Glossary · Updated November 2026

AI meeting note taker glossary

Answer in brief

AI note takers turn speech into text (transcription), work out who spoke when (speaker diarization) and use a language model to write a summary with action items. They either join the call as a meeting bot or record on your device without one.

Around that sit the privacy and security terms you meet when buying: a DPA is the processing contract under the GDPR, a sub-processor is a provider the vendor uses, data residency says where data is stored, and SOC 2, ISO/IEC 27001 and a HIPAA BAA are security reports, certificates and contracts. The 25 definitions below are factual and short.

This comparison is published by meetergo GmbH, a German software company that sells an AI note taker and is one of the tools described here. Every fact about other vendors comes from their official websites as of 9 November 2026. We did not run long-term tests; “pick” is our editorial judgment for the use case named. How this guide is made.

25 terms, from capture to compliance

Grouped from capture to compliance. Legal terms link to the official text; vendor examples come from the sources in our tool directory.

Meeting bot: A software participant that joins a video call on Zoom, Teams, Google Meet or similar to record and transcribe it. It appears in the participant list and can record meetings you do not attend.
Bot-free recording: Recording from your own device instead of a bot in the call, through a desktop app, a browser extension or a phone app. See note takers without a bot.
System audio capture: Recording the sound your computer plays, together with your microphone. This is how desktop note takers record calls without joining them.
Local transcription: Speech-to-text that runs on your own computer, so audio does not leave the device. meetergo's local mode and Notta's on-device transcription are examples.
Transcription: Converting spoken words into text, also called speech-to-text. The transcript is the raw material for summaries and search.
Speaker diarization: Splitting a recording into segments by speaker, answering “who spoke when”. Naming the speakers is a separate step, often using the calendar or participant list.
Word error rate (WER): A common measure of transcription accuracy: the number of substituted, deleted and inserted words, divided by the number of words in a correct reference transcript. Lower is better. Results depend heavily on audio quality, accents and vocabulary.
Large language model (LLM): An AI model trained on large amounts of text that can write and summarise. Note takers send the transcript to an LLM to produce summaries and answers.
Summary: A short written account of the meeting generated from the transcript, usually with topics, decisions and open questions. Many tools let you choose a template per meeting type.
Action items: Tasks extracted from the meeting, ideally with an owner and a due date. Some tools push them into a task or CRM system.
CRM sync: Automatically writing meeting notes, summaries or field updates into a customer relationship management system such as HubSpot or Salesforce, linked to the right contact or deal.
MCP (Model Context Protocol): An open-source standard for connecting AI applications to external systems. A note taker with an MCP server lets assistants such as Claude or ChatGPT read your meeting notes.
Consent to record: Permission from the people being recorded. Rules differ by country; in Germany, recording privately spoken words without authorisation is an offence under section 201 of the Criminal Code.
Controller and processor: Under Article 4 GDPR, the controller decides the purposes and means of processing; the processor processes personal data on the controller's behalf. You are usually the controller, the note taker vendor the processor.
DPA (data processing agreement): The contract between controller and processor required by Article 28(3) GDPR. It sets out the subject matter, duration, nature and purpose of processing, the type of data and the parties' rights and obligations.
Sub-processor: Another processor that your vendor engages, for example a cloud host, a transcription service or an LLM provider. Under Article 28(2) and (4) GDPR this needs your authorisation and the same data protection obligations.
Data residency: The country or region where data is stored and processed. A vendor can store data in one region and process it with providers in another, so check both. See GDPR and EU hosting.
Standard contractual clauses (SCCs): Data protection clauses adopted by the European Commission that can serve as an appropriate safeguard for transfers outside the EU under Article 46 GDPR.
EU-US Data Privacy Framework: The basis of Commission Implementing Decision (EU) 2023/1795, which finds an adequate level of protection for transfers to US organisations on the Data Privacy Framework List.
Zero data retention: A provider commitment not to keep the data it processes after returning the result. meetergo states this for its EU AI providers Mistral and melious.
Model training on customer data: Using your recordings or transcripts to improve AI models. Vendors differ: some never do it, some train their own models on de-identified data unless you opt out.
Retention period: How long recordings, transcripts and notes are kept before deletion. Article 5(1)(e) GDPR asks that personal data is kept no longer than necessary.
SOC 2: A report by an independent CPA firm, under AICPA standards, on a service organisation's controls relevant to security, availability, processing integrity, confidentiality or privacy. Ask which report type and which period it covers.
ISO/IEC 27001: The international standard that defines requirements for an information security management system (ISMS). A certificate shows that an organisation has such a system; check whether it covers the vendor itself or only its data centre.
HIPAA BAA: A business associate agreement under the US HIPAA rules: the written contract through which a covered entity obtains assurance that a vendor handling protected health information will safeguard it.

Questions and answers

What is the difference between transcription and a summary?

Transcription is the full text of what was said. A summary is a short version written by a language model from that transcript, usually with decisions and action items.

What is the difference between a DPA and a sub-processor list?

The DPA is your contract with the vendor. The sub-processor list names the other companies the vendor uses to process your data, such as hosting, transcription and LLM providers.

Is SOC 2 a GDPR certificate?

No. SOC 2 is a US audit report on security controls. It can support a vendor's security claims but says nothing about whether your recording has a legal basis under the GDPR.

Why does word error rate matter?

Every wrong word in the transcript can end up in the summary, especially names and numbers. Vendors rarely publish comparable figures, so test with your own meetings.

Continue reading

Sources

All pages accessed 9 November 2026. Definitions are simplified for readability and are not legal advice.

Cite this entry

Best AI Meeting Note Taker. “AI meeting note taker glossary”. meetergo GmbH. Content reviewed 9 November 2026. https://bestaimeetingnotetaker.com/glossary/