Build an AI Research Digest with n8n

Automate NASA research summaries with Gemini, n8n, and Google Sheets.

Introduction

30 Second Summary

Interesting research can arrive faster than you have time to read it. Useful links soon disappear into crowded tabs or an overflowing inbox.

In this project, you will build a browser-based research assistant in n8n Cloud. It turns three live NASA Science RSS items into concise notes from Google Gemini stored in Google Sheets.

What You'll Build

You will run the finished workflow to watch a clean research log fill with three source-linked summaries ready to review or share.

By the end of this project, you'll have:

A live workflow run you can start to bring exactly three current NASA Science items onto the n8n canvas.
Consistent AI research notes that give you a factual summary plus three key takeaways for every item.
A duplicate-safe daily digest you can review in Google Sheets. Repeated source URLs update their existing rows when the published workflow runs again.
Secret Mission: Add an AI relevance gate that stores only research matching a theme you define.

Are there any prerequisites?

You need a Windows computer with Google Chrome or Microsoft Edge plus a Google account. The guide covers n8n Cloud setup from scratch, so no coding experience is required.

Before We Start

Before any hands-on work, this step locks in the purpose of your research digest.

A clear purpose names the intended reader plus the research topic the digest helps them follow. Your checkpoint captures that purpose in one sentence.

Set Up Your Research Workspace

You already know who the digest serves. Now the plan needs a browser workspace where each automation step can connect.

The AI stage depends on an eligible credit balance. Confirming that balance first prevents you from building a workflow that cannot run.

In this step, you will create an n8n Cloud trial. You will finish with a blank workflow named AI Research Digest and a confirmed Gateway credits balance.

In this step, get ready to:

Create an eligible n8n Cloud trial with a confirmed Gateway credits balance.
Create a blank workflow named AI Research Digest.

Create your n8n Cloud trial

n8n Cloud hosts the workflow editor in your browser. Gateway credits give supported AI nodes access to a shared prepaid balance without a separate provider account.

The Cloud Admin Panel displays one shared balance for the workspace. It also shows how that balance is used by workflow or service.

The trial needs no credit card. This keeps the setup separate from a billing commitment.

Visit the official n8n quickstart in Chrome or Edge.
Follow the Cloud trial link on the quickstart page.
Complete the account setup until your n8n Cloud workspace opens.
Open Cloud Admin Panel in a new browser tab from the workspace account area.
Locate the Gateway credits balance to confirm that the trial includes funds for later AI runs.
Skip any prompt to add a payment method.
Skip every option to add funds.
Keep automatic top-ups disabled.

Why Check the Balance Now?

The Google Gemini step later spends from this shared balance. An amount above zero proves that the included AI path is available before you add an AI node.

Choose the outcome below that matches the balance shown in the Cloud Admin Panel:

✔️ I see a Gateway credits balance

You have cleared the main setup risk. Your workspace can use the included AI path in a later step.

ⓧ The balance is missing or zero

A missing balance means this trial does not meet the project's cost boundary yet.

Confirm that the account currently open is your newly created n8n Cloud trial.
Return to the Cloud Admin Panel after switching to that trial.
Pause the project if the balance is still missing or zero.
Ask n8n support whether the new trial is eligible for Gateway credits.

Can't Open the Workspace?

Retry the trial link in a private Chrome or Edge window.
Confirm that you are signed in with the email address used for the trial.

Ask AI to diagnose the access problem. Share account-specific questions in the n8n Community.

Create your blank workflow

A blank canvas gives the digest one home before nodes arrive. Naming it now makes later executions easy to find.

Return to the n8n workspace tab from earlier.
Match your current n8n screen to one path below:

Welcome View

Select Start from Scratch.

You should see a blank workflow canvas.

Overview Page

Select Create Workflow.

You should see a blank workflow canvas.

Replace the current workflow title in the canvas header with AI Research Digest.
Leave the canvas empty.

Before you check both browser tabs, what two signs should prove the workspace is ready? Make your prediction before continuing.

