Data Extraction & Structuring prompts

Pulls contact information from unstructured text (emails, bios, directories) into a ready-to-use table.

Read the following block of text and extract every person's name, email address, phone number, and company into a clean markdown table with the columns: Name | Email | Phone | Company. If a field is missing for a person, write N/A. Do not add any information that is not explicitly present in the text.

[PASTE TEXT HERE]

How to use: Paste any messy contact list, email thread, or directory excerpt into the placeholder.

Turns unstructured invoice text into a structured JSON object suitable for accounting or API ingestion.

You are a data extraction assistant. Parse the following raw invoice text and return a single valid JSON object with these fields: invoice_number, invoice_date, due_date, vendor_name, vendor_address, line_items (array of objects with description, quantity, unit_price, total), subtotal, tax, and total_amount. Use null for any field you cannot find. Output only the JSON, no explanation.

[PASTE INVOICE TEXT HERE]

How to use: Paste the raw text copied from a PDF invoice or email into the placeholder.

Extracts key fields from a job listing to enable comparison, database ingestion, or analysis.

Extract structured data from the job posting below. Return a JSON object with the following keys: job_title, company_name, location, employment_type, salary_range, required_skills (array), preferred_skills (array), years_of_experience, education_requirement, and application_deadline. If a field is not mentioned, set its value to null.

[PASTE JOB POSTING HERE]

How to use: Copy and paste any job posting text, whether from a website, PDF, or email.

Strips marketing language from product copy and surfaces only the technical specs in a clean table.

Read the product description below and extract all technical specifications. Return the results as a two-column markdown table: Specification | Value. Only include factual, measurable attributes (e.g., dimensions, weight, material, capacity, voltage). Ignore promotional language. If units are mentioned, include them in the Value column.

[PASTE PRODUCT DESCRIPTION HERE]

How to use: Paste any product page copy, catalog entry, or datasheet text.

Builds a structured, sortable timeline from narrative or historical text.

You are a timeline extraction engine. Read the text below and identify every discrete event that has an associated date or time period. Return a JSON array where each element is an object with: date (ISO 8601 format if possible, otherwise the original text), event_summary (one sentence), and people_involved (array of names, empty array if none). Sort the array chronologically. Output only the JSON array.

[PASTE TEXT HERE]

How to use: Works well on news articles, historical summaries, project post-mortems, or legal documents.

Cleans and normalizes a messy list of addresses into a consistent, parseable structure.

The list below contains addresses written in inconsistent formats. Normalize each one and return a JSON array. Each element should have: street_number, street_name, unit (null if absent), city, state_or_province, postal_code, country. Standardize state/province to the official two-letter abbreviation where applicable. Do not infer or guess missing components—use null.

[PASTE ADDRESS LIST HERE]

How to use: Paste any raw address list; works great as a pre-processing step before importing into a CRM or database.

Classifies all named entities in a document for downstream NLP tasks, search indexing, or analysis.

Perform named entity recognition on the text below. Identify and classify every entity into one of these categories: PERSON, ORGANIZATION, LOCATION, DATE, MONEY, PRODUCT, or OTHER. Return a JSON array of objects, each with: entity_text, category, and start_position (approximate character index in the original text). Remove duplicates—if the same entity appears multiple times, include it once with the first occurrence position.

[PASTE TEXT HERE]

How to use: Useful on news articles, contracts, research papers, or any information-dense text.

Transforms raw meeting notes into a trackable action-item table for project management.

Read the meeting notes below and extract every action item. Return a markdown table with columns: Action Item | Owner | Due Date | Priority (High / Medium / Low / Unknown). Infer priority only if it is clearly stated or strongly implied in the text. If no due date is mentioned, write 'Not specified'. List items in order of appearance.

[PASTE MEETING NOTES HERE]

How to use: Paste typed or transcribed meeting notes; works well after generating a transcript summary.

Converts a plain-text resume into a structured JSON profile for ATS systems or candidate databases.

Parse the resume text below into a structured JSON object with the following top-level keys: full_name, email, phone, location, linkedin_url, summary, work_experience (array of objects: company, title, start_date, end_date, responsibilities as array of strings), education (array of objects: institution, degree, field_of_study, graduation_year), skills (array of strings), and certifications (array of strings). Use null for missing fields and empty arrays where no items are found.

[PASTE RESUME TEXT HERE]

How to use: Paste the full text of a resume copied from a PDF or document file.

Pulls all financial figures from earnings reports, press releases, or articles into a verifiable table.

