Annotated bibliography examples and format

What is an annotated bibliography?

Two things are happening in every entry, and rubrics mark them separately. The first is the citation: a properly formatted entry, identical to what you would put in a reference list or works-cited list. The second is the annotation: your own paragraph about that source. Most lost marks come from doing the first carefully and the second thinly.

What goes in an annotation

Three moves, usually in this order.

• Summary. What does the source claim, and what is that claim built on — a sample, an archive, a set of interviews, a close reading? Two or three sentences.
• Assessment. How strong is the evidence, and where does it stop? Say what the design cannot show, not only what it shows.
• Reflection. What is this source doing in your project — which section does it feed, which claim does it support, which other source does it argue with?

Assignments differ in how many of the three they want. If the rubric only says “annotated bibliography”, write all three: a paragraph that summarizes, assesses and reflects is never marked down for doing too much.

Annotation vs. abstract

An abstract is written by the source’s own authors to describe their paper for people deciding whether to read it. It is neutral by design. An annotation is written by you, for one project, and it does the two things an abstract never does: it judges the source, and it says what you need it for.

That is also the practical reason not to recycle the abstract. A published abstract is indexed — on the publisher’s page, in the database record, on aggregator sites — so an annotation that is a light paraphrase of one is a rewrite of a document the matching engine already holds. That is how paraphrased text surfaces in a similarity report.

Descriptive, evaluative and reflective annotations

The three labels describe which of the three moves dominates, not three unrelated genres.

• Descriptive (or indicative): mostly summary. Common when the bibliography surveys a field.
• Evaluative (or critical): summary plus judgment. The most commonly assigned kind.
• Reflective: adds what the source contributes to your argument. Standard when the bibliography is a step toward a longer paper.

If your rubric names one of these, it is telling you which move to weight — not giving you permission to drop the others.

A full annotated bibliography example

Three sources, alphabetized by surname, under a page heading — what the assignment actually asks for, rather than a single entry in isolation. The annotations run 106 to 134 words each, which is a normal working length.

Notice what changes when you switch style and what does not. The sources, the order and the annotations are constant. The indents, the spacing and the punctuation of the entries are all that move.

Annotated bibliography format by style

The citation half never changes: format each entry exactly as you would in an ordinary reference list. What differs between styles is how the annotation is set off from it — the indent, the spacing, and whether the annotation starts on its own line.

One honest note on that table: APA and MLA publish a rule you can point at. Chicago does not. Where the Chicago column says “usually” or “not specified”, that is because CMOS genuinely leaves it open, not because the rule was left out.

What about Harvard?

Harvard is not a single manual — it is a family of author–date styles, and your university’s own guide is the authority for the version you are marked against. There is no Harvard equivalent of MLA §5.132, so the annotation layout comes from your department. The usual setup mirrors APA: a hanging indent on the entry, the annotation indented underneath, double-spaced. Build the Harvard reference first, then check your faculty guide for the annotation.

How to write an annotated bibliography, step by step

1. 1

   Choose and read your sources

   Work from the reading list or your database results, and skim for the argument and the method before you commit. You cannot evaluate a source you have only read the abstract of — and the abstract is the one part you must not lean on when you write.

2. 2

   Build the reference entry

   Format each source exactly as you would in a normal reference list or works-cited list. Get this right before you write anything, because the annotation hangs off it. Build the reference entry.

3. 3

   Summarize the argument and the method in your own words

   Two or three sentences: what the source claims, and what evidence it rests on — the sample, the archive, the case study. Write this from your notes with the source closed.

4. 4

   Add your assessment

   One or two sentences on how convincing it is and where it is limited, then a sentence on what it does for your project. This is the part that separates an annotation from a summary, and it is the part rubrics grade.

5. 5

   Alphabetize and format the page

   Sort by the first author’s surname, apply the hanging indent to every entry, indent the annotations, and set the line spacing your style requires.

How to set the hanging indent in Word and Google Docs

This is the step that eats the evening, and nobody explains it on the page where you need it. Do it once, at the end, on the whole list — not entry by entry.

Google Docs

1. Select every entry in the list.
2. Format → Align & indent → Indentation options.
3. Under Special indent, choose Hanging and set it to 0.5.
4. Apply.

On the ruler (View → Show ruler) the same thing is two drags: pull the triangle — the left indent — right to 0.5″, then pull the rectangle above it — the first-line indent — back to 0.

Microsoft Word

1. Select every entry in the list.
2. Home tab → the small arrow in the corner of the Paragraph group.
3. Indentation → Special → Hanging → By: 0.5″.
4. OK.

The keyboard shortcut is faster once you know it: Ctrl+T (⌘+T on a Mac) applies a hanging indent to the selected paragraphs, and Ctrl+Shift+T removes it.

The second indent MLA requires

MLA wants the annotation another half inch in, so it is a second operation on a different selection. Select onlythe annotation paragraphs — not the citation lines — and set the left indent to 1 inch with no special indent. In Word: Paragraph → Indentation → Left: 1″, Special: (none). In Google Docs: Indentation options → Left indent 1, Special indent: None.

