How To Scale AI Video Ads For Small Business Clients: One Storyline For TikTok, Reels & AI Search
All Paid Media PPC News Social MediaAdvertising Video Advertising Columns Ask A PPC ExpertNEW PPC Pulse Guide AI Max Invalid Traffic Doubled, Weeks Before Black Friday Invalid traffic on AI Max search doubled to 5.28%. Get the channel-by-channel retail benchmarks before Black Friday. Digital Marketing All Digital Marketing Analytics & Data Ecommerce Lead Generation Content Marketing Social Platforms Google YouTube Reddit LinkedIn TikTokNEW WordPress Other CMS Platforms Guide Your ChatGPT Calls Are Hiding in the 'Direct/Other' Bucket Calls from ChatGPT, Claude, and Perplexity land in 'direct/other.' This checklist shows six steps to track them as their own source. SEJ Live Webinars Rundowns PodcastNEW Ebooks All Resources Webinar Fix The Product Pages AI Shopping Agents Skip In this webinar, learn to rewrite spec-sheet product copy into the words shoppers use, and make product pages readable to AI. Pedro Cerdeira October 8, 2026 ⋅ 9 min read Pedro Cerdeira Chief Technology Officer at Leadzai Bio Follow 244 READS Editorial Reviewed: Jennifer McDonald Editorial Reviewed This post was sponsored by Leadzai. The opinions expressed in this article are the sponsor’s own. Not long ago, most of our conversations with agencies opened with the same question: should we even be doing TikTok for our small business clients? That question has quietly disappeared. Agencies now come to us asking how quickly they can bring their entire SMB portfolio onto the platform, and the first thing they want to see is the video. Both comments point to the same shift. For agencies, the bottleneck on TikTok was never the ad platform. It was the video, which is becoming an increasingly important part of digital advertising. 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” (SEJ). Google Search Console now reports on TikTok, Instagram, and YouTube posts appearing in Search and Discover (SEJ), which means Google possibly treats social video as search inventory. TikTok videos aren’t cited by ChatGPT or AI Overviews today. TikTok is where people search. YouTube is what AI cites. A storyline that runs on both covers both. From here, video picks up a second important task for brand visibility. TikTok has more than a billion users, and it’s no longer just an entertainment platform. In an Adobe Express survey reported earlier this year, 49% of U.S. consumers said they have used TikTok as a search engine. For a local plumber, salon, or restaurant, that represents another source of discovery beyond Google. 86% of advertisers said they are using or plan to use generative AI to build video ad creative, and small and mid-sized brands are adopting it faster than enterprises. Now, you have to keep up with competition’s video scale for each of your clients. Four numbers give you the target. Work through them per account, then total. Bandwidth. TikTok runs on video, and most SMBs don’t have the resources to produce it consistently. A professionally produced ad has traditionally involved scripts, shoots, and edits that can take weeks to complete for a single client. For an agency managing hundreds or thousands of SMB accounts, producing video one client at a time simply doesn’t scale. Volume. Generative AI has helped solve the volume half of that problem, but not the quality half. In a WARC study conducted with TikTok and 400 marketing leaders, 90% said they use AI as a core creative tool, yet only 45% reported significant improvements in quality. The report argues that the new creative advantage comes from relevance, not output. Relevance. Volume without relevance produces ads that could belong to any business. The fix is a pipeline built around the advertiser’s real context from the first step. The agencies scaling fastest treat video as a pipeline, not a production. Each step has one job, and the order matters. Image: pipeline diagram Caption: video ad pipeline · 3 steps and a feedback loop Start with a multimodal LLM, a model that can analyse images and text together. Give it what the business already has: a few real photos of the storefront, products, team, or finished work, along with the logo and key business information. That means what they sell, where, who they serve, and what makes them different. The model’s job is to write a storyline, a script and visual guideline that becomes the ad’s single source of truth: It also maps each scene to a reference image and identifies the key elements that must never change. This is where consistency is decided. If the guideline specifies this restaurant, with this dining room and these dishes, every downstream step inherits it. Skip this step, and you get a generic video about “a restaurant” that could belong to anyone. It’s also a learning step. When the agency or business owner asks for a punchier hook or a warmer tone, that feedback refines the storyline, and every later version improves. Next industry insight would be nice here: when does this go live? The reactions to those videos have been remarkably consistent. “It’s amazing how real the ad looks based on contextual images. It feels like it was professionally made,” one agency partner told us after reviewing ads built from a handful of a client’s own photos. Another described the part that surprised them most: “The storyline keeps getting better and better as we provide feedback.” Next, pass the storyline and the same images to a video model. Here, the field narrows quickly. Very few models accept images as references for generation rather than text prompts alone. Current options include Google’s Veo 3.1 and Gemini Omni, ByteDance’s Seedance, Kling, and Alibaba’s Wan. For SMB advertising, reference images are non-negotiable. Text-to-video can produce stunning footage, but a video advertising your restaurant that shows a restaurant that isn’t yours is worse than no video at all. For a local business, the real place is the product, so the context and visuals have to feel real. Generate in vertical 9:16, at 15 to 20 seconds. In our pipeline, each video is ready in minutes. That changes how agencies work: review, feedback, and a new version can all happen in the same sitting. Build for model churn. Don’t build your process around a single model. In recent months, OpenAI shut down the Sora 2 video API with no direct replacement, Google retired Veo 2 and Veo 3.0, and new models such as Gemini Omni and Seedance 2.5 arrived. Any agency that hard-wired one vendor has already had to rebuild. The durable assets are the ones you own: storyline guidelines, brand inputs, feedback history, and a library of hooks that work. Treat the video model as a component you can swap. Finally, turn the video into an ad. Generate the voiceover from the script with a text-to-speech model. Leading voice models now cover 70+ languages, allowing one storyline to serve a client in Lisbon, Zurich, or Dallas with native-sounding narration. Then add captions in the same language. TikTok suggests five to 10 words of on-screen text per second, and captions help carry the message for anyone scrolling with or without sound, while making the video more accessible. AI labeling is either set by the advertiser at upload or applied automatically by the platform. Many video models embed invisible watermarks, such as Google’s SynthID, that platforms can read to identify AI-generated video. Make disclosure a default setting in the workflow, not an afterthought.
Source: Search Engine Journal
This article has been carefully curated and reformatted for educational and informational purposes. Full credit goes to the original publisher.
📚 Visit more helpful articles on Joab Peters Blog
No comments