclosure
With no dedicated IEEE template, a software citation pattern fills the gap, but what IEEE actually requires is separate: a statement in the acknowledgments declaring that AI-generated content was used.
Vancouver follows ICMJE
Medical journals apply the ICMJE position directly: AI use goes in the methods or acknowledgments, authorship is not on the table, and the reference entry itself borrows the NLM pattern for software.
Ten model families, ready without setup
Every preset carries the developer's real name and web address. Shown here in APA 7; the generator renders each one in all six other styles too.
ChatGPT
OpenAI
OpenAI. (2026, July 14). Correlation and causation explained [Generative AI chat]. ChatGPT-5. https://chatgpt.com/
Claude
Anthropic
Anthropic. (2026, July 14). Correlation and causation explained [Generative AI chat]. Claude Sonnet 4.5. https://claude.ai/
Gemini
Google. (2026, July 14). Correlation and causation explained [Generative AI chat]. Gemini 2.5 Pro. https://gemini.google.com/
Meta AI
Meta
Meta. (2026, July 14). Correlation and causation explained [Generative AI chat]. Llama 4. https://www.meta.ai/
Copilot
Microsoft
Microsoft. (2026, July 14). Correlation and causation explained [Generative AI chat]. Copilot GPT-5. https://copilot.microsoft.com/
Grok
xAI
xAI. (2026, July 14). Correlation and causation explained [Generative AI chat]. Grok 4. https://grok.com/
DeepSeek
DeepSeek
DeepSeek. (2026, July 14). Correlation and causation explained [Generative AI chat]. DeepSeek-V3. https://chat.deepseek.com/
Le Chat
Mistral AI
Mistral AI. (2026, July 14). Correlation and causation explained [Generative AI chat]. Mistral Large. https://chat.mistral.ai/
Perplexity
Perplexity AI
Perplexity AI. (2026, July 14). Correlation and causation explained [Generative AI chat]. Perplexity Sonar. https://www.perplexity.ai/
Qwen
Alibaba Cloud
Alibaba Cloud. (2026, July 14). Correlation and causation explained [Generative AI chat]. Qwen3. https://chat.qwen.ai/
APA 7 is shown here; the generator produces all seven outputs for every model listed. Each version string is a current example, so substitute the one you were served.
The same exchange, every supported style
A single ChatGPT conversation formatted for every style included here. Notice where the prompt appears, where it disappears, and which styles keep AI out of the reference list altogether.
APA (7th edition)
Reference list entry
OpenAI. (2026, July 14). Correlation and causation explained [Generative AI chat]. ChatGPT-5. https://chatgpt.com/
MLA (9th edition)
Works Cited entry
"Explain the difference between correlation and causation for a first-year statistics student" prompt. ChatGPT, ChatGPT-5, OpenAI, 14 July 2026, chatgpt.com.
Chicago (notes and bibliography)
Footnote (Chicago cites AI in notes only)
ChatGPT, response to "Explain the difference between correlation and causation for a first-year statistics student," OpenAI, 14 July 2026.
Chicago (author-date)
In-text citation (no reference list entry)
(ChatGPT, 14 July 2026)
Harvard (Cite Them Right)
Reference list entry
OpenAI ChatGPT (2026) ChatGPT response to prompt 'Explain the difference between correlation and causation for a first-year statistics student', 14 July.
IEEE
Reference list entry
[1] OpenAI, ChatGPT (ChatGPT-5) [Large language model], 2026. [Online]. Available: https://chatgpt.com/ (accessed 14 July 2026).
Vancouver
Reference list entry
1. ChatGPT (ChatGPT-5) [Large language model]. OpenAI; 2026 [cited 14 July 2026]. Available from: https://chatgpt.com/
The seven entries above are generator output. Replace the prompt and the dates with the details of the conversation you are citing.
此生成器的 4 种使用方式
学生
在允许的情况下引用 AI 使用情况，且以风格手册规定的准确形式呈现，不做临时猜测。
研究人员和学者
用同一条参考文献同时满足期刊的披露要求和引用要求。
AI 辅助写作者
清楚区分模型写了什么, 它建议了什么, 以及哪些内容仍需核实。
教师和学术诚信负责人
一页展示课堂如何按 APA、MLA 和 Chicago 引用 AI, 包括分歧。
AI 来源已正确引用。
其余草稿仍需要你再检查一遍。
TextPulse humanizer 通过对整篇草稿的措辞和节奏进行处理来工作，同时包括 AI 引文在内的每一处引文都保持原位不变。Tracked changes 会覆盖每一处编辑，并附带一个免费演示。