← Back to Blog

Short answer: Do not dictate citations inline with your prose. Dictate the argument at full speed, drop a spoken placeholder where each source belongs, then insert the actual citations afterward using your reference manager. For footnotes, create the empty note first with a keyboard shortcut, then dictate its contents.

Dictation is excellent at prose and terrible at citations, and most people discover this in the worst possible way: mid-paragraph, three sources deep, watching their document fill with things like "Smith comma J period open paren 2019 close paren" rendered as literal words.

The instinct after that is to give up on dictation for academic, legal, and research writing entirely. That is an overcorrection. The fix is not to stop dictating. It is to stop treating citations as part of the sentence you are speaking.

This guide covers the technique in detail: why citations break dictation, the punctuation vocabulary that gets you through when you do need to speak one, the insert-then-dictate pattern for footnotes in every major writing app, and the placeholder workflow that keeps your drafting speed intact.

Why citations break dictation

Four things make a citation structurally hostile to speech.

Punctuation density. Ordinary prose has maybe two punctuation marks per sentence. A single citation can have eight: parentheses, commas, periods after initials, a colon before page numbers, an ampersand, quotation marks, a hyphen inside a page range. Every one has to be spoken explicitly, which means the ratio of formatting commands to actual content flips upside down.

Proper nouns with no context. Transcription engines use surrounding language to decide between similar-sounding words. Author surnames arrive with no context at all. A name you have read a hundred times in a PDF may be a name the engine has never encountered, and there is no sentence around it to disambiguate.

Abbreviations that are not words. Journal abbreviations, reporter abbreviations, "et al.", "ibid.", "cf.", "s.v.", volume and issue markers. These are visual conventions. Nobody says them aloud in normal speech, so nobody has trained on them as speech.

Formatting that lives outside the text. Italics for journal titles, superscript footnote markers, hanging indents. These are not characters you can say. They are document formatting, and dictation puts characters at your cursor, not styles.

Put those together and you get the fundamental insight: a citation is not language, it is a data structure that happens to be rendered in letters. Speaking a data structure aloud is always going to be miserable. So do not.

The core rule: draft in prose, cite in a second pass

Here is the workflow that actually works, and it takes about ten minutes to internalize.

Pass one is prose only. You dictate your argument end to end, at speech speed, without stopping. Where a source belongs, you say a placeholder instead of the citation. Nothing about your thinking gets interrupted to spell out an author's initials.

Pass two is citations only. You go back with your reference manager, find each placeholder, and insert the properly formatted citation with the plugin that already exists for exactly this purpose.

This is faster than dictating citations inline even if you type every citation by hand in pass two, for a reason that has nothing to do with typing speed. Drafting and citing are different cognitive modes. Drafting is generative and wants momentum. Citing is verification work and wants precision. Interleaving them means you pay a context-switch cost at every source, and in a paper with sixty citations that cost is enormous.

Most experienced academic writers already work this way with a keyboard. Dictation just makes the gap between the two modes more obvious.

Choosing a placeholder you can actually find

The placeholder needs three properties: easy to say, unambiguous to the transcription engine, and searchable without false positives.

Good options:

What matters most is consistency. Pick one and use it for the whole document, because a search that misses three placeholders is worse than no system at all. Before you start pass two, run a find-all and note the count. That number is your checklist.

Add your chosen placeholder to Smart Vocabulary if the engine keeps mangling it. Voice Keyboard Pro's personal dictionary supports replacement rules, so you can say a short phrase and have a longer, uglier, more searchable token land in the document. Speak "cite mark", get [[CITE]]. That is exactly the kind of thing a replacement rule is for.

Footnotes: create the note, then dictate into it

The single most common footnote mistake is trying to dictate the footnote marker. You cannot speak a superscript. What you can do is create the empty footnote with a keyboard shortcut, which drops your cursor into the note area, and then dictate normally.

Because Voice Keyboard Pro types at the cursor system-wide rather than integrating with each app, this pattern works identically everywhere. Wherever your cursor lands, that is where your words go.

