When a NotebookLM import failed, do not immediately upload the same source again. First identify the source type and failure stage, then check support, size, extractable content, permissions, web or caption availability, account limits, and the device you are using.

Google renamed NotebookLM to Gemini Notebook in July 2026. This guide keeps the former name in its title so older searches and instructions remain recognizable; the current product name is Gemini Notebook.［5］

Google's source guide lists supported inputs and source-specific restrictions, including limits of up to 500,000 words or 200 MB per source, web-page extraction boundaries, Drive access requirements, and YouTube caption conditions.［1］ A repeat upload only helps after the underlying condition changes.

Key Takeaways

Preserve the original and record the exact error before converting anything.

Verify the real content type; a familiar filename extension is not proof of parseable content.

Diagnose PDFs, Drive files, websites, YouTube videos, and audio with different branches.

Check notebook and account limits before splitting or duplicating sources.

Import a small, permitted test source to separate a source defect from an account or service problem.

Which stage broke when your NotebookLM import failed?

Write down what happened instead of relying on “it did not work.” The next check depends on the last successful stage.

Stage	Observable symptom	First question
Selection	File or source cannot be chosen	Is this input type supported on this device?
Transfer	Upload stalls or immediately errors	Is the local file readable and within size limits?
Retrieval	Drive, web, or YouTube source cannot be fetched	Does NotebookLM have access to the current version?
Extraction	Source arrives empty or incomplete	Is there extractable text, captions, or intelligible audio?
Indexing	Source remains processing or fails later	Is the content unusually large, malformed, or temporarily unsupported?
Notebook limit	Add-source control rejects another item	Has the notebook or account reached its current source allowance?

Capture the account type, browser or app, source type, filename or URL, approximate size, notebook source count, time, and exact visible message. Do not include confidential content in a support screenshot.

Step 1: Confirm that the NotebookLM source type is supported

NotebookLM supports a defined set of source paths, including common document and text formats, PDFs, copied text, web URLs, public YouTube videos with captions, audio, and eligible Google Drive files. The current list can change, so use Google's source-import page rather than a third-party list.［1］

Check the file's actual type, not only its suffix. A file renamed from another format, a damaged archive, a shortcut, or a cloud placeholder may display a supported extension without containing readable data. Open it locally in the expected application. If it cannot be opened there, recover a valid source before returning to NotebookLM.

For an unsupported application format, export from the source application to a documented supported format. Preserve the original. Review the export for missing pages, comments, tables, formulas, and encoding changes before import.

Step 2: Check size, word count, and source allowance

Google states that each source can contain up to 500,000 words or be up to 200 MB, and its overview and limits documentation describe account-dependent notebook and source allowances.［1］［2］［3］ Check the current account page because limits may change or vary by plan.

Do not split blindly. First decide whether the source has meaningful independent sections. A useful split preserves headings, page references, dates, and source identity, such as policy-part-1-v3 and policy-part-2-v3. An arbitrary split can make citations harder to interpret.

Before adding a duplicate, check whether an earlier version already occupies a source slot. Remove or replace a source only when you understand the effect on notes, citations, and collaborators.

Step 3: Diagnose PDF extraction and protection

A PDF is a container, not a guarantee of usable text. Test these conditions locally:

Can you open every page without a password?
Can you select and copy a sentence from the body?
Does search find a visible word?
Are pages scans rather than a text layer?
Does the file contain malformed pages, attachments, or unusual security restrictions?
Are you authorized to process and upload the document?

Google notes that copy-protected PDFs can fail.［3］ Do not attempt to bypass access controls. Obtain an authorized accessible copy from the owner or export the document through the approved source application.

For a scan, create an OCR-derived accessible copy only when you have permission. Review names, dates, numbers, headings, footnotes, tables, and page order; OCR errors can become plausible but false NotebookLM answers. Keep the scan and OCR version as separate, labeled files.

If one PDF fails, test a small, non-sensitive, known-good PDF. If the test succeeds, keep diagnosing the original. If it also fails, move to the account, device, and service branch instead of converting every file.

