Troubleshooting
Common problems and how to fix them.
NotebookLM Problems
"I can't add a source"
| Symptom | Solution |
|---|---|
| URL won't add | Some sites block AI tools. Try copying the text and adding it as "Copied text" instead of a URL. |
| File won't upload | Check the file type. NotebookLM accepts PDFs, Google Docs, and plain text. Convert other formats first. |
| "Source too long" | Split the document into smaller sections and add them as separate sources. |
| Peer's blog post won't add | GitHub Pages blogs sometimes have issues. Copy the post text directly and paste it as "Copied text." |
"NotebookLM's answers seem off"
| Symptom | Solution |
|---|---|
| Answers don't cite sources | Make sure you're asking about topics covered in your sources. If you ask about something not in your sources, it can't ground the answer. |
| Wrong source cited | Click the citation number to check. If it's wrong, ask a more specific question that points to the source you want. |
| Ignoring some sources | You might have too many sources. NotebookLM prioritizes some over others. Try asking specifically about the source being ignored. |
| Misinterpreting my source | Ask a follow-up: "Quote the exact passage you're basing this on." Then check if the interpretation is fair. |
"Discover Sources isn't finding good results"
| Symptom | Solution |
|---|---|
| Results are low quality | Be more specific in your query. Instead of "AI jobs," try "IMF report 2024 AI job displacement statistics." |
| Can't find a specific source | Use regular Google search to find the source, then add it manually via URL or copied text. |
| Results are paywalled | Look for free summaries, press releases, or academic preprints. Try adding "PDF" or "free access" to your search. |
Gemini Gems Problems
"My Gem isn't following instructions"
| Symptom | Solution |
|---|---|
| Skipping steps | Number your steps explicitly (1, 2, 3...) and ask "Let's go through step by step" when you start. |
| Wrong tone | Add more specific tone instructions. Instead of "be professional," try "be direct and concise, avoid jargon." |
| Not using my format | Include a sample output in your instructions: "Format your response like this: [example]" |
| Ignoring placeholders | Be explicit: "When you write the draft, mark every factual claim with [CITATION NEEDED] in brackets." |
"I can't create a Gem"
| Symptom | Solution |
|---|---|
| Don't see "Gem manager" | Make sure you're using Gemini (gemini.google.com), not Google Search AI. Gems are in the left sidebar. |
| Gem won't save | Check that you've filled in both the name and instructions. Try a shorter instruction set to test. |
| Feature not available | Gems may require a Google Workspace or Google One subscription. Check your account type. |
"I can't attach my NotebookLM notebook"
| Symptom | Solution |
|---|---|
| No attachment option | This feature requires Google One AI Premium. If you don't have it, use the manual transfer method described in the Combined Workflow page. |
| Notebook doesn't appear | Make sure the notebook is in the same Google account as your Gemini. Check that the notebook is saved (not just open). |
| Attachment fails | Try closing and reopening Gemini. If it persists, use manual copy/paste as a workaround. |
Output Quality Problems
"The draft sounds too generic"
Problem: AI-generated drafts often sound the same — proper but bland.
Solutions:
- Add voice instructions: In your Gem, describe your actual writing voice. "I use contractions. I sometimes start sentences with 'But' or 'And.' I'm slightly sarcastic."
- Use your own opening: Write the first paragraph yourself, then ask AI to continue matching that style.
- Request specifics: "Give me three different ways to phrase this opening, from most formal to most conversational."
- Plan to revise: Treat the draft as raw material, not final product. Read it aloud and change anything that doesn't sound like you.
"The summary misrepresents my peer"
Problem: AI summaries sometimes miss nuance or accidentally strawman.
Solutions:
- Ask for quotes: "Summarize their argument using their own words. Quote at least two key phrases."
- Request steelman: "Give me the most charitable interpretation of their argument."
- Check against the source: Read the AI summary, then re-read the peer post. Would your peer agree with how they're represented?
- Ask your peer (if possible): "Does this fairly capture what you were saying?"
"I can't tell what's verified and what's not"
Problem: Draft mixes sourced claims with AI-generated claims.
Solutions:
- Use placeholders from the start: Tell your Gem: "Mark every factual claim with [VERIFY] until I confirm it."
- Color-code your draft: In Google Docs, highlight everything you haven't personally verified. Don't publish until all highlights are removed.
- Keep a verification log: For each claim, note: source, whether you checked it, what you found.
Workflow Problems
"This is taking longer than just writing myself"
Problem: Learning new tools has a startup cost.
Reframe:
- The first post with these tools will be slow. By the third or fourth, you'll have reusable notebooks and Gems.
- The time you spend now on verification saves you from embarrassing errors later.
- If a particular workflow isn't helping, simplify. Use only the parts that add value.
Minimal viable workflow: If the full process is overwhelming, try just NotebookLM for research OR just a Gem for outlining — not both. Add complexity as you get comfortable.
"I'm not sure if I'm using AI ethically"
The key questions:
- Is the thesis mine? Did I choose what to argue, or did I just accept AI's suggestion?
- Can I defend every claim? If asked about any part of the post, can I explain it?
- Did I verify? Is everything sourced and checked?
- Does it sound like me? Would friends recognize my voice?
If you can answer "yes" to all four, you're using AI as a tool, not a replacement for your thinking.
Getting Help
For tool-specific issues:
- NotebookLM: Google NotebookLM Help
- Gemini: Google Gemini Help
For course-related questions:
Email Dr. Plate or bring questions to class. Technical problems with AI tools are a normal part of learning — you won't be penalized for asking.
Remember the slogan: If it breaks, breathe. Then try again. These tools are meant to help, not frustrate. If something isn't working, simplify or try a different approach.