Return to the Cloud Admin Panel tab from earlier.
Confirm that the Gateway credits balance still appears.
Return to the n8n workflow tab.
Confirm that the canvas header shows AI Research Digest.
Confirm that the canvas contains no nodes.
Leave the workflow open on the canvas.

You should see a Gateway credits balance in one tab. You should see AI Research Digest above a blank canvas in the other.

Your research workspace now has a confirmed AI credit path. Next, you will connect it to a live NASA Science feed. You will cap each run at three items.

Pull In Live Research Items

Your n8n Cloud workspace is ready. The blank AI Research Digest workflow is waiting for its first path.

The workflow needs real structured research items before AI can summarize anything. This step builds a small test path from the live NASA Science feed.

In this step, get ready to:

Add a Manual Trigger for controlled tests.
Read live NASA Science Photojournal items from an RSS feed.
Use an item limit to keep exactly three feed items.

Add a manual starting point

A trigger decides when a workflow begins. Manual Trigger keeps these first tests under your control.

Why start manually?

Manual runs let you inspect each node before the workflow becomes automatic. You choose when the source is read.

Select the add-node control in the center of the blank canvas.
Search for Manual Trigger.
Select Manual Trigger from the results.
Confirm that Manual Trigger is the first node on the canvas.

Good start. Your workflow now has a controlled entry point for every test run.

Read the NASA Science feed

An RSS feed packages recent website content as structured items. The RSS Read node brings those items directly onto your canvas.

Why use RSS for this digest?

RSS provides article details in a consistent structure. This keeps the workflow focused on research automation without extracting content from web pages.

Select the add-node control attached to the right side of Manual Trigger.
Search for RSS Read.
Select RSS Read from the results.

The canvas now shows Manual Trigger connected to RSS Read.

Enter https://science.nasa.gov/feed/photojournal/latest-content/ in the URL field.
Run RSS Read with the execution control in its configuration panel.

The live source is connected. You can now inspect multiple current NASA Science Photojournal items with article details in the output.

No RSS items in the output?

Check that the URL field contains https://science.nasa.gov/feed/photojournal/latest-content/.
Run RSS Read again after correcting the feed address.

Help me troubleshoot an RSS Read node that returns no NASA Science feed items.

Keep three items and verify the output

Small test batches make each result easier to inspect. The Limit node controls how many feed items continue through the workflow.

Select the add-node control attached to the right side of RSS Read.
Search for Limit.
Select Limit from the results.

The canvas now shows Limit connected after RSS Read.

Enter 3 in Max Items.
Select First Items under Keep.

Before you execute the full workflow, which node do you expect to show the smaller item count?

Select Execute Workflow.
Select Limit after the run to inspect its output.

That is your first live data path working. The Limit output now holds exactly three current NASA Science Photojournal items with article details.

Workflow checkpoint

Your canvas should show Manual Trigger connected to RSS Read. RSS Read should connect to Limit.

The RSS Read URL should be https://science.nasa.gov/feed/photojournal/latest-content/.

The Limit node should use 3 for Max Items. Keep should show First Items.

The Limit output should display exactly three NASA Science Photojournal items with article details.

Seeing the wrong number of items?

Check that Max Items contains 3.
Check that Keep shows First Items.
Select Execute Workflow again after correcting the settings.

Help me find why my Limit output does not contain exactly three items.

Your research input is live and easy to inspect. Next, you will turn each of these three feed items into a consistent AI note.

Generate AI Research Notes

Your n8n workflow now captures three live NASA Science Photojournal items in every test run. Those raw items still require manual reading before they become a useful digest.

Google Gemini turns each item into a short factual briefing. A source-grounded prompt tells the model to acknowledge missing detail instead of guessing.

In this step, get ready to:

Connect Google Gemini with Gateway credits.
Generate one source-grounded briefing for each feed item.
Map each result into four storage-ready fields.

Connect Google Gemini

Each incoming item needs the same AI transformation. The Google Gemini node applies one model configuration to all three items.

This test uses the Gateway balance you confirmed earlier. Your billing settings stay untouched.

