Airecmark Logo
info DEMO DATA: scores shown are template sample values pending independent evaluation · as of 2026-09-05 · Methodology
INDEX / AI_RESEARCH / KNOWLEDGE_SYNTHESIS / NOTEBOOKLM v2026.2
EVAL_RUN: 0x4f88..d92e
data_object JSON_TELEMETRY
Google Labs PRIMARY SPEC VERIFIED Mountain View, CA • Released Dec 2023 • v2026.2.4

NotebookLM

verified

Personal Knowledge Agent & Conversational Audio Overview Engine

memory GEMINI 1.5 PRO 2M CONTEXT lock ZERO CLOUD LEAKAGE graphic_eq AUDIO OVERVIEW ENGINE anchor MULTIMODAL GROUNDING
COMPOSITE AIRECMARK SCORE INDEX 2026.Q1
90.7 /100
30d Velocity
Category Rank #1 of 42

Grounded Knowledge Hubs, Retrieval Engines & Synthesis

HALLUCINATION 0.3% (MIN)
LATENCY AVG 1.42s TTFT

Deterministic Evaluation Telemetry

CONTEXT WINDOW CEILING all_inclusive
2,000,000
Tokens in Active Cache

Native Gemini 1.5 Pro massive multi-file context cache with instant cross-document needle retrieval.

AUDIO OVERVIEW REALISM graphic_eq
98.6%

Two-host conversational podcast generation with natural vocal inflections, backchanneling, and interjections.

GROUNDING INTEGRITY verified_user
99.7%
Source Attributable

Strict RAG locked exclusively to user-uploaded notebook sources. Null responses preferred over fiction.

MULTI-FORMAT INGESTION attachment
50+
Sources Per Notebook

PDFs, Google Docs, Slides, Web URLs, native YouTube spoken transcripts, and MP3/WAV audio uploads.

EMPIRICAL BENCHMARKS

Macro Intelligence Breakdown

Deterministic evaluations conducted against 1,400 multi-source cross-reference test suites.

description Massive Multi-Document Synthesis (500k+ words)
98.2%
format_image_left Source-Grounded Q&A (Zero Extraneous Hallucination)
99.4%
podcasts Audio Overview (Podcast Deep Dive) Cohesion
97.8%
format_list_bulleted Timeline & Study Guide Generation
94.5%
link Cross-Source Citation Interactivity
96.0%
verified Grounding methodology verified via Airecmark Source Integrity Protocol (ASIP-1).
RUN-ID: #892-NLM
VOICE SYNTHESIS TELEMETRY
DUAL-HOST_NEURAL
graphic_eq
Deep Dive: Quantum Topography
Synthesized Audio Overview • 11m 42s
HOST 1 (FEMALE_NATIVE) DYNAMIC INTERRUPTIONS (240ms) HOST 2 (MALE_NATIVE)
CONVERSATIONAL LATENCY 2.4 min (Generation)
PROSODIC VARIATION 99.1% Inflection
ANALOGY SYNTHESIS High (Contextual)
AUDIO BITRATE 128 kbps Opus
SYSTEM ARCHITECTURE & CAPABILITIES

Quantitative Technical Specification

memory

Gemini 1.5 Pro 2M Token Engine

ACTIVE CONTEXT CACHE INGESTION

Harnesses Google's frontier needle-in-a-haystack retrieval architecture, allowing whole textbooks, corporate charters, and technical codebases to reside simultaneously in active context without chunking loss or vector degradation.

2,000,000 Tokens Zero Lost-in-the-Middle Native TPU Execution
mic_external_on

Audio Overview Generative Engine

NEURAL MULTI-SPEAKER CONVERSATION

Transforms dense dry research documents into lively, engaging two-host conversational podcast audio discussions. Accurately synthesizes nuance, humor, interjections, counterarguments, and simplifying metaphors based purely on supplied sources.