Read the financial text or report excerpt below and extract every monetary figure mentioned. Return a markdown table with columns: Metric | Value | Currency | Period | Source Sentence. In the Source Sentence column, quote the exact phrase from the text that contains the figure. Do not round or alter any numbers.

[PASTE FINANCIAL TEXT HERE]

How to use: Ideal for earnings call transcripts, annual report excerpts, or financial news articles.

Condenses multiple product reviews into a side-by-side comparison table for quick decision-making.

Below are [NUMBER] product reviews for different [PRODUCT CATEGORY] products. Extract the product name and the following attributes mentioned in each review: price (if stated), key strengths (up to 3 bullet points), key weaknesses (up to 3 bullet points), and overall sentiment (Positive / Neutral / Negative). Present the results as a single markdown table with one row per product.

[PASTE REVIEWS HERE]

How to use: Fill in the number of reviews and product category; paste the review texts sequentially.

Structures contract clauses for legal review, compliance tracking, or contract management systems.

You are a legal data extraction assistant. Read the contract excerpt below and identify each distinct clause. For every clause, return a JSON object within an array containing: clause_number (or null if unnumbered), clause_title, clause_type (e.g., Indemnification, Termination, Confidentiality, Governing Law, Payment, Other), party_obligations (array of plain-English obligation summaries), and conditions (array of triggering conditions, empty if none). Output only the JSON array.

[PASTE CONTRACT EXCERPT HERE]

How to use: Paste one or more sections from a contract; particularly useful for NDAs, SaaS agreements, or employment contracts.

Turns a descriptive paragraph into a single CSV row based on user-defined columns for bulk data entry.

I will give you a paragraph of text and a list of column headers. Extract the value for each column from the text and return a single CSV row (comma-separated, values quoted if they contain commas). If a value is not found in the text, output an empty quoted string for that column. On the first line output the headers, on the second line output the data.

Column headers: [COMMA-SEPARATED COLUMN NAMES]

Text: [PASTE PARAGRAPH HERE]

How to use: Define the columns that match your spreadsheet or database schema before pasting the source text.

Codes qualitative survey responses into a structured dataset ready for quantitative analysis.

Below are open-ended survey responses to the question: '[SURVEY QUESTION]'. For each response, extract: response_id (sequential number starting at 1), verbatim_text, primary_theme (one short label), secondary_themes (array of additional labels, empty if none), sentiment (Positive / Neutral / Negative / Mixed), and contains_suggestion (true/false). Return a JSON array sorted by response_id.

[PASTE SURVEY RESPONSES HERE — ONE PER LINE OR SEPARATED BY '---']

How to use: Fill in the survey question for context; separate responses clearly with line breaks or '---'.

Converts recipe prose into a machine-readable JSON structure for apps, databases, or meal planners.

Extract the structured data from the recipe text below and return a JSON object with: recipe_name, servings, prep_time_minutes (integer or null), cook_time_minutes (integer or null), total_time_minutes (integer or null), ingredients (array of objects with quantity, unit, and ingredient_name), steps (array of strings in order), and tags (array of applicable labels such as 'vegan', 'gluten-free', 'quick' — only if clearly supported by the recipe text).

[PASTE RECIPE TEXT HERE]

How to use: Paste the full recipe text including ingredients and instructions.

Recovers table structure from plain-text exports of PDFs, emails, or scraped web pages.

The text below is a table that was exported from a document but lost its formatting. The original table had these columns: [COLUMN NAMES, COMMA-SEPARATED]. Reconstruct it as a properly formatted markdown table. If you encounter a row where data appears ambiguous, include it with a note in a final 'Notes' column. Do not invent or fill in missing values.

[PASTE GARBLED TABLE TEXT HERE]

How to use: Specify the expected column names based on your knowledge of the original document.

Pulls structured medication data from clinical notes for EMR population or medication reconciliation review.

You are a clinical data extraction assistant. Read the clinical note below and extract every medication mentioned. Return a JSON array where each element contains: medication_name, dosage (strength and unit), frequency, route_of_administration, indication (if stated), and status ('current', 'discontinued', 'as_needed', or 'unknown'). Do not interpret, infer diagnoses, or add any clinical judgment. Only extract what is explicitly stated.

[PASTE CLINICAL NOTE HERE]

How to use: Use only with de-identified or fictional clinical text; always have a qualified clinician verify output.

Structures glossary or definitional content for knowledge bases, training datasets, or documentation systems.

Read the glossary or definitional text below and extract each term and its definition. Return a JSON array of objects, each with: term and definition. Preserve the definition exactly as written. If a term has multiple definitions, create separate entries with the same term and each distinct definition. Sort the array alphabetically by term.

