Quick reference
An annotated bibliography is an alphabetical list of sources where each citation is followed by a 100–200 word annotation. A strong annotation makes three moves in order: summarize the source's question, method, and findings; evaluate its credibility, evidence, and limitations; and relate it to your own project. The citation format follows your assigned style — APA 7, MLA 9, or Chicago 17 — but the annotation structure is the same in all three.
The three moves every annotation must make
Most rubrics — whether they say so explicitly or not — grade an annotation on three moves. Summarize: in two or three sentences, state what the source asked, how it answered, and what it found. Evaluate: judge the source's authority — peer review, sample size, method, limitations, and how the field has received it. A fast, defensible way to make the reception claim is a forward-citation check: how many times has the source been cited, and from where? A geographic citation map shows the spread of citing institutions at a glance — global spread signals a field-defining source; a single-cluster footprint signals a niche one. Relate: say what the source does for your project specifically — which section it supports, which claim it anchors, what it cannot do for you. The worked entries below label all three moves so you can see the proportions.
Three worked annotated bibliography examples
Each example uses a real, verifiable, heavily cited paper so the citation mechanics are checkable against the published record. The annotations are illustrative — written for a hypothetical student project — but they are full-length, rubric-ready entries, not fragments.
Example 1 · APA 7
Attention Is All You Need (machine learning)
Citation (APA 7)
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., & Polosukhin, I. (2017). Attention is all you need. Advances in Neural Information Processing Systems, 30, 5998–6008.
Annotation (~150 words, three moves labeled)
[Summarize] Vaswani et al. introduce the Transformer, a sequence-transduction architecture that replaces recurrence and convolution entirely with multi-head self-attention. On the WMT 2014 English-to-German and English-to-French translation benchmarks, the model sets new state-of-the-art BLEU scores while training in a fraction of the time required by recurrent baselines, because attention parallelizes across sequence positions.
[Evaluate] The work is peer-reviewed (NeurIPS), authored by a Google Brain/Research team, and is now among the most-cited papers in computer science — a forward-citation check shows its citing institutions span every continent, a strong signal it is field-defining rather than niche. Its limitation for my purposes is that the original evaluation is confined to machine translation, and self-attention's quadratic memory cost in sequence length is acknowledged but unsolved.
[Relate] This paper anchors the architectural thread of my review of efficient sequence models. I will use its quadratic-attention bottleneck as the problem statement that my project's sparse-attention method directly addresses.
Format notes
APA 7: hanging indent on the citation, annotation as a new indented paragraph below it. Sentence case for the article title; italicize the journal or proceedings name. Order entries alphabetically by first author surname.
Example 2 · MLA 9
AlexNet / ImageNet classification (computer vision)
Citation (MLA 9)
Krizhevsky, Alex, et al. “ImageNet Classification with Deep Convolutional Neural Networks.” Communications of the ACM, vol. 60, no. 6, 2017, pp. 84–90.
Annotation (~150 words, three moves labeled)
[Summarize] Krizhevsky, Sutskever, and Hinton report a deep convolutional neural network trained on 1.2 million ImageNet images that achieved a top-5 error of 15.3 percent — against 26.2 percent for the next-best entry — in the 2012 ILSVRC competition. The paper credits GPU training, ReLU activations, and dropout regularization for making a network of this depth trainable at all.
[Evaluate] This is the retrospective Communications of the ACM version of the 2012 NeurIPS paper, written after the result had already reorganized the field; it is therefore unusually candid about which design choices mattered. The benchmark numbers have long since been superseded, and the paper is an engineering report rather than a theoretical account — readers seeking an explanation of why depth works must look elsewhere.
[Relate] In my historiography of the deep-learning turn, AlexNet is the periodization marker: every source in my bibliography is classified as pre- or post-2012 relative to this result.
Format notes
MLA 9: hanging indent throughout — citation and annotation form one indented block. Title case with quotation marks for the article title; italicize the container (journal). “et al.” after the first author when there are three or more. Alphabetical order by author surname.
Example 3 · Chicago 17
Diabetes Prevention Program RCT (public health)
Citation (Chicago 17)
Knowler, William C., Elizabeth Barrett-Connor, Sarah E. Fowler, Richard F. Hamman, John M. Lachin, Elizabeth A. Walker, and David M. Nathan. “Reduction in the Incidence of Type 2 Diabetes with Lifestyle Intervention or Metformin.” New England Journal of Medicine 346, no. 6 (2002): 393–403.
Annotation (~150 words, three moves labeled)
[Summarize] This multicenter randomized controlled trial assigned 3,234 adults at high risk of type 2 diabetes to placebo, metformin, or an intensive lifestyle-modification program. Over an average follow-up of 2.8 years, the lifestyle intervention reduced diabetes incidence by 58 percent and metformin by 31 percent relative to placebo — establishing that behavioral intervention can outperform first-line pharmacotherapy for prevention.