Dual Neural TTS Dynamic Host Dialogue Audio Export Ready
verified_user

Source-First Grounding Perimeter

ZERO EXTRANEOUS KNOWLEDGE HALLUCINATION

Unlike general web LLMs, NotebookLM answers queries exclusively from uploaded reference materials. When data is absent from the sources, it explicitly halts and reports omissions rather than guessing or fabricating claims.

Clickable In-Text Citations Source Anchor Verification Strict Boundary Guard
input

Multimodal Ingestion Pipeline

UNIFIED CROSS-MEDIA VECTOR REPOSITORY

Supports direct YouTube video URL ingestion (extracting and indexing full spoken transcripts), raw audio files (MP3/WAV), Google Drive integration, PDFs, markdown, and public web URLs into a unified searchable context index.

YouTube Transcripts Raw Audio Transcripts Google Drive Native
COMPARATIVE TELEMETRY

Direct Peer Matrix

Side-by-side technical evaluation against industry frontier knowledge engines.

UPDATED 2026.02.14 • EVAL VER 2.4
Tool / Platform Airecmark Score Context Ceiling Audio Synthesis Grounding Rigor Commercial Cost Action
N
Google NotebookLM SUBJECT UNDER EVAL
93.6
2,000,000 tokens check_circle Two-Host Native $0 (Google Account)
P
Perplexity Spaces Perplexity AI
95.4
128,000 tokens None (Text Only) 92.4% (Web + Docs) $20 / mo (Pro)
C
Claude Projects Anthropic
94.2
200,000 tokens None (Artifacts Focus) 96.8% (Artifact RAG) $20 / mo (Pro)
G
ChatGPT Projects OpenAI
93.0
128,000 tokens Voice Mode (1-on-1) 88.1% (Web Hybrid) $20 / mo (Plus)
H
Heptabase Hepta Labs
89.8
Spatial Visual Whiteboard None 94.0% (Card Map) $11.99 / mo
COMMERCIAL AVAILABILITY & INTEGRATION

Access Tiers & Pricing

STANDARD ACCESS POPULAR

Google Account Free

$0 / forever
  • check Up to 50 sources per individual notebook
  • check 500,000 words per source ingestion ceiling
  • check Full Audio Overview (Podcast) generation
  • check Gemini 1.5 Pro multimodal processing
ENTERPRISE GRADE
ORGANIZATIONS

Google Workspace / Gemini

Included in Workspace
  • check Enhanced corporate data perimeter guarantee
  • check Zero model training on proprietary enterprise files
  • check Admin console audit trails & access policy
  • check HIPAA & SOC2 compliance coverage
ACADEMIC & RESEARCH PILOT

NotebookLM Business / Edu

$0 - Pilot / institution
  • check Higher limits on concurrent audio generations
  • check Classroom and institutional notebook enclaves
  • check Curated study guide templates & FAQ exports
  • check Dedicated Google Labs feedback channel
rate_review ANALYST CONSENSUS VERDICT • AIRECMARK RESEARCH GROUP
“NotebookLM represents Google Labs' greatest AI product achievement to date—a masterclass in strict grounded retrieval that completely eliminates hallucination by locking Gemini 1.5 Pro to user-supplied documents. The conversational Audio Overview engine is not merely a novelty; it constitutes an entirely new paradigm for complex knowledge digestion.”
CHIEF EVALUATOR: DR. A. KAUFMAN • PRINCIPAL LLM ARCHITECT
DETERMINISTIC VERDICT: EXCEPTIONAL (93.6/100)
Cryptographic Merkle Audit Seal SHA-256 SIGNED

ROOT_HASH: 7a88b49e102f901193bdca781938feaa124b893e110c732b210fca448e8992da

download Export Cert
Analyst Trade-off Summary

Strengths & Limitations

thumb_upStrengths
  • check_circle严格源约束
report_problemLimitations
  • error_outline仅能基于用户导入的来源作答,无实时检索能力
  • error_outline复杂 PDF 与特殊格式解析偶有失败
  • error_outline输出结构可控性弱,长文整理需多次引导