Annotated Bibs the Right Way | AI for Real Life

The annotated bibliography shortcut to avoid.

Annotated bibs are formulaic: citation, then a paragraph or two summarizing and evaluating each source. Because the format is so regular, they're a strong temptation: paste your source list, ask Claude to write the annotations, paste back into the doc, submit. Don't do this.

What an annotated bib is actually for

Professors don't assign annotated bibs because they want a list of sources. They assign them because they want proof that you read each source and can evaluate whether it belongs in your paper. The annotation is a compressed argument: "I read this. Here's what it says. Here's why it's useful (or isn't). Here's how it fits with the others." This can't be outsourced to AI.

The four moves in a good annotation.

Most professors want some version of these four. Read your assignment sheet for the exact format and word count, but the structure is reliable across disciplines.

📖

1. Summary (what does the source say?)

2–4 sentences. The author's main argument or finding, in your words. Not a description of what the article is "about":a statement of what it argues. "Smith argues that…" not "This article discusses…"

🔍

2. Evaluation (is it credible / well-argued?)

1–3 sentences. Author's credentials, the methodology, the journal, the bias if any. Are the sources cited? Is the argument supported? What does this source do well, and what's its weakness?

🧩

3. Relevance (why does it belong in YOUR paper?)

1–2 sentences. The link to your specific topic and argument. "This is useful for my paper because…" If you can't write this sentence, the source doesn't belong on your list.

🪢

4. Connection (how does it talk to your other sources?)

1–2 sentences. Often optional, sometimes required. "Smith's framing complicates the assumptions in Jones (2019)…" This one is the early move toward the synthesis matrix in Lesson 5.4.

The honest workflow: Claude as your reading partner, not your annotator.

The trick: Claude reads the source with you (you upload it, you skim it, Claude helps you compress what you understood). You then write the annotation in your own voice, drawing on the conversation. The annotation is yours; the reading process used a tool.

Step 1 · Speed-read the source (the Lesson 5.3 move)

I'm reading [paper title, author, year] for my annotated bibliography in [class]. I've uploaded the PDF. I have about [10 / 20 / 30] minutes. Give me a structural map of this paper: 1. The main argument or claim, in 1–2 sentences. 2. The 3–4 most important pieces of evidence the author uses. 3. The methodology (if it's empirical) or the theoretical framework (if it's a humanities piece). 4. The 1–2 paragraphs I MUST read in full to actually understand the argument (give me the page numbers). 5. The author's main weakness or limitation that a critical reader would push on. Don't summarize the whole thing. The point is to tell me where to look so I can read those parts myself and write my own annotation.

Step 2 · Test your understanding (BEFORE writing)

I just read [paper title]. Test whether I actually understood it before I write my annotation. Ask me: 1. What is the author's main argument? (One sentence in my own words.) 2. What's their strongest piece of evidence? 3. What's a weakness in the argument that I noticed? 4. What does this source do that the others I've read DON'T do? 5. Why is it relevant to MY paper on [your specific topic]? Wait for each answer. Don't tell me if I'm right until after the last question: then give me a debrief on which answers were on target and which suggest I should re-read a section. Don't do this in note-taking mode. Do it in flashcard / quiz mode. The point is to find out what I retained, not to feed me the summary.

Step 3 · Critique YOUR annotation draft (the Edit-Don't-Write move from M4.5)

Here's my draft annotation for [source]. Don't rewrite it. Critique it. Annotation draft: [paste your draft, in your own words] Source: [author, title, year] Class context: [name + level] My paper's topic / argument: [1 sentence] Look for: 1. Does my summary actually capture the author's argument, or am I describing what the paper is "about"? 2. Is my evaluation specific (methodology, evidence, credibility), or is it generic ("the author makes a strong argument")? 3. Does my relevance sentence connect to MY specific paper, or could it be cut-and-pasted into anyone else's bib? 4. Anything factually wrong about the source? (Author's claim, year, methodology, conclusion.) 5. Voice check: does this sound like me, or does it sound like the AI-flat default? Tell me what to fix. Don't show me the fixed version. I'll edit it myself.

The two traps that catch students on annotated bibs.

🪤

Trap 1: the auto-annotation paste

"Here are 10 sources. Write me a 150-word annotation for each." What you get: 10 paragraphs of well-formatted, half-correct prose with the AI's hedging-everything voice ("the author's compelling argument," "this source provides valuable insights"). The voice is the tell. Graders don't have to run a detector: annotations sound like AI annotations. And there's a non-zero chance Claude has misread the source, attributed an argument to the wrong author, or invented a methodology that isn't there.

🚨

Trap 2: the Citation Trap on the source list itself

If you let Claude help you find sources before you read them, double-check every entry against the workflow in Lesson 5.2. The asymmetry from Lesson 5.1 holds: 9 real sources and 1 invented one in your bibliography looks the same as 10 real ones until your prof clicks the link. An annotated bib with an invented source is the worst-case version of the Citation Trap because every entry is supposed to have been read.

The "did I actually read this?" gut-check

Before submitting an annotated bib, look at each entry and ask: could I answer a quick question about this source from a TA right now? If you couldn't: that's the entry that needs more actual reading before submission. The annotation should be the result of reading, not a substitute for it.

Where this fits in the course.

📚 Lesson 5.2: source-checking

Before any source goes in your bib, run it through 5.2's source-checking workflow. The annotation is built on the source actually existing and saying what you think it says: both have to be verified.

📖 Lesson 5.3: speed-reading PDFs

The "structural map" prompt (Step 1 above) is the same one from 5.3. The annotation workflow is the same speed-reading + verifying process, with one extra writing step on the end.

🔬 Lesson 5.4: synthesis

An annotated bib is the early stage of the synthesis matrix in 5.4. The "connection" sentence in each annotation is what you'll later expand into agreements, disagreements, gaps, and method tensions across sources.

✏️ Lesson 4.5: Edit-Don't-Write

Step 3 of the workflow above is just M4.5's edit-don't-write critique applied to your annotation draft. Same rule: you write it, Claude critiques, you fix.

Honest Work Code · Rule 1, applied to annotations

You stay the author. The whole point of the annotation is to prove that you read the source and that you can place it in your argument. AI in the workflow as a reading partner, a quiz partner, and an editing partner is fine. AI writing the annotation for you isn't: it's the part the assignment was checking for in the first place.

Back to the curriculum

This page lives in the bonuses library. Pull it out when an annotated bib gets assigned. Most students hit one of these per term in some research-methods or upper-division class.

Back to the curriculum →