A trend scan is a short, repeatable check of native discovery surfaces, selected niche accounts, and trustworthy industry sources. Its purpose is to collect dated evidence for content decisions—not to forecast virality.

By the end of this lesson, you will have a five-day trend decision log, clear adapt / watch / reject criteria, and a weekly AI rollup prompt. The daily routine stops at 20 minutes even when the feeds keep offering more.

// concept

Build a Fixed 20-Minute Source Route

Choose the sources before opening an app. A creator who decides where to look while scrolling will follow whatever the recommendation system serves. A fixed route makes this research: the same source types, the same order, and the same stop time.

Time	Source lane	What to inspect	What to capture
0:00–6:00	Native discovery	YouTube Studio's Trends tab, TikTok Creative Center, and the Instagram Explore or search surface available in your current app version	Repeating format, topic, audio, search phrase, or audience question
6:00–11:00	Niche accounts	Five public accounts selected in advance: two peers, two accounts slightly ahead, and one adjacent niche	A pattern appearing across accounts—not somebody's script, joke, thumbnail, or footage
11:00–15:00	Industry sources	Two official or primary sources for your niche, such as a regulator, product vendor, trade body, or original announcement	Dated facts that may make a topic timely; save the original page
15:00–19:00	Log and decide	Enter no more than three candidates	Source, date, format, topic, evidence, score, and decision
19:00–20:00	Close	Set the next recheck date for anything marked watch	Close every discovery tab and app when the timer rings

Interface names change, so confirm menu paths on official help pages. If a Trends or Creative Center view is unavailable, use ordinary platform search and public discovery; do not buy a plan for this routine. When the timer rings, stop—even halfway through a lane. The scan feeds the log; it does not borrow time from production.

// concept

Record Evidence Before Deciding

Create the log in a spreadsheet or document you control rather than relying on in-app saves. Use one row per candidate:

Field	What belongs there
Date observed	Local scan date
Format	Reusable structure: “question → demonstration → result,” not a copied caption
Topic	Subject the audience would receive
Source	URL, screenshot filename, native surface, or primary document
Evidence	What you observed, including independent repetition
Niche-fit score	1–5 against the stated audience and promise
Decision	Adapt, Watch, or Reject
Reason / next check	Why, what still needs verification, and a recheck date if watching

Apply the same decision rules every day:

Adapt: fit is 4–5; two independent observations or one clear native-search signal exist; claims are verified; rights and context are safe; and you can make an original version.
Watch: fit is at least 3, but evidence appears once, local relevance is unclear, or a claim needs a primary source. Set a recheck date.
Reject: it is off-niche, unverified, sensitive, copy-dependent, privacy-invasive, or beyond production capacity.

The score organizes decisions; it does not predict performance. AI may group your observations, but it cannot convert a hunch into verified evidence.

// concept

Turn Five Days Into One AI-Assisted Rollup

After five scans, paste only that week's rows into this prompt. Redact private information.

// prompt — copy me19 lines

You are organizing my weekly trend log. Use only the rows inside <trend_log>.
Do not claim that a trend is growing, popular, or likely to perform unless the rows
contain direct evidence for that wording. Do not add trends from your memory.

My niche: [specific audience + content promise]
My production limits: [formats, hours, language, tools]

Tasks:
1. Group candidates that share the same FORMAT engine, even if topics differ.
2. List repeated TOPICS separately from repeated formats.
3. For every group, cite the row dates and source fields that support it.
4. Recommend ADAPT, WATCH, or REJECT using my criteria.
5. Flag claims, permissions, cultural context, or rights that need human checking.
6. Produce one original content brief for each ADAPT recommendation.

Output columns: pattern | supporting rows | evidence limit | decision | next action.
<trend_log>
[paste the five-day log]
</trend_log>

If the output says “trending” from one observation, correct it to “observed once” and choose watch.

// worked_example

Worked Example

This is a hypothetical sample, not current trend evidence. A Karachi home clothing seller wants practical short-form content about washable everyday wear.