Microsoft Word for Mac

Press Command + Option + F to insert a footnote. The cursor jumps to the note at the bottom of the page. Hold your dictation hotkey, speak the note content, release. Press Command + Option + Shift + F or click back into the body to return. Our fuller walkthrough of speaking into Word is in how to dictate in Microsoft Word on Mac.

Google Docs

Press Command + Option + F to insert a footnote. Same pattern: cursor moves into the note, you dictate, you click back up. Docs handles the numbering, so you never say a number.

Apple Pages

Use Command + Option + S for a footnote, or insert one from the Insert menu. The cursor drops into the note area and dictation proceeds as normal.

LaTeX and Overleaf

This one deserves care, because you are dictating markup and markup is not speech. The workable approach is to type the command structure and dictate only the human-readable content inside it. Type \footnote{}, put the cursor between the braces, then dictate the sentence. Trying to speak backslashes and braces is a losing battle that produces more cleanup than it saves. We cover the general principle in dictating in Overleaf: speak the prose, type the syntax.

Scrivener

Inspector footnotes live in a side panel rather than at the page bottom. Select your text, create the note, click into the panel, dictate. The panel is a normal text field, so nothing special is required.

The rule that covers every app: your hands create the container, your voice fills it.

When you do have to speak a citation

Sometimes there is no reference manager. You are answering a peer review comment, writing an email to a colleague, drafting a memo where one case cite has to appear inline. For those moments, here is the punctuation vocabulary you need.

Voice Keyboard Pro's transcription engine handles spoken punctuation, and these are the ones that appear in citations:

For a wider reference on speaking symbols, dictating quotes and special characters covers the full set, and dictating numbers and dates covers year and page-number behavior, which matters more in citations than anywhere else in your writing.

Worked examples

APA in-text, one author. Say: open parenthesis Smith comma 2019 close parenthesis. You get (Smith, 2019).

APA in-text, two authors with a page. Say: open parenthesis Smith ampersand Jones comma 2019 comma p period 47 close parenthesis. This one is already at the edge of what is worth speaking. If you find yourself doing it more than a few times per document, go back to placeholders.

MLA parenthetical. Say: open parenthesis Smith 47 close parenthesis. MLA is the friendliest style for dictation because the in-text form is minimal: name, space, page number, no punctuation between them.

Chicago note. Author first name, last name, comma, title in italics, publication details in parentheses, comma, page. The italics are the problem. Dictate the plain text, then apply italics with Command + I over the title afterward. Do not try to voice a style change.

Legal citation. These are the hardest, because they are dense with abbreviations that are not words. A reporter cite has a volume number, an abbreviated reporter, a page, and a parenthetical with court and year. Speaking that aloud produces something you will spend longer fixing than typing. Our honest recommendation for legal writers: dictate the argument, type the cites, and put your recurring case short-names into Smart Vocabulary so that at least the ones you use forty times in a brief come out right. There is more on the surrounding workflow in voice to text for lawyers.

Let the reference manager do the formatting

This is the part people skip, and it is the part that makes everything else unnecessary.

Zotero, Mendeley, EndNote, Papers, and the citation tools built into Google Docs and Word all exist to solve the exact problem you are trying to solve by mouth. They store the bibliographic data once, insert correctly formatted citations with a keystroke, renumber automatically when you move a paragraph, and rebuild the bibliography when you change styles.

No amount of careful dictation competes with that, because dictation cannot renumber your footnotes when you cut a section, and it cannot convert your document from APA to Chicago because a journal asked.

So the division of labor is clean:

A researcher who works this way gets the full speed benefit of dictation on the 90 percent of the document that is prose, and pays nothing for the 10 percent that is not. Comfortable speech runs about 130 to 150 words per minute against roughly 40 for average typing, so that 90 percent is where the entire gain lives. Fighting over the other 10 percent is the wrong optimization.

Smart Vocabulary: teach it your sources once

If you are writing in one field, the same twenty surnames, journal names, technical terms, and case names will appear over and over across every paper you write for years. Fixing them by hand every time is a slow leak.

