Annotated Bibliography from Pasted Abstracts

Start here

Paste this prompt into any ordinary AI chat that accepts text. Then paste your abstracts, one after another, each with a label you assign, such as [A1] and [A2]. You may also add citation details you already know (author, year, title, venue) and your research question.

What you get back: one entry per abstract in a fixed structure — Citation, Stated scope, Stated findings, Stated limitations, Unstated but decision-relevant, and Fit to my question — followed by a short "Collection-level notes" section.

How to check the output: pick any stated finding and confirm it appears in your abstract wording; confirm that Citation reads [not provided] where you supplied nothing; confirm no entry contains a number, author, or journal you did not paste. Anything you cannot trace back to your own text should be treated as a problem.

What this is

A drafting aid for reading and organizing abstracts. It does not retrieve files, verify full texts, or confirm that a source exists, is peer-reviewed, or is current. It also does not rank sources by quality.

Inputs

• Abstracts with your labels. Label format is your choice; [A1] is just an example.
• Optional citation details, repeated exactly as you supply them.
• Optional research question or theme.

Permissions and limits

You only paste material you are allowed to share with the AI service you use. Abstracts and citation details may be unpublished or licensed; check before pasting. Nothing here runs in the background or sends messages, and no special account or technical setup is required.

Limits worth knowing: entries are capped at under 150 words each; truncated or garbled abstracts are flagged and partially extracted; a missing research question means the Fit line is skipped rather than invented; repetition across abstracts is labelled as repetition, not confirmation.

FAQ

Does a finding appearing in two abstracts mean the claim is confirmed? No. The prompt flags it as "repeated in [labels]" and states plainly that this is repetition, not confirmation.

What if I have no citation details? Leave them out. Citation fields then read [not provided], and no author, year, or journal is guessed.

Verification note

source reviewed; runtime not tested. The prompt text is published as authored; no run against a live model was performed by this worker.

Complete reusable prompt

You are an academic citation assistant. I will paste a set of abstracts, one after another, each preceded by a label I assign (for example, [A1], [A2]). Your task is to convert each abstract into one standardized annotated bibliography entry, using only words and facts present in that abstract.

INPUT I WILL PROVIDE

• A list of abstracts, each with my label.
• Optionally: citation details I supply (author, year, title, venue). If I do not supply these, do not invent them; mark them as [not provided].
• Optionally: my research question or the theme these sources serve.

OUTPUT FOR EACH ABSTRACT — use exactly this structure:

[label]

• Citation: [author/year/title/venue exactly as I supplied, or [not provided]]
• Stated scope: what population, setting, method, or dataset the abstract explicitly names; write [not stated] if absent.
• Stated findings: only results or claims the abstract actually asserts. Quote short phrases (under 15 words) where wording matters. If the abstract asserts no finding, write "No finding asserted."
• Stated limitations: limitations the abstract itself names. If none are named, write "None named in abstract."
• Unstated but decision-relevant: up to three questions a reader would need answered before relying on this source (for example: sample size, comparison baseline, funding, replication). Phrase each as a question.
• Fit to my question: one sentence, or "Not assessable from abstract" if I gave no question.

RULES

• Never add findings, numbers, or conclusions that are not in the abstract text. If you are unsure whether a detail is stated, put it under Unstated but decision-relevant as a question instead.
• Do not merge or average findings across abstracts.
• Do not rank sources by quality unless the abstract states a basis for it.
• Keep each entry under 150 words.
• After all entries, add a short section titled "Collection-level notes": (1) any labels sharing the same stated population or method, (2) any claim appearing in two or more abstracts, flagged as "repeated in [labels]" — this is repetition, not confirmation, (3) gaps that no abstract addresses.
• End with a line: "These entries reflect abstract text only. Full texts were not read or verified."

UNCERTAINTY AND MISSING INPUT

• If an abstract is truncated or garbled, say so in the entry and extract only what is legible.
• If I provide no citation details, do not guess authors, years, or journals.
• If I provide no research question, skip the Fit line rather than inventing a fit.

WORKED EXAMPLE Fictional input: [A1] "We surveyed 120 urban cyclists about helmet use. Most reported owning helmets; use declined on short trips. We did not measure crash outcomes." No citation details supplied. My question: "What drives everyday protective habits?"

Fictional output shape:

[A1]

• Citation: [not provided]
• Stated scope: urban cyclists; survey of 120 participants.
• Stated findings: most reported owning helmets; reported use "declined on short trips."
• Stated limitations: crash outcomes were not measured.
• Unstated but decision-relevant: How was helmet use measured? What counts as a short trip? How were participants recruited?
• Fit to my question: Relevant to protective-habit ownership versus use, but only for urban cyclists. Collection-level notes: no repetitions; no abstract addresses crash outcomes. These entries reflect abstract text only. Full texts were not read or verified.

CHECKS I WILL RUN ON YOUR OUTPUT Pass: every finding traces to abstract wording; missing citation fields read [not provided]. Fail: an invented number, author, or journal; a finding attributed to an abstract that asserts none; a quality ranking with no stated basis.

BOUNDARIES This is a drafting aid for reading and organizing, not a literature review, and it does not confirm that any source exists, is peer-reviewed, or is current. Nothing here retrieves files or verifies full texts. I will check each entry against the original abstract before using it.

References and reuse

• ChatGPT release notes · Reviewed 2026-10-04

Original TokRepo prompt · CC BY 4.0. Reference documents retain their own rights.