[Evaluate] The trial's randomized design, large and ethnically diverse sample, and pre-registered endpoints make it the methodological benchmark in its literature; two decades of forward citations across clinical, policy, and behavioral-science venues confirm its field-defining status. Its main limitation is external validity: the lifestyle arm's one-on-one coaching is far more intensive than most health systems can deliver at scale.
[Relate] My thesis evaluates digital diabetes-prevention programs, so this trial supplies both the canonical 58 percent effect size my comparisons are measured against and the delivery-cost problem my intervention claims to solve.
Format notes
Chicago 17 (bibliography style): hanging indent on the entry; annotation as an indented paragraph below. Headline-style capitalization with quotation marks for the article title; italicize the journal. First author inverted, subsequent authors in normal order. Alphabetical by author surname.
Descriptive vs evaluative vs reflective annotations
Assignments name three annotation types, and they nest: a descriptive annotation makes only the summarize move; an evaluative annotation adds the evaluate move; a reflective annotation adds the relate move. When the assignment sheet doesn't specify, write evaluative-plus-reflective — the full three-move entry shown in the examples above — because no grader has ever complained that an annotation was too critical or too connected to the project.
Descriptive (summary) annotation
Reports what the source covers — research question, method, findings — without judging it. Common in lower-division coursework and preliminary reading logs.
Example: “Vaswani et al. (2017) propose the Transformer, an attention-only sequence model, and report state-of-the-art BLEU scores on two WMT 2014 translation benchmarks with substantially reduced training time.”
Evaluative (critical) annotation
Adds a judgment of the source's authority, methodology, and limitations. The default expectation in upper-division and graduate assignments.
Example: “The trial's randomized design and 3,234-person sample make it the strongest causal evidence in this literature, but the 2.8-year follow-up leaves long-term durability of the lifestyle effect unestablished.”
Reflective annotation
Adds an explicit statement of how the source serves your project — which section it supports, what claim it anchors, what it cannot do for you.
Example: “I will use AlexNet as the periodization marker in my review's historical section; it cannot, however, support claims about why depth works, so the theoretical section relies on other sources.”
Formatting rules: APA 7 vs MLA 9 vs Chicago 17
The annotation content is style-agnostic; the citation mechanics are not. Five dimensions cover nearly every formatting point a grader will check.
| Rule | APA 7 | MLA 9 | Chicago 17 |
|---|---|---|---|
| Hanging indent | Citation has a hanging indent; annotation starts as a new paragraph, indented under the citation. | Entire entry — citation plus annotation — is one hanging-indented block. | Citation has a hanging indent; annotation is an indented paragraph below it. |
| Entry order | Alphabetical by first author's surname (year breaks ties). | Alphabetical by author surname; title used when no author. | Alphabetical by author surname; 3-em dash for repeated authors is discouraged in current practice. |
| Title capitalization | Sentence case for article titles; title case and italics for journal names. | Title case in quotation marks for articles; italics for containers. | Headline-style (title) case in quotation marks for articles; italics for journals and books. |
| Author formatting | Surname + initials, ampersand before the last author; up to 20 authors listed. | First author inverted; “et al.” for three or more authors. | First author inverted, the rest in normal order; “et al.” in notes for four or more. |
| Annotation length | 100–200 words, one paragraph (assignment-dependent). | Typically 100–150 words; some instructors allow 2–3 sentence descriptive entries. | 100–200 words; longer reflective paragraphs accepted in graduate work. |
When the assignment sheet and the style manual disagree, the assignment sheet wins — instructors routinely customize annotation length and grouping.
From annotated bibliography to literature review
The annotated bibliography is not busywork — it is the raw material of a literature review, and the conversion is mechanical if your annotations made all three moves. Each annotation becomes a row in a synthesis matrix: one column for the source, one for its key finding (from your summarize move), one for its methodological strength or weakness (from evaluate), and one for the theme or sub-question it speaks to (from relate). Sort the rows by theme, and each theme's cluster of rows becomes one paragraph of the review — with a topic sentence that makes a claim across the cluster rather than summarizing each source in turn.
- Annotate 10–30 sources with the three-move structure above.
- Tabulate each annotation as a synthesis-matrix row: source · finding · method quality · theme.
- Group rows by theme; three to six themes is the normal range for a course paper or thesis chapter.
- Synthesize each group: what do these sources agree on, where do they conflict, and what has none of them done? That last answer is your gap statement.
If the next step is the review itself, our how to write a literature review guide walks the synthesis stage in detail, the literature review outline templates give you the section skeleton, and the worked literature review examples show what the finished argument looks like in five fields.
Common grader complaints
Six failure modes account for most lost points on annotated bibliography assignments. All six are checkable before you submit.
The annotation is just a re-worded abstract. Copying the paper's abstract — or paraphrasing it sentence-by-sentence — covers only the summarize move. Graders look for the evaluate and relate moves; without them the entry earns summary-level credit at best.
