Manthan Patel ("I teach AI Agents and Lead Gen | Lead Gen Man(than) | 100K+ students") recently shared something that caught my attention: "This guy is running AI street interviews as paid ads with 4.5x ROAS" followed by a bold setup: "No creator. No film crew. No budget. Just a prompt + Calico AI."

That combo of performance (4.5x ROAS) and simplicity (prompt-first production) is exactly why his post spread. It is not just about a shiny tool. It is a signal that the creative supply chain for paid social is changing fast, and most teams are still operating like it is 2022.

In this post, I want to expand on what Manthan pointed out: the best brands are already using AI-generated video formats that look and feel like familiar social content, but are produced faster, localized instantly, and tested at a much higher volume than traditional workflows allow.

The shift Manthan is highlighting: distribution now rewards velocity

Manthan listed "11 AI video tactics brands are using to create viral ads right now" and added, "Most marketers have no idea these exist yet." That line matters, because when a format is both unfamiliar and effective, early adopters get an outsized advantage.

Paid social platforms reward two things relentlessly:

Relevance (the ad looks like something people already watch)
Iteration speed (you find the winning angle faster than competitors)

AI video is not automatically better. But it makes iteration so cheap and fast that you can run the kind of creative testing most brands only talk about.

Key insight: AI video is less about replacing creators and more about multiplying creative shots on goal.

The 11 AI video ad tactics, expanded

Here is Manthan's list, with context on why each works and how marketers can apply it responsibly.

1) Street interviews (indistinguishable from real ones)

Street interviews are a performance staple because they feel unscripted, fast-paced, and socially native. The viewer expects imperfection, which paradoxically increases trust. AI makes it possible to generate variations: different interviewers, different locations, different hooks, and different product angles.

What to test:

The opening question (curiosity vs controversy)
The first 3 seconds of reaction
One clear product moment (show, do not tell)

2) Podcast-style product endorsements

Podcast clips work because they borrow authority and long-form credibility, then compress it into a short-form hit. With AI, brands can create "host + guest" style explanations, with a tight problem-solution arc.

What to test:

A strong claim + quick proof
"I used to think X, now I do Y" framing

3) Viral food videos

Food formats have built-in retention: the viewer wants to see the final result. Brands outside food can still use the structure: a satisfying process and a reveal.

What to test:

Process shots that loop seamlessly
A reveal tied to your product benefit (before-after, time saved, cleanup, cost)

4) Try-on videos with zero models

Try-on is inherently product-forward. AI can simulate outfits, cosmetics, or accessories in a way that lets you scale styles, skin tones, and contexts without scheduling shoots.

What to test:

"Day to night" or "work to weekend" transformations
Different personas (minimalist vs bold)

5) Shock hooks (47% higher completion rate)

Manthan mentioned shock hooks and a specific performance claim. Whether the number is your benchmark or not, the principle holds: pattern interrupts increase completion rate. The key is to earn the shock with a real payoff.

What to test:

A surprising statement that is resolved within 5-7 seconds
Visual contradiction (show the opposite of what people expect)

6) Instant localization into any language

This is one of the most practical advantages. Localization used to be expensive and slow, so most brands did "English-first" and called it global. AI lets you match language, cultural references, on-screen text, and even pacing.

What to test:

Country-specific offers
Local slang versus formal tone

7) Product commercials

Classic commercial structure still works: problem, product, proof, offer. AI reduces the cost of producing multiple "polished" variants, which is useful for top-of-funnel prospecting and retail partners.

What to test:

Feature-led vs outcome-led scripts
Different hero shots and end cards

8) POV adventure content

POV is sticky because it puts the viewer inside the story. AI helps generate settings that would be expensive to film. Even B2B can borrow POV: "A day in the life" of a role using your tool.

What to test:

Fast scene changes every 1-2 seconds
One recurring visual motif (the product as a companion)

9) Real estate transformations

Transformation content is universal: messy to clean, old to new, empty to furnished. This is obvious for real estate, but the structure applies to software onboarding, closets, garages, and workflows.

What to test:

Split-screen before-after
A timed reveal ("Watch this change in 10 seconds")

10) Anatomical animations

If your product affects the body (fitness, wellness, medical adjacent), anatomy visuals can explain benefits quickly. The win is clarity: show what changes, where, and why it matters.

What to test:

Simplified visuals that prioritize understanding over realism
One claim, one mechanism, one disclaimer

11) Unboxing videos

Unboxing is ritual content. It signals quality, anticipation, and detail. With AI, you can generate unboxing variants for bundles, limited editions, or localized packaging.

What to test:

ASMR-style pacing vs fast cuts
Emphasizing what is included versus what problem it solves

A practical way to use Manthan's list without getting lost

A common mistake is trying all 11 tactics at once. The smarter approach is to treat these as format buckets and run a focused sprint. Here is a simple system I recommend:

Step 1: Pick 2 formats that match your offer

If your product needs trust: street interviews, podcast endorsements, unboxing
If your product needs demonstration: try-on, transformations, commercials
If your product needs education: anatomical animations, podcast explanations

Step 2: Create a "prompt kit" with variables

Manthan mentioned he compiled prompts in a free PDF and asked people to comment "AI ADS" to get them. Even if you build your own prompts, the idea is the same: treat prompts like reusable templates.

Include variables like:

Audience persona
Hook type (curiosity, shock, social proof)
Setting and language
Offer and CTA

Step 3: Test like a performance marketer, not a filmmaker

Run 5-10 variations per format with tight measurement:

Hook rate (3-second view or thumbstop)
Hold rate (25%, 50%, 95% view)
CTR and CPA

Then keep winners and iterate one change at a time. AI makes this feasible.

The ethics and platform reality check

When Manthan says "indistinguishable from real ones," it is worth pausing. The goal should be "native" and "compelling," not deceptive. If you are generating synthetic people or scenarios, follow platform policies, avoid impersonation, and consider disclosure where appropriate. Long-term trust beats short-term tricks.

Best practice: Use AI to scale storytelling and testing, not to mislead viewers about who is speaking or what is real.

Why this matters for your 2026 ad strategy

Manthan's post is a snapshot of a bigger change: the bottleneck in paid growth is moving from media buying to creative throughput. The teams that win will not just have better targeting or budgets. They will have a repeatable pipeline for generating, localizing, and iterating creatives that look like the content people already binge.

If you take one thing from his 11 tactics, let it be this: pick a proven social format, build a prompt-driven production loop, and test relentlessly until performance tells you what to scale.

This blog post expands on a viral LinkedIn post by Manthan Patel, I teach AI Agents and Lead Gen | Lead Gen Man(than) | 100K+ students. View the original LinkedIn post →