Your First Cycle: Steal. Create. Launch.

Written By Yuri P.

You bought CopyThief to ship more — not to research more. This is the 10-minute walkthrough that turns the 5 apps into one cycle.

If you follow this end-to-end on day one, you’ll have a live test in Meta before your first session ends.

The cycle in one diagram

  1. Discover → Find what’s already winning in your market

  2. 2. Save → Steal it into Swipes (boards + AI Analysis)

  3. 3. Track → Promote winners’ brands to Raven (autopilot from here)

  4. 4. Generate → Lab turns 1 validated input into 20 variants

  5. 5. Launch → Co-Pilot batches them into Meta in 3 minutes

Step 1 — Find what’s winning (Discovery)

  1. Open Discovery

  2. 2. Search your niche or a specific brand

  3. 3. Filter to last 30 days + Longest Running — this surfaces validated ads, not just live ones

  4. 4. Browse 10–20 results. Stop there. You’re identifying 3–5 patterns worth modeling.

If you spend more than 10 minutes here, you’re researching. Move to step 2.

Step 2 — Steal into Swipes

For each pattern worth modeling:

  1. Click the ad → open the drawer

  2. 2. Note the AI Analysis breakdown (hook / angle / creative / body / CTA)

  3. 3. Click Save to Swipes → assign to a board

  4. 4. Tag at save-time (curiosity-hook, ugc-style, founder-pov)

  5. 5. Rate 4★ or 5★ if it’s a “ship this” candidate

Save 5 max in this session. More = you’re hoarding.

Step 3 — Track the winners (Raven)

For the brands behind your top 2–3 saves:

  1. Click the brand avatar on any of their ads in Discovery

  2. 2. Click Track Brand in Raven

  3. 3. Pick a folder (direct-competitors or inspiration)

  4. 4. Confirm

From now on, Raven captures every new ad they launch — twice a day, no clicks required.

Step 4 — Generate variants (Lab)

Pick ONE of your 4★+ saves as the input. Don’t try to generate from 5 references at once on your first cycle.

  1. Open the saved ad → click Send to Lab

  2. 2. In Lab, the ad pre-fills as your input

  3. 3. Add your 30% — your offer details, your product specs, your voice notes

  4. 4. Set Creativity Amplifier to 5 (middle ground — adjust later)

  5. 5. Pick output type: Copy variants (start with this) or Image/Video

  6. 6. Click Generate → 20 variants in ~60 seconds

Review the output. Star 5 that match your gut + your data.

Step 5 — Launch in Meta (Co-Pilot)

  1. From Lab, click Send to Co-Pilot with your 5 starred variants

  2. 2. Pick (or build) a campaign template — for first launch, use Test — Bid Cap Broad

  3. 3. Map ads → ad sets (Co-Pilot auto-suggests groupings)

  4. 4. Review the structure once

  5. 5. Click Launch

Ads hit your Meta account in under 5 minutes. Co-Pilot follows your naming convention automatically.

What “done” looks like for cycle #1

  • 5 ads tagged + rated in Swipes

  • - 2–3 brands tracked in Raven

  • - 1 Lab generation with 5 starred variants

  • - 1 live test in Meta

Total time: 30–45 minutes for the first cycle. After 2–3 cycles, you’ll do it in 15.

The Eternal Researcher trap

You will be tempted to: save 50 ads instead of 5, track 30 brands instead of 3, generate 100 variants and never pick, or say “I’ll launch tomorrow when I have more ammo.” That’s the trap. CopyThief was built to break the trap, not feed it. Stop hoarding what works. Start launching it.

What to do after cycle #1 ships

Wait 5–7 days for ad data to come in. Then:

  1. Check the test in Meta — kill underperformers

  2. 2. The survivor(s) tell you which component worked (hook? creative? angle?)

  3. 3. Run cycle #2 modeling that specific component

  4. 4. Keep narrowing until you find a hook+creative+LP combo that scales

FAQ

Do I have to do this in order? The first cycle, yes. After you’re comfortable with each app, you’ll skip around.

How often should I run a cycle? Weekly is the sweet spot. Daily is overkill. Monthly is too slow — competitors saturate angles in 2–3 weeks.

What if my first cycle’s test loses money? That’s the cost of the data. The point is to learn what doesn’t work as fast as what does. Cycles 2–4 use what cycle 1 killed.