Smart Vocabulary on Mac is a personal dictionary with replacement rules, and this is the highest-value use of it for academic and legal writers. Load it with:

The setup cost is ten minutes once. After that, the specific words that used to break every paragraph simply stop breaking. If you want the deeper version of this, how custom vocabulary learns your words walks through it.

Fixing citation errors without retyping

Even with placeholders, some cleanup is inevitable. Two things make it faster.

Find and replace is your primary tool. If the engine consistently renders a name wrong in the same way, one replace-all fixes every instance in the document at once. This is much faster than catching them one at a time while reading, and much more reliable.

On iPhone, Voice Edit handles the fiddly repositioning. Editing citation-heavy text on a phone is genuinely painful because tapping precisely between two parentheses is a fine-motor task. Voice Edit lets you speak the change you want rather than placing a cursor. For a broader treatment of the editing pass, see how to edit dictated text without retyping the whole thing.

One more technique worth knowing: if a proper noun comes out wrong and you need it right immediately, you can spell it out. Spelling words while dictating covers when this works and when it is slower than just fixing it later.

A realistic session, start to finish

Here is what a chapter-drafting session looks like with all of this in place.

  1. Before you speak, have your sources open. Not to cite from, but so your brain is not reaching for half-remembered details mid-sentence. The number one cause of dictation stalling is not knowing what you are about to say.
  2. Draft the section in one pass. Hold the hotkey, speak in full paragraphs, say your placeholder wherever a source belongs. Do not stop to fix transcription errors. Do not stop to look anything up. Momentum is the entire point.
  3. Read it back and fix prose only. Wrong words, sentences that went sideways, paragraph breaks. Still no citations.
  4. Find-all your placeholders. Note the count. Work through them with the reference manager, one at a time.
  5. Insert footnotes last. Create each note with the shortcut, dictate the content, move on. Because your reference manager numbers them, you can insert in any order.
  6. Final search for stragglers. Search the placeholder token again. It should return zero. If it does not, you just caught the mistake that would have made it to a reviewer.

For the surrounding habits that make long-form academic drafting work by voice, dictation for academic writing and voice typing for PhD students both go further than this post does on structure and outlining.

Common questions

Can I dictate an entire bibliography?

You can, but you should not. A bibliography is pure structured data, which is the worst possible input for speech. Every reference manager exports one automatically from records you already have. If you genuinely have no manager, typing is faster than dictating here, and it is one of the few places in writing where that is true.

Will dictation insert footnote markers for me?

No. Footnote markers are a document feature, not text. Your word processor creates them, which is good, because it also renumbers them when you rearrange sections. Create the note with the shortcut, then dictate into it.

How do I dictate "et al."?

Say it as et al period and check the result, because engines vary on whether they treat it as a known abbreviation. If yours does not get it reliably, a Smart Vocabulary replacement rule solves it permanently.

What about italics for journal and book titles?

Dictate the title as plain text, then select it and press Command + I. Styling is not something you can speak, and any workflow that promises otherwise is adding a layer you will fight with.

Does this work the same on iPhone?

The placeholder method does, using the Voice Keyboard Pro keyboard's mic button in whatever app you are drafting in. Footnote insertion depends on the app, and mobile word processors vary widely in whether they support notes at all. For serious citation work, the Mac is the better surface, and phones are best used for capturing prose you will cite later.

Is any of this specific to one style guide?

The placeholder-then-cite workflow is style-agnostic, which is precisely its advantage. You are not encoding APA or MLA or Chicago into how you speak, so if a journal changes the required style, your reference manager reformats everything and your drafting habits never change.

The takeaway

The reason dictating citations feels impossible is that you are asking speech to do a job it was never suited for. Citations are structured data wearing the costume of text. Prose is the thing your voice is genuinely faster at, and in most academic, legal, and research documents prose is the overwhelming majority of the words.

Split the job along that line. Speak the argument, mark the sources, cite in a second pass, let the tools that were built for citation formatting do the citation formatting. You keep the speed where the speed matters and you lose nothing where it does not.

Voice Keyboard Pro has a free tier if you want to test the workflow on a section you are already working on. Draft one with placeholders, cite it afterward, and compare how it felt against your last citation-heavy paragraph typed straight through.

Keep reading