Select the connector after Limit to add the next node.
Search for Google Gemini in the node picker.
Select Google Gemini.

Google Gemini now appears directly after Limit on the canvas.

Choose Use Gateway credits when the node asks for a credential.
Select Text.
Select Message a Model.

The node now shows the model settings for a text completion.

Open the Model dropdown.
Choose one of the models shown in the live list.

Why the Live Model List Matters

The live Model dropdown only shows models available to your account. Choosing from this list keeps the workflow aligned with current Gateway availability.

How Three Items Stay Separate

Limit passes three separate items into Google Gemini. The node applies the same message template once to each incoming item.

This item-by-item pattern produces three distinct responses from one node configuration.

Build a source-grounded prompt

A useful digest stays anchored to the information supplied by the feed. The instruction also gives Gemini a safe response when an article lacks enough detail.

Select the Google Gemini node.
Place the cursor in the message editor.
Drag the item title from INPUT into the message.
Drag the available description or summary from INPUT beneath the title.
Add the instruction beneath the source content by pasting this text:

Using only the supplied feed item, write SUMMARY as two factual sentences, then KEY TAKEAWAYS as three bullets. If the item lacks enough detail, say so instead of guessing.

What Does This Prompt Control?

The title anchors the response to the article topic.
The description or summary supplies the available article detail.
The source-only instruction limits the response to supplied information.
The requested structure makes every briefing easier to scan.

Before you execute Google Gemini, do you expect every feed item to contain enough detail for a complete briefing?

Execute Google Gemini using the node's execution control.
Select each output item to inspect its response.

You'll see three responses. Each response uses SUMMARY followed by KEY TAKEAWAYS.

A sparse feed item may acknowledge that it lacks enough detail. That response proves the grounding instruction is working.

That is the reading bottleneck handled. Each source now has a compact briefing you can scan.

Missing or Weak Responses?

Confirm Google Gemini is connected directly after Limit if fewer than three responses appear.
Confirm both source fields are present in the message if a response lacks context.
Return to the Cloud Admin Panel from earlier if the execution mentions unavailable Gateway credits.

Help me diagnose my Google Gemini output.

Map the digest fields

Field mapping turns variable model output into a stable record. Edit Fields (Set) keeps each article's source details beside its AI response.

Select the connector after Google Gemini to add the next node.
Search for Edit Fields (Set) in the node picker.
Select Edit Fields (Set).

Edit Fields (Set) now appears after Google Gemini on the canvas.

Choose Manual Mapping mode.

Why Shape the Data Here?

Stable field names give every item the same structure. Source details remain easy to trace back to the original article.

The AI response becomes one reviewable field instead of an unlabelled model result.

Add a field named Title.
Drag the incoming RSS title from INPUT into Title.
Execute Edit Fields (Set) using the node's execution control.

You'll see three items containing a populated Title field.

Add a field named Source URL.
Drag the source article URL from INPUT into Source URL.
Execute Edit Fields (Set) using the node's execution control.

You'll see each title paired with its original source URL.

Add a field named Published.
Drag the feed's publication value from INPUT into Published.
Execute Edit Fields (Set) using the node's execution control.

You'll see the publication value beside each article's title and source URL.

Add a field named AI Notes.
Drag the Gemini response from INPUT into AI Notes.

Before you run the full workflow, how many structured items do you expect at the final node?

Run the full flow by selecting Execute Workflow.
Select Edit Fields (Set) after the run completes.

You'll see exactly three items. Each item contains Title plus Source URL.

Each item also contains Published plus AI Notes.

You have turned three unpredictable feed items into a consistent dataset. The final node is ready for durable storage.

Final Output Missing Fields?

Return to Limit if the final output contains the wrong number of items.
Confirm Max Items shows 3.
Confirm Keep shows First Items.
Check each mapping for an empty source value if one final field is blank.

Help me fix my structured n8n output.

Your AI research notes are structured. Next, you'll give them a durable home in Google Sheets.

