Social PublishingIntermediate

How to Do YouTube Keyword Research with AI

Combine YouTube autocomplete with an AI clustering step to find low-competition search topics your channel can actually rank for.

9 minIntermediate

YouTube is the world's second-largest search engine, and a chunk of your views can come from search long after you publish. The trick is targeting phrases people actually type that big channels have ignored. This guide pairs free autocomplete data with an AI clustering pass to surface those gaps.

What you need

A seed topic for your channel
Access to YouTube search (logged out, to avoid personalized results)
ChatGPT, Claude, or Gemini for the clustering step

Step 1: Mine YouTube autocomplete

Type your seed topic into the YouTube search bar and run through the alphabet: 'home espresso a', 'home espresso b', and so on. Autocomplete reflects real search demand. Copy every suggestion into a plain text file.

YouTube - Search autocomplete

home espresso setup

home espresso under 300

home espresso vs cafe

home espresso beginner mistakes

home espresso machine cleaning

Each autocomplete suggestion is a phrase real people search for.

Step 2: Cluster the phrases with AI

Paste the raw list into your AI tool and ask it to group the phrases by search intent: buying guides, how-to, comparison, troubleshooting. Clustering turns a messy list into a clear content map and reveals which intents you have not covered yet.

cluster-prompt.txt

Here is a raw list of YouTube autocomplete phrases. Group them by
search intent (how-to, comparison, buying-guide, troubleshooting).
For each cluster, suggest one video title that targets the whole
cluster and note the likely difficulty (low/med/high) based on how
broad the phrase is.

<paste your phrase list here>

Claude - Intent clusters

You

Group these YouTube autocomplete phrases by search intent and suggest a title per cluster...

Agent

Troubleshooting cluster (low difficulty): cleaning, channeling, sour shots. Title: '5 Espresso Mistakes Ruining Your Shots (Easy Fixes)'.

The model groups phrases and proposes one umbrella title per cluster.

Step 3: Validate competition by hand

AI suggests demand, but you confirm competition manually. Search each candidate title and look at the top results. If the first page is all channels with millions of subscribers and recent uploads, skip it. If you see thin, old, or low-view videos, that is your opening.

Low view count can be a green light

A search where the top videos have only a few thousand views often means low competition, not low demand. If autocomplete suggested the phrase, people are searching it. A focused video can climb fast.

Step 4: Build a ranked content list

Score each validated topic on demand and competition, then sort. Tackle the low-competition, decent-demand topics first so your channel banks search wins early while it is still small.

Result

You finish with a prioritized list of titles tied to real searches, not guesses. Each entry maps to a cluster of phrases, so a single video can rank for a dozen related queries at once.