Step 4: Check Google Drive ownership and access

For a Drive source, confirm you are signed into the intended Google account and can open the exact file there. A link visible in a browser tab does not prove that the NotebookLM account has permission.

Google says you need view access to import an eligible Drive file. Eligible Drive sources update automatically and sync every few minutes; a manual sync control may also be available. Comments and footnotes are not imported.［1］ If the latest content is missing, verify that the original file and access are still valid, allow the automatic sync to complete, and use the documented manual sync instead of adding ambiguous duplicate copies.

Check:

correct account and workspace;
file ownership and sharing scope;
organization restrictions on external or AI services;
link target is a file, not a folder or shortcut chain;
the file has finished uploading and opens normally;
expected content is in the document body, not only comments or unsupported elements.

Do not ask an owner to make a private file public merely to make import easier. Use an approved access path.

Step 5: Treat web URLs as extraction requests

NotebookLM does not import a web page exactly as your browser renders it. Google says web URL import extracts text and does not bring in page images, embedded videos, nested pages, or content behind paywalls.［1］

A URL may therefore fail or produce a thin source when:

login, consent, subscription, or paywall access is required;
the useful content loads only after scripts or interaction;
the URL points to a home page, index, redirect, or embedded viewer;
the article is mostly images, tables rendered as graphics, or video;
robots or site controls prevent retrieval;
the content changed or was removed.

Open the URL in a clean browser context you are authorized to use. Confirm it is the canonical article and that the relevant text is present without relying on your private session. Never copy paywalled or restricted material into another format to evade access controls.

If you own the content, export a clean document or paste only the permitted text, then label the source with its URL and retrieval date. For research method after a successful import, use the NotebookLM research workflow; it is separate from this intake diagnosis.

Step 6: Verify YouTube caption eligibility

Google's source guide says NotebookLM can import public YouTube videos with captions, and that it imports the text transcript rather than the audiovisual content. Videos without speech are not supported, and recently uploaded videos may not yet be available for import.［1］

Check:

the video is public and opens while signed out;
captions or a transcript are available;
the spoken language is represented well enough for your task;
the video is not newly uploaded and still processing;
the URL points to the video, not a playlist, channel, clip, or embedded redirect;
you understand that visual-only facts will be absent.

Do not use a transcript source to claim that NotebookLM examined charts, demonstrations, on-screen labels, or visual evidence. If the video imports successfully, the Audio Overview verification guide and Video Overview coverage guide explain later output checks.

Step 7: Inspect audio quality and structure

For audio, confirm the format is on the current supported list and the file plays completely outside NotebookLM.［1］ Check duration, corruption, clipped speech, overlapping speakers, long silence, extreme background noise, and language support.

Create a clean authorized export rather than repeatedly transcoding a damaged file. If you trim silence or split a long recording, preserve chronological order and label every segment. Do not enhance audio in a way that invents words or obscures who said what.

Importing speech creates a text interpretation of audio; it does not prove speaker identity, consent, quotation accuracy, or factual correctness. Review important names, numbers, commitments, and quotations against the recording.

Step 8: Separate mobile limitations from source failures

The mobile app and desktop web experience may not expose identical source types or controls. Google's overview documents platform-specific behavior and recommends checking current availability.［2］ If a supported source cannot be selected on mobile, test the official desktop web interface with the same authorized account before changing the source.

This is a diagnostic comparison, not a workaround for policy. A managed device, browser extension, storage provider, or organization rule may intentionally restrict upload. Do not disable security controls to force the import.

Step 9: Run a controlled import test

Use a small, public or synthetic source with known content:

Title: NotebookLM import diagnostic
Version: 1
This source contains three facts:
1. The test identifier is ORANGE-27.
2. The review owner is Example Team.
3. The source date is 2026-09-06.

Save it in a currently supported plain format and import it into a notebook with available capacity. Then ask NotebookLM to identify the three facts and inspect the citations.

