How to Do TikTok Ad Research to Find Winning Ads and Hooks

TikTok ad research means finding ads that are genuinely overperforming, not just ads from big accounts, and pulling out the hook, format, and transcript so you can adapt them to your own product. The key condition: judge every ad against the creator's own baseline reach, not raw view count. A 50K-view video from a creator who normally gets 5K is a stronger signal than a 2M-view video from an account that always gets 2M. Tools like ViralSpy score posts this way, and you can also do a rougher version manually.

Step 1: Define your research goal and niche

Before you search, decide what you're actually looking for, because "TikTok ads" is too broad to produce a usable pattern.

• Category: DTC ecom, apps, TikTok Shop, or brand ads. ViralSpy filters by exactly these segments.
• Product type: physical product, app, service, or content.
• What you need: a hook to test, a format to copy, a creator to hire, or a full creative brief.

If you sell a physical product on TikTok Shop, filtering to TikTok Shop ads only removes a lot of irrelevant noise. If you run an app, filter to app ads.

Step 2: Separate real breakouts from big-account noise

This is the step most people skip, and it's the one that determines whether your research is useful.

Every post should be scored against its creator's normal reach. ViralSpy's breakout library uses this scale:

A post at 12× the creator's baseline is a genuine signal even if the absolute view count is modest. A post with 3.5M views from a creator who always gets millions tells you almost nothing.

If you're doing this manually: open a creator's profile, note their typical view range across recent posts, then flag any post that clearly exceeds it. It's slower, but the logic is the same.

Step 3: Filter by ad type, format, and hook

Once you have a list of breakouts, narrow it to patterns you can reuse.

• Ad type: paid ad, UGC-style, demo-led, or organic.
• Format: POV, demo, before/after, listicle, talking head.
• Hook: the opening line. ViralSpy pulls hooks, formats, and transcripts out for you.

A concrete example from ViralSpy's breakout library: xTool, a laser cutter brand, ran a paid ad with the opening line "I turned my 3D printed animals into action figures." The reusable hook template is "I turned my [product] into [new format]." The format was curiosity-driven, demo-led, UGC-style. That template is portable to almost any physical product.

Another example: Suno, at 12× breakout, used the hook "POV: your new church group turns out to be a cult." That's a POV curiosity hook, not a product demo — useful if your product fits narrative or entertainment framing.

Step 4: Extract hooks, formats, and transcripts

For each breakout ad you keep, write down three things:

1. The hook — the first 1–3 seconds of spoken or on-screen text.
2. The format — how the video is structured (demo, POV, testimonial, skit).
3. The transcript — what's actually said, so you can see the full argument the ad makes.

Then rewrite each one for your product. Keep the structure, swap the specifics. "I turned my 3D printed animals into action figures" becomes "I turned my [your product] into [a new use case your customer cares about]."

Step 5: Validate before you commit

One breakout ad can be luck. A pattern that repeats is a signal.

• Does the same hook template appear across multiple creators?
• Does it keep working across multiple weeks?
• Does it show up in more than one brand in your category?

ViralSpy's library defaults to the last 14 days, which is a reasonable validation window — long enough to see repetition, short enough to stay current. If a hook only appears once, treat it as a test candidate, not a proven formula.

Step 6: Turn research into a brief or a test

Your research output should be one of two things:

• A creator brief: the hook, the format, the transcript structure, and the product angle, handed to a creator so they know what to make.
• Your own ad test: the same elements, shot in-house.

ViralSpy's workflow maps to this directly: discover what works, recruit creators with a brief, then ideate with an agent that turns this week's patterns into fresh ideas. You can also pull the data via API or from Claude Code and Codex if you want to feed it into your own process.

Common sticking points

• Sorting by views instead of breakout score. This is the most common mistake and it buries the real signals.
• Copying the ad instead of the pattern. The hook template transfers; the exact script usually doesn't.
• Skipping validation. A single 100× post from a small creator might be an outlier, not a repeatable format.
• Researching without a goal. If you don't know whether you need a hook, a creator, or a format, you'll collect data you can't act on.

ViralSpy offers a 24-hour trial at $0, then $99/month, cancel anytime — enough to run this process end to end once and see whether the breakout scoring surfaces anything your manual search missed.