Session 8.2: Stage 1: Research and Source Gathering

Course → Module 8: The Pipeline

Session 2 of 10

Research Is the Foundation

Skip research, and you are building on sand. Every factual claim, every statistic, every "according to experts" in your final content needs a traceable source. If you let AI generate from its training data alone, you get plausible-sounding text backed by nothing. That is not content. That is noise.

The research stage exists to gather verified inputs before any prose is written. The output of this stage is not content. It is raw material: facts, sources, data points, and perspectives that your content will be built from.

The Research Workflow

In Module 7, you built a research workflow using search APIs. Now that workflow becomes Stage 1 of your production pipeline. The process follows a consistent sequence.

flowchart TD A["Define Research Questions"] --> B["Query Search APIs"] B --> C["Filter by Relevance"] C --> D["Extract Key Data"] D --> E["Verify Claims"] E --> F["Compile Research Brief"] C -- "Irrelevant results" --> B E -- "Unverifiable" --> G["Flag for Manual Check"] G --> F style A fill:#222221,stroke:#c8a882,color:#ede9e3 style B fill:#222221,stroke:#6b8f71,color:#ede9e3 style C fill:#222221,stroke:#8a8478,color:#ede9e3 style D fill:#222221,stroke:#c8a882,color:#ede9e3 style E fill:#222221,stroke:#c47a5a,color:#ede9e3 style F fill:#222221,stroke:#6b8f71,color:#ede9e3 style G fill:#222221,stroke:#c47a5a,color:#ede9e3

Defining Research Questions

Before you search anything, write down what you need to know. Not "research topic X." Specific questions that your content must answer.

For an article about remote work productivity, your research questions might be:

What do the most-cited peer-reviewed studies say about remote work productivity?
What is the sample size and methodology of each major study?
Which companies have published internal data on remote vs. office productivity?
What are the most common metrics used to measure productivity in these studies?
What do the critics of remote work research say about methodological flaws?

Each question becomes a search query. Specific questions produce specific results. Vague questions produce noise.

API-Assisted Search

Your Tavily or Google Search Grounding setup from Module 7 handles the query execution. Feed each research question as a query. Collect the top results. For each result, extract: title, URL, publication date, a key excerpt, and a relevance score.

The automation matters here. Manual research for 5 questions across 10 sources takes 1 to 2 hours. API-assisted research takes under 5 minutes. The time savings compound at scale.

The Research Brief Format

The output of Stage 1 is a structured document called a research brief. This is what gets handed to Stage 2 (Outline) and Stage 3 (Draft Generation). Its format should be consistent across every piece of content you produce.

Section	Contents	Purpose
Topic Summary	2-3 sentences defining the content's focus	Scope definition
Key Findings	5-10 bullet points of verified facts	Core claims the content will make
Source List	Title, URL, date, reliability rating for each source	Citation and fact-checking
Data Points	Specific numbers, percentages, dates with sources	Evidence for claims
Counterarguments	Opposing views with their sources	Balanced coverage
Gaps	Questions that could not be answered by search	Flags for manual research

Source Quality Assessment

Not all search results are equal. A peer-reviewed study from a major university is not the same as a blog post from an anonymous author. Your research workflow needs a source quality filter.

A simple three-tier system works:

Tier 1 (use directly): peer-reviewed journals, government data, official company reports, established news outlets
Tier 2 (verify before using): industry publications, well-known blogs with named authors, conference papers
Tier 3 (corroborate with Tier 1): anonymous sources, social media, opinion pieces, aggregator sites

API search results arrive without quality ratings. Assigning them is a human judgment step. It takes 30 seconds per source and prevents entire categories of errors downstream.

The quality gate for Stage 1: does the research brief contain enough verified information to support every claim the content will make? If the answer is no, keep researching. Do not advance to outlining.

Timing and Benchmarks

Your first research brief will take 30 to 60 minutes. With a tuned API workflow and consistent brief format, the target is 15 to 20 minutes per piece. At scale, when producing a batch of related content, much of the research overlaps, and the per-piece time drops further.

Time the process every time. Not to pressure yourself, but to measure improvement. A pipeline you cannot measure is a pipeline you cannot improve.

Assignment

Execute Stage 1 for a real piece of content. Use your research workflow from Module 7 to produce a complete research brief with the following sections:

Topic summary (2-3 sentences)
Key findings (5-10 verified facts)
Source list with reliability ratings (Tier 1, 2, or 3)
Data points with citations
At least one counterargument with its source
Any gaps requiring manual follow-up

Time yourself from start to finish. This is your benchmark for research speed. You will improve on it.

Stage 1: Research and Source Gathering