Save the Digest and Find Duplicates

Your n8n workflow already turns three live NASA Science items into structured AI notes.

Those notes need a durable home where you can review each result. Google Sheets provides a searchable research log that remains available after each workflow execution.

A repeatable automation must also handle the same source more than once. This step tests the first storage design against that requirement.

In this step, get ready to:

Create the research spreadsheet with four structured columns.
Connect n8n to Google Sheets through managed OAuth2.
Execute the unchanged workflow twice to inspect its repeat behaviour.

Create the Google Sheet

The spreadsheet gives every digest item the same structure. Its headers must match the four fields produced by Edit Fields (Set).

Open a new browser tab in Chrome or Edge.
Go to Google Sheets in that tab.
Create a blank spreadsheet.

You should see an empty spreadsheet grid that is ready to become your research log.

Rename the spreadsheet to AI Research Digest.
Enter Title in cell A1.
Enter Source URL in cell B1.

The first two columns now identify each research item and its original location.

Enter Published in cell C1.
Enter AI Notes in cell D1.
Confirm that row 1 displays all four headers from columns A through D.

Why These Four Columns?

The Title column makes each article easy to scan.
The Source URL column preserves the link to the original item.
The Published column records the feed's publication value.
The AI Notes column stores the generated summary and takeaways.

Connect n8n to Google Sheets

A managed OAuth2 credential lets n8n connect through your Google sign-in. You can authorize the connection without creating a separate Google Cloud credential.

Account authorization can feel sensitive. This flow stores the managed credential in n8n without asking you to paste your Google password or an API key into the workflow.

Switch back to the open AI Research Digest workflow.
Select the connector after Edit Fields (Set).
Search for Google Sheets in the node picker.
Select the Google Sheets node.

The Google Sheets node should open beside Edit Fields (Set) on the canvas.

Create a new credential in the node's credential selector.
Choose the managed OAuth2 sign-in option.
Sign in with your Google account when the authorization window opens.
Authorize n8n to connect to Google Sheets.

How Managed OAuth2 Helps

Managed OAuth2 keeps the Google connection inside n8n's credential system. The workflow node can reuse that authorized connection without storing your Google password in a field.

This keeps the authentication step separate from the data mappings you configure next.

Select the AI Research Digest spreadsheet in the Google Sheets node.
Choose Sheet Within Document as the resource.
Choose Append Row as the operation.
Choose Map Each Column Manually as the mapping mode.

The mapping area should now show the four headers from your spreadsheet as destination columns.

Map the incoming Title field to the Title column.
Map the incoming Source URL field to the Source URL column.
Map the incoming Published field to the Published column.
Map the incoming AI Notes field to the AI Notes column.
Confirm that the canvas ends with Edit Fields (Set) connected to Google Sheets.

Google Sign-In Did Not Complete?

Check whether your browser blocked the authorization window. Allow the window for n8n before trying the credential flow again.

Confirm that you selected the Google account that owns the AI Research Digest spreadsheet.

Help me troubleshoot the managed Google Sheets connection.

Run the Same Feed Twice

The first execution proves that workflow data reaches the spreadsheet. The second execution tests what happens when the source items remain unchanged.

Switch to the workflow canvas.
Select Execute Workflow.
Switch back to the AI Research Digest spreadsheet.
Refresh the spreadsheet tab.

You should see three data rows beneath the headers. Each row should contain values across all four columns.

Your first three AI notes now have a durable home in the sheet.

Before the second execution, do you think the number of rows will stay the same or increase?

Switch back to the n8n workflow.
Select Execute Workflow a second time.
Return to the spreadsheet from the first execution.
Refresh the spreadsheet tab again.

You should now see six data rows beneath the header. The Source URL column contains repeated values from the unchanged feed.

The Duplicates Are Intentional

The Append Row operation adds another row for every incoming item. It does not check whether the same source URL already exists.

Running the unchanged feed twice exposes the limitation that a repeatable automation must solve.

Your digest now stores real AI notes and exposes the weakness in append-only storage. Next, you will make repeated runs duplicate-safe and automate the daily schedule.

