What is named entity recognition?
Quick answer
Named entity recognition (NER) is the AI task of finding and labeling specific things in text: people, companies, dates, amounts, places, product names. It turns a messy email or transcript into structured pieces, such as owner, due date and amount, that can fill a table, a CRM record or a redaction list.
Last updated
Take the free module →Module 3 of the free Intro to AI Automation course, about 9 minutes. No account needed.
Updated · By Robert Breen
Why it matters for a small business
NER used to need a dedicated model trained for each kind of entity. Language models now do it from a plain request: "list every person and the task they own, with a date". That makes everyday extraction jobs, like action items, contract parties or invoice details, quick to set up.
The risk is invented entities. If a date isn't in the text, a model may supply a plausible one. Tell it to write "Not stated" instead, and check the extracted values against the source.
Entity extraction also feeds privacy work. Before you paste a document into an AI tool, a quick pass that lists every person, account number and address shows you what to remove or replace.
In a real lesson: Summarize a Meeting Transcript with ChatGPT
In Summarize a Meeting Transcript with ChatGPT, the Harbor Street Bakery Monday meeting goes in, and the request asks for an action items table with task, owner and due date, plus one rule: only use what is in the transcript, and write "Not stated" when an owner or date is missing.
Pulling owners and dates out of rambling talk is entity recognition. Two due dates come back Not stated because nobody said them out loud; you add them in one line and the table updates.
Try this lesson free or read the step-by-step guide.
Common confusions
NER vs keyword extraction
Keyword extraction finds important words or topics. NER finds specific named things and labels their type, such as person or date.
NER vs PII redaction
Redaction removes personal details. NER is often the step that finds them first.
Tips
- Name the exact entity types you want: owner, date, amount.
- Require "Not stated" for anything missing.
- Check extracted names and numbers against the source.
Related terms
More AI basics terms
Where you use it: free lessons
- Summarize a Meeting Transcript with ChatGPT (ChatGPT, 9 min)
- Turn a Client Intake Call into Next Steps and a Conflict-Check List (Law Firms) (ChatGPT, 9 min)
- AI Receipt Extractor: Receipts to Google Sheets with n8n (n8n, 12 min)
Frequently asked questions
- Can ChatGPT do named entity recognition?
- Yes. Ask it to list the people, companies, dates or amounts in a text, ideally as a table, and check the results.
- What is NER used for at work?
- Action items from meetings, parties and dates from contracts, vendor and total from invoices, and finding personal details to redact.
- Is NER accurate?
- On clear text it is usually good. It struggles with unusual names, abbreviations and missing details, so check the results that matter.