If you apply the hanging indent to the annotation too, its first line juts back out to the margin and the annotation stops looking like an annotation. That is the single most common MLA formatting error on this assignment.

How long should each annotation be?

You will find 50–200 words on one site, 100–300 on the next and “about 150” on a third, with nobody explaining why they disagree. Here is why: almost none of those numbers come from a style manual.

APA §9.51 sets the layout of an annotation but not its length. Purdue’s guidance is explicitly that length depends on purpose and ranges from a couple of sentences to a couple of pages. MLA is the only one of the three that commits, and it commits in the opposite direction from most of the internet: §5.132 asks for succinct phrases and rules out rehashing detail, quoting the author, or recounting an argument step by step.

So: the rubric first. If your assignment gives a word count, that number outranks every guide, including this one. If the rubric is silent, one paragraph of 100–200 words is the safe default for APA and Chicago, and shorter for MLA. Structure it as two or three sentences of summary, one or two of assessment, one of reflection — which is what the examples on this page do.

Length is a proxy anyway. A 90-word annotation that names the method and says what the source does for your argument marks better than a 250-word one that retells the introduction.

Annotation templates and sentence starters

Skeletons for when the blank page is the problem. Fill the blanks from your notes, then delete any sentence that ended up saying nothing.

Common mistakes that cost marks

• Paraphrasing the abstract

  The most common mistake and the most expensive one. It produces a bland annotation that says nothing about your project, and it is the single biggest source of similarity matches in an annotated bibliography.

• Summarizing without evaluating

  A paragraph that only describes what the source says is a summary, not an annotation. Rubrics almost always ask for judgment — how strong, how limited, how useful to you.

• Losing the hanging indent

  Pasting an entry from a database or a generator usually drops the indentation. Apply the hanging indent to the whole list at the end, in one go, rather than fixing entries individually.

• Indenting the annotation the same amount as the entry in MLA

  MLA is the one style with two different indents: 0.5″ hanging on the entry, 1″ on the annotation. If both sit at 0.5″ the annotation stops being visually distinct, which is the whole point of the rule.

• Sorting by the order you found things

  Entries go in alphabetical order by surname, not in reading order or by importance.

• Mixing editions

  A Chicago bibliography built from a 17th-edition guide will include a place of publication that the 18th edition dropped for books published after 1900. Check which edition your rubric names.

Will an annotated bibliography get flagged by Turnitin?

Partly, yes — and the part that gets flagged is usually not the part students worry about. Turnitin produces two separate reports on a submission, a similarity report and an AI writing report, and an annotated bibliography can trip each of them for a different reason.

Why the reference entries match, and why that is usually fine

A bibliography is built out of strings that thousands of indexed documents contain verbatim: author names, article titles, journal names, volumes, page ranges, DOIs. A matching engine finds repeated strings, so it finds these. On a short assignment that is mostly bibliography, the entries alone can account for a visible share of the similarity score.

That is a measurement, not an accusation — the score counts text overlap and decides nothing about misconduct. Instructors can also switch on a filter that excludes bibliographic material from the score, and many do for this kind of assignment, which is why two students with identical formatting can see very different percentages. What a similarity score actually counts is worth reading before you panic about a number.

The annotations are where the real similarity risk is

If you write your annotations by paraphrasing each source’s published abstract, you are rewriting a document that is already indexed — on the publisher’s page, in the database record, on aggregator and preprint sites. A light paraphrase keeps most of the noun phrases and the order of the clauses, and that is precisely the shape a matching engine catches.

It scales badly, too. One abstract-derived annotation is a small match. Twenty of them is a match inside every entry, spread evenly through the document — the pattern that looks worst in a report and is hardest to explain away. And it is self-defeating academically: an abstract cannot say what the source does for your argument, so an abstract-derived annotation is thin on exactly the sentence the rubric grades.

The fix is procedural rather than verbal. Read the source, close it, then write the summary from your notes. If you need a specific formulation from the abstract, quote it in quotation marks and cite it — a short, properly attributed quotation was never the problem.

AI-written annotations and the AI writing report

The AI writing report is separate from the similarity score and works differently: it looks at how the prose is written, not at what it matches. An annotated bibliography is structurally one of the worst assignments to hand to a chatbot, because summarizing a source in neutral academic register is the register a language model produces by default — even, hedged, evenly paced, with no idiosyncratic judgment in it. That flatness is the signal.

Two practical notes. The AI report needs roughly 300 words of qualifying prose before it will run at all, and Turnitin surfaces scores above about 20%, so a three-entry bibliography may be too short to score while a twenty-entry one is well past both thresholds. And language models routinely misdescribe real papers — invented findings, wrong sample sizes, methods the study never used — so an unverified AI annotation can make claims about a source you have said you read. How Turnitin’s AI detection works covers what the indicator does and does not measure.

Frequently asked questions