Make the Digest Reliable and Automatic

Your second Google Sheets test exposed the weakness in append-only storage. Every repeated source URL became another row.

Reliable automation needs idempotency, which lets the same input run again without creating extra records. In this step, you will make n8n update matching rows automatically.

In this step, get ready to:

Make repeated runs update rows matched by Source URL.
Replace the Manual Trigger with a daily Schedule Trigger.
Prove the published workflow runs without adding rows for stored source URLs.

Make repeated runs duplicate-safe

The Google Sheets node currently appends every item as a new row. A matching field gives the node a stable value for finding an existing row before it writes.

Select the Google Sheets node on the canvas.
Change Operation from Append Row to Append or Update Row.
Choose Source URL as the matching field.

The node now has a unique value for locating each stored article. New source URLs still create rows.

Keep the remaining Google Sheets settings unchanged.

What stays the same?

The node continues using your existing managed credential. It also continues using the AI Research Digest spreadsheet.

The required settings are:

Resource remains Sheet Within Document.
Mapping mode remains Map Each Column Manually.
The existing mappings remain Title, Source URL, Published, and AI Notes.

What happens to the earlier duplicates?

Rows created by the earlier append-only tests may remain in the sheet. The new operation controls future runs by updating a row when its Source URL already exists.

Return to the AI Research Digest spreadsheet from earlier.
Record the current last occupied row here: current last occupied row number.

Before you run the workflow twice, consider whether the last occupied row will move.

Switch back to the AI Research Digest workflow.
Select Execute Workflow.
Wait for the execution to complete.
Select Execute Workflow again.
Return to the AI Research Digest spreadsheet.
Refresh the spreadsheet page.

The last occupied row should still be current last occupied row number for the same three source URLs. The matching rows receive the latest values instead.

That closes the duplicate gap. Future reruns can refresh stored articles without growing the sheet for those URLs.

Did the row count increase?

Confirm that Operation shows Append or Update Row.
Confirm that the matching field shows Source URL.
Confirm that the incoming Source URL value is still mapped to the sheet column with the same header.

Help me stop repeated source URLs from creating new Google Sheets rows.

Replace manual runs with a daily schedule

The duplicate-safe storage path can now run repeatedly. A scheduled trigger removes the need to remember each daily execution.

Select the Manual Trigger node on the canvas.
Press the Delete key.
Use the node picker on the canvas to add Schedule Trigger in the empty starting position.
Connect the Schedule Trigger output to the RSS Read input.

Your canvas now begins with a trigger designed for automatic runs. The existing research path remains connected after it.

Select the Schedule Trigger node.
Set Trigger Interval to Days.
Set Days Between Triggers to 1.

The trigger panel now reflects a daily interval. Next, you will choose the time for that run.

Set Trigger at Hour to 8am.
Set Trigger at Minute to 0.

The trigger panel now reflects an 8am run time. The workflow timezone decides which local 8am n8n uses.

Check the workflow timezone in the workflow settings.
Set the workflow timezone to your local timezone if the values differ.

Why does the workflow timezone matter?

Schedule Trigger uses the workflow timezone when one is set. A matching local timezone keeps the daily run aligned with your intended morning hour.

The schedule is ready when the trigger settings line up with the finished workflow below.

✔️ My schedule matches

Your daily trigger is ready for publication.

ⓧ I'd like to double check

Confirm that the canvas starts with Schedule Trigger.
Confirm that the node order remains Schedule Trigger, RSS Read, Limit, Google Gemini, Edit Fields (Set), and Google Sheets.
Confirm that Trigger Interval shows Days.
Confirm that Days Between Triggers shows 1.
Confirm that Trigger at Hour shows 8am.
Confirm that Trigger at Minute shows 0.
Confirm that the Google Sheets operation shows Append or Update Row.
Confirm that the Google Sheets matching field shows Source URL.

Publish the finished workflow

Publishing activates the daily schedule. Manual execution remains available for a final controlled check.

Scheduled runs use Gateway credits from the balance you confirmed earlier. Your no-billing setup stays in place.

