I Built a Gemini Workflow That Killed My Research Addiction and Tripled My Writing Speed
And the exact prompts you can steal today
Twelve browser tabs open. Zero words written.
The cursor blinks at you like it’s judging your life choices.
You stare back, questioning those choices.
You had the idea when you sat down. The conviction was there. The take was loaded and ready.
Then you opened that first tab.
One became five. Five became twelve. Twelve became a graveyard of half-read studies, bookmarked threads, screenshots you’ll never look at again.
Forty-five minutes gone. You haven’t written a single sentence. The fire is out. The edge is dull.
You close the laptop. Tell yourself you’ll finish tomorrow.
Tomorrow never comes. And all this research turns kilobytes into megabytes, into gigabytes of wasted digital space.
I spent 20 years in intelligence work turning raw information into actionable analysis under time pressure. The way most creators do research is the equivalent of a government analyst on the edge of a DOGE purge Googling random keywords and hoping those actions justify their job.
There is a better way.
I’m going to walk you through the exact workflow I use to cut research time by 70% while producing BETTER content than the people spending all day buried in browser tabs.
Stop Researching Like a Hoarder
Research is the most socially acceptable form of procrastination in the creator economy.
You can convince yourself you’re working while accomplishing nothing.
FACT: A 2023 University of California study found that the average knowledge worker spends 9.3 hours per week searching for information. Nearly a quarter of the workweek burned on finding things instead of using them.
Most creators have no system. No process. They mix collection and creation into one chaotic activity that produces nothing but exhaustion.
Intelligence agencies figured this out decades ago. You separate collection from analysis. Always.
The people who skip this separation produce content that reads like a term paper nobody asked for.
The Research Assistant Workflow (Step by Step)
I use AI as a research assistant.
A research assistant does the tedious collection work so your voice stays sharp when you sit down to write.
I think of the relationship between a lead detective and a team pulling case files, running background checks, compiling evidence, conducting surveillance, and meeting with informants. The detective still makes the call. The detective still sees what nobody else sees.
You are the detective. AI pulls the files.
STEP 1: Define Your Thesis BEFORE You Research
Most people research first and form opinions second.
Backwards.
You already have a take. You already know what you believe. You’ve been thinking about your topic for weeks, months, sometimes years. Your short-form content has already validated what resonates with your audience.
Write your thesis in one sentence before you touch any research tool.
Example: “Most creators waste 70% of their research time because they treat collection and analysis as the same activity.”
That sentence is your anchor. Everything from here serves the anchor or gets cut.
STEP 2: Deploy Gemini for Collection
I use Gemini with deep research toggled on. Gemini and I have found this to be the best tool for deep research right now.
Open a new Gemini chat. Drop this exact prompt:
“I’m writing an article arguing [YOUR THESIS]. Act as my research assistant with deep research enabled. Find me:
(1) Two to four recent data points or studies published in the last 18 months that support this argument. Provide direct URLs for each source.
(2) The strongest counterargument someone credible is actually making against my position right now. Include the source URL.
(3) One historical or cross-domain analogy that illustrates this concept. Cite where you found this connection.
(4) Any surprising connections between this topic and adjacent fields that most people miss. Include source URLs.
For every piece of information you provide, include the direct URL so I can verify the source.”
The fourth request is critical. Gemini’s deep research can surface lateral connections from obscure corners of the internet your brain would never find. A human researcher stays in their lane. Gemini crosses lanes by default.
The source URL requirement is non-negotiable. This catches hallucinations before they end up in your article.
STEP 3: Filter Through Your Lens
99% of people using AI go wrong right here.
They take what Gemini gives them and paste the facts into their article. You end up sounding like everyone else. You produce content that feels like a Wikipedia entry wearing a hoodie.
Your job at this stage is to run everything through YOUR filter.
I ask myself two questions for every piece of research Gemini surfaces:
Does this support or challenge my lived experience? If I’ve seen something different in 20 years of work, that tension is where the best content lives.
Would my audience care about this in 30 seconds or less? If a data point requires three sentences of context before it makes sense, cut the data point. Find a better one.
YOUR INTERPRETATION of the research is the content.
AI has no scar tissue. No stories from the field. No 3 AM realizations that changed how you see the world.
Those are yours.
STEP 4: Build Your Contrarian Angle
Every piece of research Gemini surfaces will come with an implied consensus. The mainstream take. The obvious interpretation.
Your job is to find where the consensus is wrong, incomplete, or misleading.
I use a follow-up prompt in the same Gemini thread:
“Based on the research you’ve compiled, what is the mainstream consensus on this topic? Now search for contrarian perspectives. What are credible experts saying that challenges the consensus? What are the blind spots in the mainstream narrative? Find me the dissenting voices. Include source URLs for all contrarian perspectives.”
This prompt alone has generated more original article angles than any brainstorming session I’ve ever done.
Because Gemini has deep research enabled, when you ask it to find contrarian perspectives it can pull from niche publications, academic papers challenging mainstream assumptions, and minority expert opinions that rarely surface in a standard Google search.
You get perspectives in 90 seconds that would take you hours to find manually.
Then you decide which contrarian angles match your experience and your audience. You are still the editorial filter.
Doing the Work
This workflow only works if YOU have something to say.
Gemini will hand you all the data. All the angles. All the connections.
But if you don’t have a perspective shaped by real experience, you’ll produce content that sounds informed BUT says nothing.
The research assistant workflow is a force multiplier.
The creators who will dominate the next five years are not the ones who use AI the most. They are the ones who have done the work, lived the lessons, and use AI to AMPLIFY what they already know.
You need the years in the field. The failures that taught you what the textbook got wrong. The phone screen cracked from being up at 4 AM working on your thing while the rest of the world sleeps.
AI gives you speed.
Experience gives you substance.
Your Move
If you’re reading this newsletter, you are already ahead. You understand that AI is a tool. You understand that the goal is leverage.
Take your next piece of content. Write your thesis in one sentence before you open a single browser tab. Then open Gemini with deep research enabled and run the research assistant prompt I gave you above.
Time yourself.
Compare the results to your old process.
I guarantee you’ll never go back.
If you want the full prompt library I use for research, outlining, editing, and repurposing, along with the systems that let me produce content around a full-time job and two daughters who need their dad present, that is what this newsletter is for.
Paid subscribers get the complete prompt library, the step-by-step workflows, and real-time updates on what’s working. No theory. No fluff. Systems you build once and run forever.
The subscription is $15/month or $100/year.
Click the link below to join:
Ryan


Excellent!
Love how you don't just say what to do with the AI, but how to think about things AND hammer the importance of using AI PROPERLY, not just MORE.
Great work, Ryan. My system is solid but complex and at times cumbersome. I have all my highlights linked in project files within Claude, including hundreds of pages from articles and Kindle, but these prompts that bring in reliable outside research for additional support and idea sparks will be a great addition.