Google’s April 2026 Notebooks update finally made me move my working life out of ChatGPT.
The feature brought NotebookLM-style project notebooks into the main Gemini chat interface, starting with paid Google AI subscribers before expanding more broadly.
I had experience with both Gemini and ChatGPT, and I didn’t see a major gap in what either one could do.
Both worked for my needs. What changed the equation was mostly integration.
OpenAI kept building ChatGPT as more of a standalone interface, while Google was pulling Gemini into the ecosystem I already work in. So I moved.
Move your writing style into Gemini first
Most AI assistants sound like a customer-service script until you tell them otherwise.
I’d spent a lot of time training ChatGPT’s voice to stop apologizing and repeating my prompt back to me before answering. That was the first thing I moved.
In Gemini, you set the baseline from Settings & help > Personal context > Your instructions for Gemini. Those instructions act like a persistent preference layer for regular Gemini chats.
I added my core rules first. Don’t over-explain basic answers. Use direct language. Preserve technical formatting. Don’t summarize my question back to me unless I ask.
I added more, but the rest is out of scope here and worth covering on its own.
Rebuild your Custom GPTs as Gemini Gems
Custom GPTs were the part of ChatGPT I was most worried about leaving behind. I relied on them for repeatable jobs, especially for my code reviews.
Gemini Gems aren’t a perfect one-to-one copy of Custom GPTs. Gems also don’t have the same public marketplace, but they take less technical setup to use inside everyday Google apps.
Nonetheless, I copied the core system prompts from my ChatGPT custom bots, cleaned them up, and rebuilt them as Gem instructions.
A good Gem spec comes down to four things:
|
Spec pillar |
What it does in a Gem |
Example |
|---|---|---|
|
Persona |
Defines the role and expertise |
You are a senior developer specializing in Python performance. |
|
Task |
Defines the job |
Audit code snippets for vulnerabilities and suggest modern alternatives. |
|
Context |
Gives the model the background it needs |
I manage high-traffic applications where load speed is a priority. |
|
Format |
Controls the output |
Present findings in a table with Issue, Severity, and Fix. |
That structure kept my Gems from turning into vague prompt dumps, and my Gems now behave much like my old GPTs.
Give Gemini a real project memory with NotebookLM
Chat history is a bad project archive. The useful chats you remember from three weeks ago disappear unless you named them perfectly. Especially if you keep opening new tabs as I do.
Google’s bidirectional sync between Gemini and NotebookLM fixes much of that. Inside Gemini, you can now group PDFs, live Drive documents, and past chats into a Notebook from the side panel.
Sources added to Gemini appear in NotebookLM, and files added to NotebookLM sync back to Gemini.
If a research chat turns useful, I use the overflow menu and click Add to notebook. Gemini treats that like a container move.
The chat leaves the main recent-history list and is filed inside the project space instead. That changes how the AI answers.
Gemini can base responses on the specific data inside the Notebook, including the verified documents I’ve uploaded.
Connect Gemini to the Google apps you already use
Gemini becomes much more useful after Workspace extensions are on. Go to Settings and Help > Connected Apps > Google Workspace to activate them.
After that, typing @ in the prompt box opens a live menu for Google Drive, Google Docs, Google Sheets, and Gmail.
You can ask Gemini to summarize an email thread from this morning or pull action items from a meeting document.
Gemini can also search for supported Workspace content and return source links you can review against the original files.
You get better results when you name the exact file or folder you want Gemini to search for.
Pointing it at a specific Drive folder works like handing it a curated dossier. It reduces the chance that the model pulls details from unrelated personal files.
Gemini won because it sits closer to my work
For my workflow, the model race matters less than the plumbing around the model.
ChatGPT still has a place, and it can connect to some Google apps through its extensions and support system. But Gemini is closer to my final output.
It covers more of Google Workspace, and the overall experience is better because it is Google talking to Google.
