Search and discoverability
Some content is found by recommendation. Some is found because someone typed a question. This skill is for the second kind, and for making recommended content easier to understand.
What this skill does not have
No search-volume numbers. Showrunner does not include a keyword database. Say so. Use what the person can see themselves (search suggestions as they type, "people also ask", related searches, the videos already ranking), a web search tool if one is available, or a connected keyword tool. vidIQ says its MCP connector works with Claude, ChatGPT and Cursor on every plan including Free (facts_search id=tooling.vidiq.mcp; low confidence, and Showrunner is not affiliated with vidIQ). If it is connected, use its keyword and competitor data and name it as the source.
Ask first
- The topic, and the one outcome someone wants when they search for it.
- The platform (search works differently on each one).
- Whether the person has search or traffic data from their own channel. Their own data beats any guess.
Method
- List the questions. Write ten plain-language searches a beginner would type. Include the exact words they would use, not the expert terms.
- Mark the intent of each: learn, fix, compare, choose, or just watch.
- Check what ranks. Search each phrase and note what appears. If the top results are huge channels with the exact answer, pick a narrower phrase (add the audience, the tool, the year, the problem).
- Pick one primary phrase per piece and one or two secondary ones. One piece, one main target.
- Place them.
| Where | What to do |
|---|---|
| Title (YouTube) | Primary phrase near the start, written for a person, within the 100-character limit |
| First line of caption or description | Primary phrase in a normal sentence |
| Spoken opening | Say the phrase out loud in the first sentence or two |
| On-screen text | Show it once, big |
| Description (YouTube) | A real summary of two or three sentences that includes the phrase and the secondary ones; chapters with descriptive names |
| Hashtags | A few relevant ones. See the notes below |
Platform notes (with status)
- YouTube. Titles up to 100 characters, descriptions up to 5,000, and with more than 15 hashtags YouTube ignores all of them (all official). YouTube has not published its ranking signals (
youtube.ranking.unpublished), so what you control is packaging and retention. - TikTok. Search is described as using caption keywords, hashtags, on-screen text and spoken words. Low confidence (
tiktok.search.inputs). Cover all four. - Instagram. Mosseri said he would not think of hashtags as a way to get more distribution (
instagram.hashtags.distribution, reported). Put keywords in the caption and on screen, and keep tags few. - Google. Not covered by the facts database. Ask the person to check their own search results.
Check
Tool title_score with keyword set, and caption_check with keywords (sr-title-thumbnail, sr-caption). CLI:
python scripts/sr.py title --keyword "junior developer interview" --file title.txt
python scripts/sr.py caption --file caption.txt --platform youtube --keywords "junior developer interview"
Output
A table the person keeps:
Phrase | Intent | Platform | Where it goes | Notes
Gate
approve. Then the pieces go through the normal path (sr-title-thumbnail, sr-caption).
Do not
- Stuff keywords. Write for the person first.
- Promise ranking. Nobody can.
- Use a trending tag that has nothing to do with the post.