# Research the moment

Use this when choosing clips from an episode or checking what will matter this week. Start from the audience and the transcript. A trend can raise interest in a good moment; it cannot supply missing substance.

## Build a small evidence set

Write the date, target region, and audience. Take two or three strong candidate ideas from the episode. Search in the audience's own words.

Use three kinds of evidence as needed:

| Evidence | What to seek | Limit |
| --- | --- | --- |
| Current event | Original announcement, study, interview, or report; event date as well as article date | Old news can resurface with a new headline |
| Audience question | Recent comments, search suggestions, forum threads, or questions sent to the show | A loud thread may represent a small group |
| Comparable clip | Public clip on a similar topic, with publish date, format, title, and visible response | Public views cannot prove full-episode conversion |

For a timely pick, seek both a verified event or debate and a sign the target audience cares. Several articles copied from one source count as one signal. For a lasting pick, use repeated questions or relevant past performance. Label weak evidence plainly.

If the Benchmark MCP is connected, inspect its available research operations and use the ones suited to public posts, accounts, or links. Pull comparable episodes and inspect their title-thumbnail pairs. Do not invent private retention data or connector fields. Without it, use public pages and record access limits.

## Judge an outlier fairly

Choose three to five comparable shows. Match the viewer need and format first. Then consider size and guest recognition.

Compare a candidate clip with nearby posts from the same account, in the same format and at a similar age. If the data permits, divide views by the median views of that matched group. Name the group and sample size. Treat the ratio as a research clue.

If only current public totals exist, mark the age mismatch. Do not imply you know the first-week performance of an old upload. Do not compare a Short's views with a two-hour episode's views as if they measure the same thing.

Inspect the actual hook and payoff when accessible. The title alone cannot show whether the clip is complete. Record paid promotion or guest fame as possible factors when known. Do not assume a clever title caused the result.

## Turn interest into a choice

Make a compact table:

| Candidate | Audience question | Source and date | What the clip adds | Shelf life | Decision |
| --- | --- | --- | --- | --- | --- |

Choose among:

- **Post soon:** The event is real, the guest adds a useful view, and the cut is ready. Give a specific review-by date.
- **Evergreen:** The question lasts after the event. Use words that will still make sense later.
- **Hold:** The idea fits but the news context or a claim needs a check.
- **Skip:** The trend reaches the wrong audience or the episode has little to add.

Do not label a past interview as a reaction to a later event. Put the recording date in the post copy when the time gap affects meaning. If a new study is the hook, read what it actually found. Keep its limits in the claim.

## Check freshness before release

Recheck timely hooks before posting. If the news has moved on, change the framing or hold the clip. Avoid words such as "today" in a saved edit unless the release date is fixed.

For *Better Questions*, a sleep-score discussion may connect to a current wearable launch. Verify the launch and the audience debate. If the guest never discusses that device, frame the clip around the habit of checking scores. Do not imply a review of the new product.

Return a few useful links in the task's research note, with the claim each supports. Keep the recommendation separate from the evidence. Say when the choice rests on editorial judgment.