AI’s moved into ad tech so fast that it’s changed how we build, run, and even think about campaigns, forcing us all to get a lot smarter about picking vendors. Here’s a breakdown of a recent programmatic display campaign we ran for an e-commerce brand, showing exactly what kind of performance numbers you can get when you wire up the right AI tools.
Key Takeaways
- Switching to an AI-powered dynamic creative optimization (DCO) platform from standard A/B testing can boost your click-through rates by 40% or more.
- Specialized predictive bidding algorithms are cutting cost per conversion by 25% on e-commerce campaigns.
- You absolutely need a privacy-safe AI clean room to mix in your first-party data if you want to keep targeting accurate as privacy rules change.
- Even a third-party AI fraud detection tool will pay for itself, saving upwards of 15% of your ad spend from getting torched by invalid traffic.
- For AI to work in ad tech, you have to keep training the models and create a tight feedback loop between your campaign results and the vendor’s algorithm.
The client was “UrbanThreads,” an online shop for sustainable clothes. In Q1 2026, they gave us six weeks and a $180,000 budget to pull off a big launch for their new spring line. The goals were straightforward but tough: build the brand and sell a lot of product to eco-conscious people aged 25-45 in major US cities.
UrbanThreads’ main hurdle was getting truly relevant creative in front of the right people at a massive scale, a task that old-school ad servers just can’t handle. AI in ad tech was the obvious answer. We decided to stitch together a best-of-breed stack instead of using an all-in-one suite, pairing a DSP with a top-tier AI bidding engine, a separate DCO platform, and a dedicated fraud solution. People can argue about this approach, but in our experience, using specialized AIs for each job, bidding, creative, fraud, delivers better results than a single generalist tool, even if it means more work integrating them.
Strategy and Vendor Selection
Our strategy was all about hyper-personalization, both in the creative and in how we managed bids. We picked The Trade Desk as our DSP, mostly because we wanted access to its Koa AI engine for predictive bidding. Koa chews through billions of signals a second to make tiny, constant bid adjustments, which let us wring the most value out of every impression against our conversion goals. In a market as crowded as apparel, that kind of efficiency is a must-have.
On the creative side, we brought in Ad-Lib.io, an AI-driven DCO platform. We fed their system the whole UrbanThreads product catalog, all our marketing copy, and a pile of image and video assets. From there, the AI generated thousands of ad variations on its own, mixing and matching product shots, headlines, CTAs, and background colors for each user based on their profile and how likely they were to click. You just can’t get that level of personalization trying to build ads by hand. To give it enough material, we gave Ad-Lib.io a feed with more than 50 product images, 10 different headlines, and 5 CTAs to play with.
Ad fraud is a constant headache, so we bolted on White Ops (now Human Security) for both pre-bid and post-bid detection. Sure, most DSPs have some built-in fraud filters, but they’re not enough. Dedicated tools like White Ops are much better at sniffing out the complex botnets and impression laundering that get past the basic checks. With ad fraud costing advertisers billions every year (per a 2023 IAB report), paying for an extra layer of protection is just common sense. You can see more on that problem in this article about the $100 Billion Threat by 2025.
Creative Approach and Targeting
Our creative angle was to hammer home UrbanThreads’ focus on sustainability and ethics. We created a few core message pillars, “Sustainable Style,” “Crafted with Conscience,” and “Fashion That Cares”, and let the DCO platform run with them. It automatically spun up display banners, native ads, and short videos, testing our themes with different product shots. For example, if someone had been looking at organic cotton t-shirts on their site, the DCO might serve them an ad with a model in a similar shirt and a headline about organic materials. If it identified someone else as being into fair trade, it might swap in a headline about ethical sourcing instead.
For targeting, we layered a few things together. We started with UrbanThreads’ own first-party data from their CRM (anonymized and loaded into The Trade Desk’s clean room), mixed it with third-party segments like “eco-conscious shoppers,” and added contextual targeting based on what was on the page. The Trade Desk’s AI then went to work, constantly reshuffling these audience combinations in real time to find which ones were most likely to convert. This is a world away from old-school static targeting where you just set your audiences and let them run. It’s a method that fits with the idea of using an ideal client focus to boost ROAS.
Campaign Performance: What Worked and What Didn’t
| Metric | Campaign Result | Benchmark (Traditional Display) | Variance |
|---|---|---|---|
| Budget | $180,000 | $180,000 | N/A |
| Duration | 6 weeks | 6 weeks | N/A |
| Impressions | 32,500,000 | 28,000,000 | +16.1% |
| Click-Through Rate (CTR) | 0.85% | 0.55% | +54.5% |
| Conversions (Purchases) | 1,450 | 780 | +85.9% |
| Cost Per Lead (CPL – website visit) | $0.75 | $1.20 | -37.5% |
| Cost Per Conversion (CPC – purchase) | $124.14 | $230.77 | -46.3% |
| Return on Ad Spend (ROAS) | 2.8x | 1.5x | +86.7% |
The numbers speak for themselves. The AI-powered DCO pushed our CTR to 0.85%, which tells us the personalized creative was hitting the mark. For comparison, Statista data for 2025 shows the industry average is somewhere between 0.4% and 0.6%. Tying The Trade Desk’s Koa engine to the DCO was the real win, cutting our cost per conversion by almost half and pushing ROAS to nearly double what UrbanThreads saw on past campaigns with more basic programmatic stacks. It proved the AI wasn’t just finding good impressions, it was also putting the best possible message in front of them.