Select Publish in the workflow toolbar.

The workflow should now show a Published status. Its daily schedule is active.

Before the final manual run, consider whether the sheet's last occupied row will move.

Select Execute Workflow.
Inspect the completed execution path on the canvas.
Return to the AI Research Digest spreadsheet.
Refresh the spreadsheet page.
Compare the last occupied row with current last occupied row number.

You should see a successful execution path from Schedule Trigger through Google Sheets.

The last occupied row should remain unchanged for the same stored source URLs. Your published digest now updates matching research every day.

Is the final check different?

Confirm that the workflow still shows Published after your latest changes.
Confirm that Schedule Trigger connects directly to RSS Read.
Confirm that the Google Sheets matching field remains Source URL.

Help me diagnose my published n8n schedule or duplicate-safe Google Sheets update.

Secret mission

Add an AI Relevance Gate

Your digest already writes source-grounded notes to a duplicate-safe sheet. Add a relevance gate so only research matching your chosen theme reaches storage.

Clean Up Your Resources

This project stays at $0 while it uses the trial's included Gateway credits with no billing changes. Decide whether to keep the n8n Cloud workflow running, pause its daily schedule, or delete its saved resources.

Cost warning

The published schedule can keep using Gateway credits after you leave this project.

Do not add funds.
Do not enable automatic top-ups.
Choose Pause or Delete if the trial or included balance ends.

Resources you used:

The published AI Research Digest workflow with its daily Schedule Trigger.
The Google Sheets managed OAuth2 credential stored in n8n.
The AI Research Digest spreadsheet stored in Google Drive.

Keep everything running

No action is needed. Choose this if you still want fresh research notes added to your digest each morning.

Leave AI Research Digest published in n8n.
Check the Gateway credits balance in the Cloud Admin Panel regularly.
Review the execution history after scheduled runs.
Review AI Research Digest in Google Sheets to confirm that relevant items continue appearing.

Pause - I'll come back to this later

Pausing stops daily AI requests while preserving the workflow, credential, and spreadsheet for your return.

Return to AI Research Digest in n8n.
Select Unpublish to stop the daily schedule.
Check that the workflow no longer shows Published.

The workflow, credential, and spreadsheet remain ready for your return. Scheduled runs stop using Gateway credits while the workflow is unpublished.

Delete - I don't want to use this again

Deleting the project can feel drastic. These actions target only the workflow, saved credential, and spreadsheet created for this digest.

Return to the workflow list in n8n.
Locate AI Research Digest.
Use its workflow actions menu to delete it.
Go to the credentials area in n8n.
Locate the Google Sheets managed OAuth2 credential used by AI Research Digest.
Use its credential actions menu to delete it.
Confirm AI Research Digest no longer appears in the workflow list.
Confirm the managed Google Sheets credential no longer appears in the credentials area.

That clears the automation side of the project. The spreadsheet is the final saved resource.

Open Google Drive in the browser signed into your Google account.
Locate the AI Research Digest spreadsheet.
Move the spreadsheet to the Google Drive trash.
Open the trash area in Google Drive.
Permanently delete AI Research Digest.
Confirm the spreadsheet no longer appears in the trash.

Cleanup is complete. Your daily digest no longer runs or stores new results.

Nice Work!

You did it! Your published AI Research Digest now turns current NASA Science feed items into a daily research log in Google Sheets. Its duplicate-safe storage keeps each source link reviewable across repeat runs.

You've learned how to:

Build a published n8n Cloud workflow that reads a live NASA Science RSS feed every day. Keep each run focused on three research items.
Use Google Gemini to produce factual two-sentence summaries. Generate three key takeaways without guessing beyond the supplied feed item.
Save structured notes through managed OAuth2 access to Google Sheets. Replace duplicate-producing appends with Append or Update Row using Source URL as the matching field.
Secret Mission: Added an AI relevance gate with an If branch so only items labeled RELEVANT: YES reach the duplicate-safe digest.

Ready to quiz yourself?