Skip to content
CitationZ

How to Cite an AI-Generated Image

Updated October 2026 · Reviewed October 2026

Citing an image you made with Midjourney, DALL·E, Firefly or another generator? There is no human artist to name, and that is exactly where APA, MLA and Chicago stop agreeing.

One question decides the whole entry: what goes where your prompt would? APA leaves the prompt out of the reference entirely. MLA and Chicago put it in, quoted. Get that right and the rest follows.

Your prompt: out in APA, in for MLA and Chicago

APA addresses this directly, and the example it chooses is an image prompt. After showing a writer asking Gemini to “create an image of students in a classroom studying grammar concepts,” APA says: the reference you create would be for the chat, not the prompt.

It gives four reasons. A prompt is not one of APA's four reference elements (author, date, title, source). It does not help anyone retrieve the work. It is usually several prompts refined across a conversation, not one line. And it is typically too long — APA notes that it truncates social-media titles at 20 words, and a prompt cut to 20 words would stop being useful.

So in APA the prompt still matters, it just lives elsewhere: in your text, the Method section, an appendix, or the note under the figure.

MLA and Chicago go the other way. MLA's Title of Source is a description of what the tool generated, and the usual way to write that is your prompt, in quotation marks and not title-cased. A shortened description is explicitly allowed — MLA offers Theme of nature in Austen's Mansfield Park prompt in place of the full sentence — so quote the whole prompt or describe it, but do not tidy it into a title. Chicago quotes it too. So the same image produces visibly different entries in the three styles. That is correct, not a slip, and it is the single thing most generators get wrong.

APA: cite the tool — or the chat, when pointing at it helps

Company. (Year). Tool or model [Description — e.g., Text-to-image model]. https://tool-url

APA does not currently publish a dedicated example for this exact source type. The format below applies APA's current AI citation principles to this source.

What APA does publish is two worked AI forms: an AI tool — Google. (2025). Gemini 3 Flash [Large language model]. https://gemini.google.com — and an AI chat, which adds the day and a title and moves the tool's name after the brackets. See APA's AI references examples page and its reference-formats post and judge for yourself. The entry below is APA's tool form, element for element — the company as author, the year alone, the model or tool in the title slot, a bracketed label and the link — plus, on this entry, the version in parentheses, which APA permits but stopped asking for by default in 2025. What is ours is the wording inside the brackets: [Text-to-image model] appears in no APA example. APA permits exactly that choice — "other descriptions are possible; use the description that best fits the type of AI model or tool used" — so treat [Text-to-image model] as the wording Citation[Z] chose under that permission, and use your instructor's if they prefer another.

For an image you are putting in a paper, the tool form is the usual answer. APA lists creating “tables, figures, or other visual representations” among the uses where you cite the tool generally and record the AI use in the figure note — so reach for this form even if you could share the conversation. Having a share link does not oblige you to use it: APA says to cite the tool generally whenever citing the specific chat would be “unhelpful, unethical, or otherwise inappropriate.” Use the chat form instead when you have a link and pointing a reader at that particular conversation genuinely helps them.

The date is the year the tool was last updated — a fact about the tool, not the day you generated your image. If the tool does not say, APA accepts the copyright year on its site.

The bracket describes whatever the title names, under APA's own rule that “other descriptions are possible; use the description that best fits the type of AI model or tool used.” For a dedicated generator like Midjourney, [Text-to-image model] fits better than APA's usual [Large language model]. But if you made the image inside a chat tool — ChatGPT, Gemini — then [Large language model] is APA's own published wording for that tool and is correct, not a mistake. APA does not call either choice wrong; use your instructor's wording if they prefer another, and stay consistent.

APA 7 — APA's tool form, applied to an image

Midjourney. (2026). Midjourney (6.1) [Text-to-image model]. https://www.midjourney.com

In-text(Midjourney, 2026)

Cite an AI image — free

Set up for this page. Paste and go.

We'll identify it and build your citation.

Already have a reference list? Format it →

Citation style

Publication Manual of the American Psychological Association, 7th edition (2020). The 7th dropped the publisher location and the "Retrieved from" prefix that the 6th required.

MLA and Chicago: the prompt is the title

MLA does not treat the tool as an author at all. The prompt goes in quotation marks as typed — and the word prompt sits just outside the closing quotation mark, which is MLA's own placement, not a typo. Then the tool as the container, the model or version, the company, the date and the link.