[PASTE GLOSSARY OR DEFINITIONAL TEXT HERE]

How to use: Works on formatted glossaries or inline definitions scattered through technical documents.

Extracts and categorizes competitor mentions from market reports or industry articles for competitive analysis.

You are a competitive intelligence analyst. Read the article or report below and identify every mention of a competitor company. For each mention, return a JSON array element with: competitor_name, mention_context (the full sentence containing the mention), sentiment_toward_competitor (Positive / Negative / Neutral), and attribute_discussed (e.g., pricing, product quality, market share — use 'general' if unclear). De-duplicate by keeping all unique contextual mentions even if the same competitor appears multiple times.

[PASTE ARTICLE OR REPORT HERE]

How to use: Paste industry news, analyst reports, or competitive blog posts to surface intelligence quickly.

Quickly tabulates metadata from academic abstracts or citation lists for literature reviews.

From the list of research paper abstracts or citations below, extract the following fields for each paper and present them in a markdown table: Title | Authors (semicolon-separated) | Year | Journal or Conference | Key Finding (one sentence) | Study Type (e.g., RCT, meta-analysis, case study, survey, review). If a field cannot be determined from the text, write N/A.

[PASTE ABSTRACTS OR CITATIONS HERE]

How to use: Paste multiple abstracts separated by blank lines or citation entries from a reference manager export.

Creates a decision log from an email thread to support audit trails, project tracking, or onboarding new team members.

Read the email thread below and extract every decision that was made, proposed, or deferred. Return a markdown table with columns: Decision | Made By | Date (if determinable) | Status (Agreed / Proposed / Deferred / Rejected) | Related Action Items. Base all entries strictly on the content of the emails. Do not infer decisions that were not explicitly stated.

[PASTE EMAIL THREAD HERE]

How to use: Copy the full email chain including sender names and dates for best results.

Converts unstructured event listings into structured calendar data for import or display.

Extract all events from the text below and return a JSON array. Each element should represent a single event with: event_name, description (one sentence summary), start_datetime (ISO 8601 if possible), end_datetime (ISO 8601 if possible, null if not stated), location (physical address or 'Online' or null), organizer (null if not stated), registration_url (null if not stated), and cost (null if not stated, 'Free' if explicitly free). Use the assumed timezone [TIMEZONE, e.g. 'America/New_York'] if none is specified.

[PASTE EVENT LISTINGS TEXT HERE]

How to use: Specify your local timezone in the placeholder; paste newsletter content, website text, or event guides.

Structures experimental data from scientific text for meta-analysis, research databases, or reproducibility review.

You are a scientific data curator. For each experiment described in the text below, extract the following and return a JSON array: experiment_id (sequential), hypothesis (quoted exactly if stated, otherwise null), independent_variable, dependent_variable, control_conditions (array), sample_size (null if not stated), key_result (one-sentence summary), and conclusion_supported (true / false / unclear). Separate multiple experiments if the text describes more than one.

[PASTE EXPERIMENT DESCRIPTION OR LAB REPORT TEXT HERE]

How to use: Works on lab reports, methods sections of papers, or science journalism summarizing experiments.

Uses few-shot examples to teach the model a custom extraction pattern for any domain-specific schema.

I need to extract specific fields from text snippets. Here are two examples showing the input and the desired JSON output:

Example 1:
Input: "[EXAMPLE INPUT 1]"
Output: [EXAMPLE JSON OUTPUT 1]

Example 2:
Input: "[EXAMPLE INPUT 2]"
Output: [EXAMPLE JSON OUTPUT 2]

Now apply the exact same extraction pattern to the following new input. Return only the JSON, no explanation.

New Input: "[PASTE NEW TEXT HERE]"

How to use: Fill in two representative input/output examples that demonstrate your extraction schema before pasting the real input.

Uses explicit chain-of-thought steps to improve extraction accuracy on complex or ambiguous source text.

I need you to extract structured data from the text below by following these steps explicitly:

Step 1 – Read the full text and list every entity of type [ENTITY TYPE, e.g. 'software bug', 'transaction', 'complaint'] you can identify, in plain bullet points.
Step 2 – For each entity, note any attributes mentioned: [ATTRIBUTE LIST, e.g. 'severity, date reported, affected module, status'].
Step 3 – Use your findings from Steps 1 and 2 to produce a final JSON array. Each element should have a key for each attribute; use null for missing values.

Output all three steps visibly so I can verify your reasoning.

[PASTE TEXT HERE]

How to use: Fill in the entity type and attribute list to match your use case; particularly useful when the text is dense or noisy.