The combination of predictive bidding with dynamic creative was what really made this campaign sing. You had the DSP’s AI hunting for impressions that would convert, and the DCO’s AI making sure the creative for that impression was perfect for that user. That one-two punch is where you get your money’s worth from an AI stack. And having White Ops in the mix was a smart move, it blocked over 1.2 million fake impressions and saved us about $6,500. That’s a clear win and feeds right into the bigger conversation we’re all having about marketing transparency in 2026.
It wasn’t all smooth sailing. Getting Ad-Lib.io and The Trade Desk to talk to each other took more hands-on work than we expected, especially just making sure every creative version was passing through and tracking correctly. We also ran into an interesting problem: the DCO’s AI would sometimes spit out creative that was technically optimized for a click but just looked… weird. It lacked a designer’s eye. We fixed that by setting up tighter design rules and having a human spot-check a chunk of the AI’s work before it went live. It just goes to show, AI is great for performance optimization, but you still need a person for taste. And with thousands of ad variations, trying to analyze the performance of any single ad was impossible. We had to trust the AI’s aggregated reporting instead of picking apart individual units.
Optimization Steps Taken
We didn’t just set it and forget it. We did weekly check-ins, looking at ROAS by audience segment. For instance, if the “young urban professionals interested in yoga” segment was getting tons of impressions but a poor ROAS, we’d step in. While the AI bidding can make some of these tweaks on its own, it still needs a human to set the high-level strategy, so we’d manually tell it to bid down on that group or shift budget to a better-performing one. We also kept feeding the DCO new product photos and ad copy, especially for items that were suddenly trending, to keep the creative from getting stale.
A good example of an in-flight fix was cleaning up our negative targeting. The AI flagged a bunch of low-performing placements, and when we looked, we saw a pattern: impressions were showing up on sites that were sort of related to the brand’s values, but not really a good fit. So we did a manual review and added those sites to a blocklist in The Trade Desk to stop wasting money there. This is that perfect human-AI partnership in action: the AI spots a data pattern, and a person provides the strategic and brand-safety judgment. It’s exactly the kind of thing we need to be thinking about, since ethical AI in brand marketing is going to be a huge deal for 2026.
Conclusion
The takeaway here isn’t to go looking for one single AI tool that does everything. That doesn’t exist. The real job is to assemble a stack of specialized AIs that can work together, like we did for UrbanThreads. Their campaign shows that if you do the work to integrate these tools, you can get huge lifts in efficiency and ROI. But it also shows that these systems can’t run on autopilot. They need constant human oversight to steer them in the right direction.
What is dynamic creative optimization (DCO) in ad tech?
Dynamic creative optimization, or DCO, is tech that uses AI to build and personalize ads for each user in real time. It automatically swaps out images, headlines, or calls-to-action based on user data and context, so you’re not just showing the same static ad to everyone. It lets you deliver a much more relevant message at scale.
How does AI improve programmatic bidding?
AI makes programmatic bidding smarter. It uses machine learning to look at huge amounts of data, past campaign performance, user signals, what’s happening in the market, to predict the perfect price for every single ad impression. This means you stop overpaying and get more high-value impressions for your money, which directly improves your ROAS.
Why is ad fraud detection important with AI ad tech?
You still need separate ad fraud detection because the fraudsters’ bots are getting smarter too, and they can often fool the basic filters inside a DSP or ad server. A dedicated AI fraud tool is built specifically to spot and block sophisticated non-human traffic and other scams, which protects your budget and makes sure your campaign data is based on real people.
Can AI ad tech replace human marketers?
No, AI isn’t going to replace human marketers. It’s great at automating tedious work and crunching numbers, but you still need a person to handle strategy, provide creative direction, make ethical calls, and figure out what the data actually means for the business. Think of it as a very powerful tool that makes marketers better at their jobs, not something that makes them obsolete.
What are the main challenges when integrating multiple AI ad tech vendors?
The biggest headaches are technical and logistical. You have to make sure data can flow correctly between the different platforms (which often means wrestling with APIs), you’re stuck logging into multiple dashboards with different reports, and sometimes the vendors’ AIs can even work against each other. Getting it right requires good planning on the tech side and a clear plan for who owns what data.