Interpret the result carefully:

Test succeeds, original fails: source-specific format, access, extraction, or size problem.
Both fail in one notebook: notebook limit, notebook state, or account restriction may be involved.
Both fail across notebooks: device, account, service availability, or broader platform issue is more likely.
Import succeeds but facts are wrong: ingestion completed, but content extraction or answer verification still failed.

The NotebookLM access guide covers platform reachability and account availability. Do not confuse that separate problem with a malformed source.

Step 10: Re-upload only after one condition changes

Record the change and result:

Attempt	Changed condition	Result	Next decision
1	Original source	Exact error	Choose branch
2	Authorized clean export	Success/failure	Validate content or continue
3	Small known-good source	Success/failure	Source vs account diagnosis
4	Desktop/current browser	Success/failure	Surface vs broader issue

Avoid repeated identical uploads. They create duplicate names, consume source allowances, and erase the evidence needed for troubleshooting.

If the failure persists, collect only safe metadata for support: account category, platform, browser/app version, source type, approximate size, time, reproducible steps, and redacted error. Do not send the sensitive source unless an approved support process explicitly requires it.

Summary

Record the exact source and failure stage.
Check current supported types, per-source limits, and notebook allowance.
Test PDF text extraction and protection without bypassing controls.
Confirm Drive access and allow automatic or manual synchronization to complete.
Treat web and YouTube imports as bounded text extraction.
Validate audio and use desktop to isolate mobile feature gaps.
Re-upload only after a documented condition changes.

Frequently asked questions

Why does a PDF open normally but fail in NotebookLM?

It may be password-protected, copy-protected, malformed, scan-only, too large, or contain no usable text layer. Test local selection and search, then obtain an authorized accessible copy.

Can I import a paywalled article by URL?

NotebookLM's web importer does not promise access to paywalled content. Do not bypass the publisher's controls; use material you are authorized to access and submit.

Why is a Drive document incomplete after import?

Google says comments and footnotes are not imported. Eligible Drive sources update automatically and sync every few minutes; if the current version is missing, verify the original file and access, allow that sync to complete, and use the available manual sync control when needed.

Why does a YouTube video fail to import?

Confirm it is public, has captions, contains speech, is not still newly processing, and uses a direct video URL. NotebookLM imports transcript text, not the visuals.

Should I split a large source into many files?

Only after checking current limits and identifying meaningful sections. Preserve order, headings, source identity, and citation context; arbitrary splits can make evidence harder to trace.

Does successful import mean every detail was understood?

No. Inspect extracted text and citations, then verify names, numbers, tables, quotations, and visual-dependent claims against the original source. NIST identifies confabulation, privacy, information integrity, and human-AI configuration as generative-AI risks, so successful ingestion does not replace source verification.［4］

Is the mobile app the best place to troubleshoot?

Not always. If a source type or control is missing, compare the current desktop web interface with the same permitted account before converting the file.

Will repeated uploads eventually fix a temporary error?

They may only create duplicates. Change one condition, record the result, and use a known-good source to distinguish source-specific failure from account or service behavior.

Disclaimer: This article provides general troubleshooting information. Follow copyright, privacy, organization, and platform rules for every source you submit.

Sources

Google, Add or discover new sources for your notebook — https://support.google.com/notebooklm/answer/16215270?co=GENIE.Platform%3DDesktop&hl=en-GB
Google, Get started with NotebookLM — https://support.google.com/notebooklm/answer/16164461?hl=en
Google, NotebookLM limits and source FAQ — https://support.google.com/notebooklm/answer/16269187?hl=en
NIST, Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile — https://www.nist.gov/publications/artificial-intelligence-risk-management-framework-generative-artificial-intelligence
Google, NotebookLM is now Gemini Notebook — https://blog.google/innovation-and-ai/products/gemini-notebook/notebooklm-gemini-notebook/

Sources checked 6 September 2026.

NotebookLM Import Failed: What to Check Before Re-uploading