No evaluation of source quality. An annotation that never says whether the source is peer-reviewed, what its sample or method was, or where it is weak reads as uncritical. One sentence on study design and one on limitations is the minimum.
No connection to your project. The relate move is the most commonly skipped and the most heavily weighted. Every annotation should end by naming what the source does for your specific research question — which section it supports, which claim it anchors.
Citation format errors. Wrong capitalization style, missing hanging indent, ampersand in MLA, “and” in APA — graders treat citation mechanics as a proxy for care. Check each entry against the style's quick-reference before submitting.
Twelve sources from the same lab or country. Breadth requirements are real. A shortlist where every paper shares an author cluster, venue, or geography signals a shallow search. A geographic citation map of your candidate sources makes this failure visible in seconds — and fixable before the deadline.
Sources too old or all secondary. Unless the assignment is historical, instructors expect a recency mix (several sources from the past five years) and at least some primary research articles, not only reviews and textbooks.
A five-minute breadth check before you submit
The breadth complaint — twelve sources from the same lab — is the easiest to fix and the most embarrassing to receive. Before submitting, run a quick audit on your shortlist's lead authors: drop each one's Google Scholar profile into CitationMap and compare the geographic footprints. If every map shows the same three clusters, your bibliography is sampling one school of thought; deliberately add sources from the regions and institutions your maps leave empty. The same forward-citation view also feeds the evaluate move directly — "cited across 40+ countries" and "cited almost exclusively within the authors' own collaboration network" are both one-sentence evaluative claims you can defend.
Frequently asked questions
What is an annotated bibliography?+
An annotated bibliography is a list of sources in a required citation style (APA, MLA, or Chicago) where each entry is followed by a short paragraph — usually 100–200 words — that summarizes the source, evaluates its credibility and limitations, and explains how it relates to your research project. Unlike a literature review, it does not synthesize across sources: each entry stands alone.
How long should each annotation be?+
100–200 words is the standard range for an evaluative annotation; 150 words is a safe default that fits all three moves (summarize, evaluate, relate) in roughly two sentences each. Purely descriptive annotations can run 2–3 sentences. Always check the assignment sheet first — some instructors specify an exact range, and exceeding it signals you can't prioritize.
What's the difference between APA, MLA, and Chicago annotated bibliographies?+
The annotation content is identical across styles — only the citation mechanics differ. APA 7 uses sentence-case article titles, surname-plus-initials authors, and an ampersand before the final author. MLA 9 uses title case in quotation marks, “et al.” for three or more authors, and formats the whole entry as one hanging-indented block. Chicago 17 (bibliography style) uses headline-case titles, inverts only the first author's name, and places the annotation as an indented paragraph below the citation. All three order entries alphabetically by author surname with hanging indents.
How is an annotated bibliography different from a literature review?+
An annotated bibliography is a list — each entry summarizes and evaluates one source independently, with no argument connecting entries. A literature review is an argument — it groups, compares, and synthesizes sources to motivate a research question. The practical relationship: a good annotated bibliography is the raw material for a literature review. Each annotation's evaluate and relate sentences become rows in a synthesis matrix, which then become the thematic paragraphs of the review.
How many sources should an annotated bibliography include?+
Whatever the assignment specifies — 10 to 15 entries is the most common undergraduate range, 20 to 50 for graduate preliminary-exam or thesis-proposal bibliographies. Coverage matters more than count: instructors expect a mix of foundational sources, recent work (past five years), methodological sources, and at least one counter-argument source, spread across more than one research group and country.
Can I write annotations in the first person?+
For the relate move, yes — “I will use this study's effect size as my benchmark” is clearer and more honest than contorted passive constructions, and most style guides and writing centers explicitly permit it in reflective annotations. The summarize and evaluate moves conventionally stay in the third person. If your instructor bans first person entirely, recast the relate sentence around your project: “This trial supplies the benchmark effect size for the present study.”
Should annotations evaluate how often a source has been cited?+
It strengthens the evaluate move considerably. Citation count tells you whether the field has accepted the work; the geography of those citations tells you whether it is field-defining (cited across many countries and institution types) or niche (cited mostly by the authors' own network). A quick forward-citation check — Google Scholar's “Cited by” count plus a geographic citation map of the citing institutions — gives you one defensible evaluative sentence per source.
What order do entries go in?+
Alphabetical by the first author's surname in all three major styles. Some assignments instead ask for thematic grouping (entries clustered under topic headings, alphabetical within each group) — this is common when the bibliography is explicitly a stepping-stone to a literature review. Never order by date or by when you read the source unless the assignment says so.
Keep going
- Literature review examples — five worked cases showing what your annotations become.
- How to write a literature review — the step-by-step synthesis process.
- Literature review outline — section skeletons by format and field.