Best-performing YouTube videos on a topic in Google Sheets

What it is

This n8n workflow finds the YouTube videos on a topic that did well over the past year and logs them in a Google Sheet. You type the topic in n8n’s chat. The YouTube node searches for it, an HTTP request fetches each result’s statistics, and a Code node works out its age, views per day and engagement ratios. Videos that clear four thresholds go to OpenAI’s gpt-4o-mini, which keeps the ones with an English title on your topic. Each of those becomes a row with eleven columns. Nothing changes on YouTube.

1. You type a topic in n8n’s chat
2. YouTube returns videos on it from the past year
3. Each video’s counts become ratios and a daily view rate
4. Four thresholds and an OpenAI check drop the weak ones
5. The rest are appended to your Google Sheet

Who it suits

YouTubers and strategists who choose topics by what already performs, and agencies that research niches for clients. It gives you a list of proven videos to study. It writes no titles or ideas of its own. Skip it for research in another language, since the AI check drops every title that is not in English.

Where to use it

Any n8n that can reach Google and OpenAI will run it, on n8n Cloud or your own server. The results live in a spreadsheet, so sorting, filters, charts and sharing all happen in Google Sheets.

• n8n Cloud: import the file into a new workflow there
• Your own n8n server: the same file and steps
• n8n’s chat: the panel that opens from the canvas is where you type each topic
• Google Sheets: sort the rows, chart them, or share the sheet with a team

Setup

Get it saves a zip with workflow.json and Mindket’s setup notes in your downloads folder. Its author puts setup at about 13 minutes. Make the sheet first: a tab named YT Topic Ideas with the eleven headers below in its first row.

Topic | Title | View Count | Like Count | Commet Count | Like Ratio | Comment Ratio | Views per day | Engagement Score | Days ago | Youtube URL

1. In n8n, create a workflow, open its menu, choose Import from File and pick workflow.json.
2. Add a YouTube OAuth2 credential and select it in Get Videos and in Get Video Data.
3. In Get Video Data, replace YOUR_YOUTUBE_API_KEY in the query with a YouTube Data API key, or delete that row. The OAuth credential signs the call on its own.
4. Add an OpenAI credential and select it in ChatGPT.
5. Add a Google Sheets credential and select it in Update Data to Sheet, then replace YOUR_SHEET_ID with the ID in your sheet’s address, between /d/ and the next slash.
6. Save, open the chat, type a topic and check the rows that land.

Commet Count is spelled that way in the node. Keep that spelling in the sheet, or fix it in both places.

How to use it

Type a topic of a few words, as you would in YouTube’s search box. The search covers videos published in the last 12 months, ordered by relevance with moderate safe search. Every video that passes lands as a new row. A second run on the same topic adds its rows again, because the sheet node appends and never looks for a match.

Example

You type home espresso. Say one result has 250,000 views, 9,000 likes and 600 comments, and went up 100 days ago. The Code node gives it a like ratio of 3.60, an engagement score of 0.0384 and 2,500 views per day. All four thresholds pass. OpenAI answers Yes for its English title, and this row lands, trimmed here:

{ "Topic": "home espresso", "View Count": 250000, "Like Ratio": "3.60", "Comment Ratio": "0.24", "Engagement Score": "0.0384", "Views per day": 2500, "Days ago": 100 }

For newcomers and experts

If you are new. n8n draws a workflow as boxes joined by lines, and each box, a node, does one job. This one has 11 nodes and no sticky notes. A ratio here is a share, such as likes per hundred views. A credential is a key n8n keeps for you and hands to the node you pick it in. Try a narrow topic first, so the first run stays small.

For experts. Performance Ratio Calculation scores one item at a time, and two of its fields come out as strings with fixed decimals. Filter Bad Videos compares them with loose type validation. All four must hold.

Remove Duplicates compares whole items, so it only drops a video that came back twice with the same numbers. Filter Title & Language holds the model to {"Answer": "Yes or No"} through a structured parser, and the IF after it keeps an exact Yes. The Code node also counts likes per day, which no column takes.

Good to know

The search can use up YouTube quota quickly, and repeat runs add duplicate rows.

• Quota: Get Videos has Return All on, so it pages through every result YouTube offers, and a search costs far more quota than a lookup of one video
• Billing: one OpenAI call for each video that clears the thresholds
• Repeat runs: the same video can land in several rows, so sort or dedupe in the sheet
• Language: the check keeps English titles only, so rewrite its prompt for another language
• Who can run it: the chat trigger is not public, so only people who can open the workflow in n8n can start a run
• Credentials: the JSON holds none, and nothing from its author’s accounts