Chicago also keeps the prompt quoted and as typed. One honest caveat: what Chicago publishes for AI images is a credit line for a figure, not a bibliography entry. The entry below applies Chicago's principles in the notes-bibliography form our generator uses — Chicago has published no counterpart to check it against, so treat it as our application rather than Chicago's printed form. For the same reason the generator shows no footnote for it.

MLA 9 — the prompt as Title of Source

"Create a bioluminescent forest at night with glowing mushrooms and deep blue mist" prompt. Midjourney, version 6.1, Midjourney, 12 Mar. 2026, https://www.midjourney.com.

In-text("Create a bioluminescent")

Chicago bibliography — our application, not a published form

Midjourney. "Create a bioluminescent forest at night with glowing mushrooms and deep blue mist." Midjourney. March 12, 2026. https://www.midjourney.com.

Footnote: Chicago's note form for this source type is not verified yet, so we don't show one.

The pieces

  • Author — in APA and Chicago, the company behind the tool (Midjourney, OpenAI, Adobe), never the model and never the image. APA's reasoning settles every argument about this: an AI “cannot be an author because it is not a living, conscious human who can give consent”, while the author element is whoever is “responsible for a work”. MLA has no author element here at all. You are the prompter, which is a third thing again.
  • Title — this is the element that splits. APA's title is the tool or model, in italics. MLA's and Chicago's is your prompt, in quotation marks and as typed — not tidied, not title-cased.
  • Model or version — name the one you actually used. APA puts a model in the title slot and permits a genuine version number in unitalicized parentheses after it; MLA labels the element “model” or “version” according to which it is.
  • Bracketed description — APA's, describing what the title names. The wording is adjustable by APA's own rule; be consistent across your reference list.
  • Where it lives — the tool's address for APA's tool form, or a share link when you are citing a specific conversation.

If the image goes in your paper, it is a figure

APA 7's figure parts are a bold figure number, then a brief title in italic title case on the line below it, then the image, then a note underneath. APA 7 does not use the word “caption”, and the descriptive line sits above the image rather than below it — a leftover from the 6th edition that still trips people up.

APA says to indicate that you used AI in the table or figure note. Since the note is where that disclosure lives, it is also the natural place to record the prompt — which is exactly where APA's exclusion of prompts from references leaves it.

Before you cite it: are you allowed to use it?

Formatting the citation correctly is not the same as being permitted to use the image. Whether AI-generated work may appear in your assignment at all depends on your instructor, your department or your publisher — and APA is blunt that submitting AI-generated work as though you produced it yourself is academic misconduct. Check the policy first, then cite.

Citing a specific tool

The form is the same for every generator — what changes is the company name, how the tool writes its model names, and where it lives. The tool guides fill those in: Midjourney, DALL·E, and Adobe Firefly.

Cite the tool as it was when you used it. If a product has since been renamed or retired, that does not reach back and change the reference for work you did earlier — name what you actually used.

Keep every citation you make.

A free account saves your citation history and organizes sources into projects with notes and tags.

Create a free account

Find the sources you should be citing.

Premium searches 250 million scholarly works by topic, recommends citations for your claims, and flags statements in your writing that need support.

Go Premium — $5/month

Frequently asked questions

Do I put my prompt in the citation?
It depends on the style, and this is the most common mistake. APA: no — APA states that the reference is for the chat, not the prompt, and gives four reasons; your prompt goes in your text, the Method section, an appendix, or the figure note instead. MLA and Chicago: yes — the prompt is the quoted title. MLA asks you to describe what was generated, so the full prompt or a shortened description of it both work; what you must not do is title-case it or tidy it into a name.
Who is the author of an AI-generated image?
In APA and Chicago, the company behind the tool — Midjourney, OpenAI, Adobe. Not the model: APA holds that an AI cannot be an author because it is not a living, conscious human who can give consent, and defines the author element as whoever is responsible for a work. MLA has no author element for generated work at all. You are the prompter, which is not the same as being the author.
Should I cite an AI image, or put it in as a figure?
Usually both. If the image appears in your paper it is a figure, so it needs a figure number, an italic title above it and a note underneath recording the AI use — and the reference points at the tool you made it with. Check your assignment: some instructors want only the figure and its note.
What if there is no public URL for the image?
That is the normal case, and it is why APA's tool form exists: the source element is the tool's own address, not a link to your picture. If the image exists only in your project, show it as a figure and say in the note how it was made.
Which year do I use?
For APA's tool form, the year the tool was most recently updated — not the date you generated the image. APA says to ask the AI if you are unsure, and to fall back on the copyright year shown on the tool's site. MLA and Chicago take the date you generated it.

Related

Generate a citation now — free