Google gave my team and me early access to Veo within the Google Ads interface, so I ran it through a handful of real client accounts and am reporting back to ya’ll!
Here’s how it works.
You upload up to two static images, Veo generates up to 10 seconds of video from them (roughly five seconds per image if you’re using two), and each generation gives you two variations to choose from. From there, you add two headlines, every 30 characters, drop in your logo and branding details (font family, font weight & brand colour), and the platform assembles a ready-to-serve video ad. It can automatically pull your primary brand colour from your logo. There are also five template options for how the text and transitions are laid out, though these are optional. Music selection is built in, and there are quite a few options to preview; if you’d rather skip the selection process, Google will just choose one for you.
What it lacks is any way to guide the output. You upload, it generates, you pick. That limitation will come up often in what follows.
Here is how testing the tool went for me.
Specialty Grocer
The grocery client produced the strongest results across all the tests I ran.
We tested two images for this client: prosciutto on a plate and a latte. For the prosciutto, the first variation added ambient motion in the background while keeping the subject mostly still; a simple but effective way to bring a static image to life. The second variation introduced human movement, and while it was impressionistic rather than precise, it felt like it belonged in the frame.
The latte performed similarly; both variations captured the foam and milk movement in a way that felt natural, with slight differences in how the motion unfolded. The AI clearly understood what a coffee pour looks like and which movements make sense in that context.
The one thing that stuck with me from this client was the tongs. In one session, a variation generated tongs picking up the prosciutto in a way that looked genuinely great. I tried to recreate it 8 more times in subsequent sessions, but couldn’t get it back. That’s the thing about working without a feedback loop; when something works, you have no reliable way to ask for it again.
Wet Shaving Kits
This client sells wet shaving supplies; think shaving creams, brushes, and grooming oils. We ran two images through Veo: a brush with shaving cream on it sitting next to the product packaging, and a brush sitting in a bowl of shaving cream.
The brush and packaging image produced two variations, both of which were a little puzzling. The first had a hand enter the frame and drizzle what appeared to be orange jelly onto the brush. The second had a hand come in holding another brush, which then appeared to take some cream from the brush in the original image. Both introduced movement, which is technically what we asked for, but neither made sense for the product, and neither would have been usable as an ad without some serious questions about what the viewer is actually looking at.
The bowl shot was a different story and, honestly, a pleasant surprise. In an earlier session, Veo had removed the brush from the scene entirely and replaced it with other items unrelated to the product. This time, it understood what it was looking at; both variations had a hand entering the frame to interact with the cream in a way that actually made sense. It was a meaningful improvement, and it gave me some optimism that the model can get the context right when the image is clear enough.
Mechanical Models
This client is a US distributor for a line of intricate mechanical model kits; think buildable trains, boxes with latches, and highly detailed moving parts. Not your typical e-commerce product to animate.
This one genuinely impressed me. For the train model, Veo generated movement within the model itself, animating it in a way that felt appropriate to the product and what makes it interesting. For another product that featured a box with a latch, it understood the latch’s mechanics and treated it accordingly. It didn’t just add generic ambient motion around the product; it engaged with the product itself.
Consumer products with a clear visual identity and inherent motion logic are where Veo seems to perform best, and this client was a good example. The product tells you what it wants to do, and Veo picked up on that.
What the Experience Taught Me
A few things became clear across all of these tests that are worth calling out separately.
The feedback loop problem is real. When something works, and the prosciutto tongs are the best example of this, you have no way to ask for it again. You can run the same image through eight more times and get eight completely different results. That’s not necessarily a flaw in isolation; generative AI is probabilistic by nature. But inside an ad workflow where brand consistency matters, it creates a practical problem that’s hard to work around. The video generation feature is different than Nano Banana. NB is a separate AI image-generation and editing tool within Google Ads that lets you prompt iteratively, refine outputs, and guide the model toward what you actually want. Veo currently has none of that. You upload, it generates, and you pick from what you get. For a tool sitting inside the same platform, that gap feels significant.
Image quality and subject clarity matter quite a bit. Google’s own best practice guidance recommends high-resolution images with clear subjects and minimal text. The clients where Veo performed best, the grocer and the mechanical models, both had clean, unambiguous source images. The wet shaving tests were more compositionally complex, and the results reflected that; the model understood the context but not always the specifics.
One workflow detail worth noting: the videos don’t go directly into your campaign. They are uploaded to a YouTube channel first, either your own or a YouTube-created video ad storage channel, and then become available as ad assets from there. It’s not a major friction point, but it’s worth knowing before you’re mid-workflow and wondering where your video went.
Finally, and this applies specifically to the uploaded videos: they’re going to a YouTube channel, which means they’re technically public. Something to flag with clients before you start generating.
Who Is This Actually For?
After running Veo across multiple client accounts and product categories, the answer to that question is pretty clear to me.
Consumer product brands with strong visual assets are the obvious fit. Food, lifestyle goods, toys, intricate physical products; anything where the subject has inherent motion logic and the imagery is clean and unambiguous. The grocer and the mechanical models both performed well because the products told Veo what they wanted to do, and it picked up on that. For small businesses in these categories that don’t have a design budget, can’t easily access contractors, and aren’t in a position to commission video production, this tool removes a real barrier.
E-commerce brands should approach it with more caution. The wet shaving results are a good illustration of why. The tool understood the category but not the specific product, and when brand accuracy matters, that gap is a problem. If your creative depends on showing the right product doing the right thing, you need to QA every single output before it goes anywhere near a live campaign.
For lead-gen clients, I’m genuinely skeptical that this format will get you very far. It’s a more fundamental mismatch between what Veo produces and what lead gen creative needs to accomplish. Two static images with generated motion around them doesn’t tell a service story. It doesn’t build trust, communicate expertise, or give a prospective customer a reason to convert. That’s not a limitation that better image selection is going to fix.
The tool is interesting and in some cases genuinely useful. It just needs iteration controls, clearer subject handling, and some kind of feedback mechanism before it becomes something you can rely on consistently across a range of clients.
Here’s a video of what the videos can look like:
Great article on Google Ads! I really liked how you explained campaign structure and targeting. It helped me understand how to optimize ads more effectively.
Very Nice blog