AI Research Brief and Content Drafts

What this workflow does

Collect recent stories from trusted RSS and Atom feeds, build a source register, and turn that evidence into an AI research briefing plus LinkedIn and newsletter drafts for editorial review.

The workflow fetches real TechCrunch and The Verge feeds by default. It removes repeated URLs, skips stale or undated stories, ranks keyword relevance and gives each selected source a citation ID. Two Gemini stages handle research and drafting separately. Unknown citations and malformed AI responses are routed to review rather than treated as approved drafts.

What you receive

• An inactive, importable n8n workflow with 33 processing nodes and setup notes.
• Configurable feeds, keywords, audience, lookback window and source limits.
• A downloadable JSON review bundle containing research findings, LinkedIn/newsletter drafts, Markdown, source CSV, selection statistics and feed failures.
• An optional fictional demo for testing the complete processing path without AI charges.

Setup

1. Import the workflow JSON into n8n.
2. Open Settings and customize the topic, audience and keywords. The defaults are mode: live and enableAI: false.
3. Click Execute workflow to collect live evidence. Inspect Render Review Bundle for selected sources and any feed failures.
4. Select your own Google Gemini credential in Create Research Brief and Draft Content Pack. Confirm the selected model is available in your account.
5. Set enableAI: true to generate the research briefing and both content drafts. Gemini requests may be billable.
6. Download research-content-bundle.json from Download Review Bundle. Review the original sources, uncertainties and wording before using the drafts.

To skip previously reviewed stories on the next run, copy the bundle's seenUrls array into Settings knownUrls. Cross-run deduplication uses this explicit ledger; it is not automatically persisted.

Requirements and costs

No credentials are required for live feed collection or the optional demo. Live generation requires a Google Gemini API credential. The template uses built-in n8n nodes and needs no database, community nodes or sub-workflows.

The normal live AI path uses two Gemini requests. Retries may result in additional attempts. Charges, quotas and model availability depend on your account. Exact n8n server version was not recorded; the exported node versions and setup documentation are included in the package.

Output and failure handling

The review bundle returns one of four statuses:

Status	Meaning
review_ready	Briefing and drafts passed the structural/citation checks; human review is still required
needs_review	An AI request or validation gate failed; inspect the recorded review reasons
evidence_only	Live sources were collected with AI disabled
no_stories	No recent unseen stories matched your settings

Feed requests run sequentially with timeouts and retries. Failed feeds remain visible in the final bundle so you can assess incomplete coverage. The workflow does not post content, send messages or activate a schedule.

Tested behavior and limitations

Live generation passed both Gemini stages, produced eight findings and both content drafts, and generated a downloadable review bundle. A fresh recreation of the clean export passed live collection, the normal demo, empty feeds, malformed-feed handling and invalid-citation rejection.

Research uses feed titles and excerpts, not full articles. Citation checks establish whether a source ID exists; they do not verify factual accuracy. Broad keywords such as AI can include promotional, entertainment or infrastructure stories. Narrow your keywords and review the source register for your intended audience before publishing any generated content.