Agencies have stopped asking whether small business clients belong on TikTok. The question now is how quickly the whole SMB portfolio can get there—and the first thing they want to see is the video.
That shift carries a practical consequence: the bottleneck is no longer the ad platform. It is video production at portfolio scale, and video now pulls double duty in social feeds and search results.
Video is becoming search inventory
Google has adjusted ranking to surface more short-form video. Liz Reid, Google’s VP of Search, said users “are going to short-form video, they are going to forums, they are going to user-generated content a lot more.” Search Console now reports on TikTok, Instagram and YouTube posts appearing in Search and Discover, which suggests Google may treat social video as search inventory.
TikTok videos are not currently cited by ChatGPT or AI Overviews. The split to plan around: TikTok is where people search; YouTube is what AI cites. One storyline that runs on both covers both surfaces.
That search behaviour is already measurable. An Adobe Express survey reported earlier this year that 49% of U.S. consumers have used TikTok as a search engine. For a local plumber, salon or restaurant, that is another discovery channel beyond Google.
The annual video math
Leadzai’s planning starts with four numbers per account, then totals them across the client base.
- Start with four creatives per campaign, inside TikTok’s recommended three to five per ad group.
- Set a quarterly refresh cycle, or sooner when click-through or conversion rate drops.
- Multiply by active accounts: four creatives × four refreshes × 300 clients is 4,800 videos. At 500 clients, it is 8,000.
- Hold the workflow to that number. If it cannot hit it at consistent quality, fix the pipeline before adding accounts.
If that number feels high, that is the point. The agency model breaks when video is still treated as a one-off production.
A three-step pipeline that protects relevance
Step 1: Build the storyline from the client’s real context. Start with a multimodal LLM that can analyse images and text together. Feed it a few actual photos of the storefront, products, team or finished work, plus the logo and basic business facts. The output is a storyline with a hook, problem, narrative and call to action, with each scene mapped to a reference image. Consistency is decided here.
Step 2: Use a video generator that accepts reference images. Very few models do this well. Current options include Google’s Veo 3.1 and Gemini Omni, ByteDance’s Seedance, Kling and Alibaba’s Wan. Text-to-video alone can produce impressive footage, but an ad for your restaurant that shows someone else’s restaurant is worse than no ad. Generate vertical 9:16 clips of 15 to 20 seconds, and build for model churn rather than locking into one vendor. OpenAI shut down the Sora 2 video API, Google retired Veo 2 and Veo 3.0, and new models arrived within months.
Step 3: Add voiceover and captions in the client’s language. Leading text-to-speech models cover 70+ languages, so one storyline can serve Lisbon, Zurich or Dallas. TikTok suggests five to ten words of on-screen text per second, and captions carry the message for sound-off scrolling. Make AI disclosure a default workflow setting, not an afterthought.
The relevance warning: in a WARC study with TikTok and 400 marketing leaders, 90% said they use AI as a core creative tool, but only 45% reported significant quality improvements. The new creative advantage comes from relevance, not raw output.
One storyline, every channel
The same foundation powers TikTok, Reels, YouTube Shorts and Performance Max. TikTok needs native, creator-style creative with sound on. Reels should work without sound, with captions telling the story. YouTube Shorts now average 200 billion daily views according to YouTube’s 2026 outlook, and for Performance Max you should supply 16:9, 1:1 and 9:16 assets yourself—otherwise Google may auto-generate video featuring a product that does not match the landing page.
What to do next
The agencies best positioned to scale will not necessarily own the newest model. They will own the best inputs, the tightest feedback loop, and a pipeline that turns each client’s real world into a video worth watching. People bring creativity; the pipeline makes it scale.
Source: Search Engine Journal