Date	Candidate: format + topic	Evidence saved	Fit	Decision
Sample Mon	Question card → close-up test: “Will this colour bleed?”	Two public niche-post URLs saved as M1/M2; both use a question before a demonstration	5	Adapt: use the seller's own fabric and test footage
Sample Tue	Fast outfit transition using a specific audio clip	One native discovery result saved as T1; no repeat in the fixed accounts	3	Watch: recheck Friday; confirm music rights in-app
Sample Wed	Packing-table sequence: order → fold → care card	Three independently saved public examples W1–W3; format repeats, scripts differ	4	Adapt: film an original packing process with no customer label visible
Sample Thu	News-style post claiming cloth prices will rise	One unsourced repost H1; no original announcement located	2	Reject: unsupported price prediction
Sample Fri	Comment-on-screen → answer: summer fabric care	Two recurring questions from the seller's own redacted comments F1/F2	5	Adapt: answer the actual question without exposing usernames

Running the weekly prompt on those rows could produce this sample output excerpt:

// prompt — copy me16 lines

Pattern: Question-led proof
Supporting rows: Mon (M1/M2), Fri (F1/F2)
Evidence limit: Repetition appears in this sample log; no performance data is supplied.
Decision: ADAPT
Next action: Draft two original briefs—fabric colour test and summer-care answer—using
the seller's own footage and redacted first-party questions.

Pattern: Fast transition
Supporting rows: Tue (T1)
Evidence limit: Observed once; local relevance and music rights are not established.
Decision: WATCH until Friday's recheck.

Candidate: Price-rise claim
Supporting rows: Thu (H1)
Evidence limit: Repost only; no primary evidence.
Decision: REJECT.

In the bad first attempt, the creator opened an endless feed, saved unrelated clips, and recorded no dates. The fix was a written route, a 20-minute alarm, a three-candidate limit, and a completed row before any save counted. AI entered only after evidence existed.

// failure_cases

Failure Cases to Diagnose

7 cases to diagnose

Topics have no formats.
“Summer” does not explain the content engine; add a structure such as “myth → test → proof.”
One post is called a trend.
Mark it watch until another independent signal appears.
A summary replaces the source.
Find the original announcement or reject the factual claim.
Scanning exceeds 20 minutes.
Shorten the account list; never skip the evidence field to catch up.
AI invents momentum.
Delete “rising” or “popular” unless dated rows support it.
Adapt means copy.
Keep only the format engine; replace script, footage, hook, examples, and treatment.
Everything becomes adapt.
Reapply capacity, rights, niche-fit, and sensitivity criteria.

// pakistan_angle

Pakistan Angle

A Roman-Urdu format recurring in Karachi accounts is not evidence for an Urdu-first Rawalpindi audience or a Pashto-first Khyber Pakhtunkhwa audience. Record the observed city and language when visible; otherwise write “unknown.” Never treat one city's feed as a national sample.

For patchy connectivity and load-shedding, open lightweight official pages first, save URLs instead of videos, and use compressed screenshots only when needed. Scan video on Wi-Fi and update the log offline. Redact phone numbers, delivery labels, CNIC details, and WhatsApp identities; obtain permission before featuring a person or message.

Religion, politics, disasters, and sectarian topics require separate sensitivity and fact-check review. For PKR prices, JazzCash, easypaisa, Daraz, taxes, or regulations, use the current official source on the scan date—not another creator's claim.

// hands_on

Hands-On Exercise

6 steps

Build the running log and complete five daily scans:

Name one narrow audience and promise. List three discovery surfaces, five fixed accounts, and two primary industry sources.
Create the nine-field log and define your 1–5 niche-fit scale.
On five days, follow the route with a 20-minute timer; enter at most three candidates daily.
Attach a URL, dated screenshot, or first-party observation to every candidate.
Choose adapt / watch / reject; every watch row needs a recheck date.
Run the weekly prompt, correct claims beyond the evidence, and save the rollup. Done means the file contains five dated scan sessions, traceable evidence, decisions with reasons, and one corrected weekly rollup—and no scan exceeded the hard time cap.

// completion_rubric

Completion Rubric

6 checks — tick as you verify

0/6

// sources

Sources

4 official sources — check every claim yourself

// check_yourself

Check yourself

4 questions · answers and options are taken word-for-word from this course

0/4

1 / 4 · diagnose
Your work shows this failure mode: “Topics have no formats.” The lesson describes it like this: ““Summer” does not explain the content engine;.” What does the lesson tell you to do about it?