Workflow 1: Keyword Research & Curation

<aside> 🔎

This stage turns a manually chosen topic and search intent into a reviewable keyword database. The automation gathers and scores candidate terms; the human decides which semantic field is actually worth approving.

</aside>

Workflow map

flowchart LR
A["Manual start<br>Topic + intent"] --> B["Google Autocomplete"]
A --> C["Keyword Research API"]
B --> D["Clean + aggregate"]
C --> E["Clean metrics"]
D --> F["AI: Filter relevant keywords"]
E --> F
F --> G["Format keyword database"]
G --> H["Google Sheet<br>PRE_APPROVE"]
H --> I["Human review"]
I --> J["Move to Approved folder"]

What is automated

Start the workflow passes a topic and a detailed intent into the pipeline.
Google Autocomplete and the keyword API collect two different candidate sets: natural query suggestions and keywords with volume/competition metrics.
Cleaning, splitting, aggregation and merge nodes convert both streams into one database-ready structure.
Filter out relevant keywords1 uses AI to select 15–20 terms by semantic fit, search intent, search volume and competition rather than by exact-match wording alone.
The selected data is written to a new Google Sheet, moved into the Pre-Approved Keyword Databases folder, and announced in Telegram.

Human checkpoint

The sheet is deliberately not treated as approved merely because the metrics look good. A person can remove irrelevant terms, add missing language, correct the intent, or reject the semantic field. Moving the reviewed sheet into the Approved Keyword Databases folder is the explicit approval event that starts Workflow 2.

Selected node details

Node	Why it matters
`Set Topic & Intent1`	Defines the editorial problem before any keyword collection begins. The intent can contain detailed comparison rules, evidence requirements and product constraints.
`Autogenerate Keywords1`	Calls Google autocomplete, giving the workflow real query language rather than only API keyword rows.
`Find Keywords from Google1`	Pulls keyword candidates with monthly search and competition data.
`Filter out relevant keywords1`	AI selection layer. It combines semantic relevance with volume, competition and commercial/actionable intent.
`Format the Keyword Data for Database1`	Deterministically joins topic, intent, autocomplete terms and scored keyword rows into the final sheet schema.
`Move file1`	Places the generated sheet into the pre-approval inbox instead of silently triggering the next stage.

<aside> 🤖

The exact AI prompt for Filter out relevant keywords1 is stored in Prompts & Code Guide.

</aside>