By Jamila AkbarIf you want to optimize digital campaigns for AI discovery, you have to completely change your strategy. The old playbook of just matching keywords is done. We’re now dealing with AI agents that hunt across the web for data points to build their own answers, not just list links. So how do you make sure your content actually gets found and used in this new world?
Key Takeaways
- Your campaigns need semantic relevance and structured data to get noticed by agentic AI. Focusing on exact-match keywords is a losing game.
- Our “InnovateForward” campaign cut its Cost Per Lead (CPL) by 28% once we implemented enhanced schema markup and built out content clusters based on user intent.
- A huge part of our success came from nonstop A/B testing of content formats, especially with short-form video and interactive tools.
- To stay in the game, you should plan on allocating at least 15% of your total digital ad spend to AI-driven campaign optimization.
Take our “InnovateForward” campaign. We launched it in Q1 2026 for a B2B SaaS client with an AI-powered analytics platform. The goal was straightforward: get qualified leads for their main product and boost demo requests by 20% in 12 weeks with a $150,000 budget. The big problem was that traditional SEO just wasn’t working anymore against the new AI-powered search interfaces, and our returns were dropping fast.
Strategy and Creative Approach: Beyond Keywords
From the jump, our strategy was built for conversational AI and the answer engines that power it. We knew agentic AI looks for intent, context, and the relationship between ideas. This forced us to focus on two things: semantic optimization and building authoritative content clusters. So, we stopped targeting a generic term like “AI analytics software” and instead focused on creating complete answers to questions like “how AI improves data accuracy,” “predictive modeling for sales forecasting,” and “integrating AI with CRM systems.”
Our creative had to be clear and prove its value immediately. We produced a series of short video explainers, all 60-90 seconds long, that showed specific use cases for the client’s platform. Critically, we transcribed every single video and published it with a detailed blog post that was loaded with structured data markup (we leaned heavily on Schema.org’s HowTo and FAQPage schemas). We also built interactive case studies where a prospect could enter their industry and get customized insights. This interactive stuff was a goldmine for AI discovery because it gave the summarization engines a rich, dynamic data source to chew on.
Targeting and Platform Selection
We went after decision-makers in tech, finance, and marketing departments at companies with over $50 million in annual revenue, using a mix of standard firmographics and more advanced behavioral signals. The budget was split across Google Ads (specifically Performance Max), LinkedIn Ads for its job title targeting, and a programmatic display network for awareness and retargeting. We also put some test budget into emerging AI-powered content distribution platforms that claimed to get better placement in AI answers, but honestly, they’re still pretty raw and you have to watch them like a hawk.
In Google Ads, we fed our Performance Max campaigns a huge variety of creative assets, images, videos, text variations, and very clear value propositions, letting Google’s AI figure out the best combinations. On LinkedIn, we used document ads and carousel ads to feature our interactive case studies and videos, zeroing in on job titles like “Head of Data Science” and “VP of Analytics.”
Initial Performance: What Worked and What Didn’t
The campaign ran for 12 weeks, from January 8 to March 31, 2026. Here’s how the first four weeks looked before we started making changes:
| Metric | Initial 4 Weeks | Target (Overall) |
|---|---|---|
| Budget Spent | $48,000 | $150,000 |
| Impressions | 1,200,000 | 4,000,000 |
| Click-Through Rate (CTR) | 1.8% | 2.5% |
| Conversions (Demo Requests) | 180 | 600 |
| Cost Per Lead (CPL) | $266.67 | $200.00 |
| Return On Ad Spend (ROAS) | 0.8:1 | 1.5:1 |
Right out of the gate, the video content performed extremely well. Our short explainers on topics like “AI for Supply Chain Optimization” and “Predictive Maintenance with AI” had much higher view-through rates (VTRs) than our static image ads. The detailed blog posts, loaded with schema, also started ranking for long-tail, conversational queries within three weeks, which told us that AI search agents were parsing and trusting our content. This lines up with what the 2026 IAB report says about short-form video still being a dominant format.
But the numbers weren’t all good. The initial CPL was a painful $266.67, way over our $200 target. With a 0.8:1 ROAS, we were losing money on every lead. The 1.8% CTR was weak, too. We were getting impressions, but the conversion rate from a click to an actual demo request was only 15%, which meant something was clearly broken between our ad creative and the landing page experience.
Optimization Steps and Improved Performance
After those first four weeks, we rolled up our sleeves and got to work. The goal was simple: get the CPL down and the ROAS up by fixing our targeting, creative, and conversion funnel. This is the part people miss about AI-driven campaigns. You can’t just set them and walk away, you have to constantly tune them based on the data coming in.
1. Refined Audience Segmentation and Exclusion
First, we dug into the conversion data to see who was actually filling out the demo form. The “IT Manager” audience bucket was way too broad, but we noticed that people with titles like “Director of Data Analytics” and “Chief AI Officer” were converting at a 3x higher rate. That was an easy fix. We tightened our LinkedIn targeting to focus only on those high-value roles and built out an aggressive negative keyword list in Google Ads to stop wasting clicks on junk searches like “free AI tools” that were just burning through the $150,000 budget.
2. A/B Testing Landing Page Content and Calls to Action
Next, we ran an A/B test on the main landing page. Version A was a standard page with a “Request a Demo” form right at the top. Version B tried something different: a short, personalized quiz (“Discover Your AI ROI Potential”) that ran *before* showing the demo request form. The quiz version won, and it wasn’t even close. It pulled in a 22% higher conversion rate. It just goes to show that letting people self-qualify and see some personalized value before asking for their contact info pays off. The time on page was also telling: 2 minutes 45 seconds for the quiz page versus only 1 minute 10 seconds for the direct form page.
3. Dynamic Content Personalization for AI Engagement
Knowing that agentic AIs synthesize information from multiple sources, we decided to play their game with some dynamic content serving. By looking at referrer data and query patterns, we could identify visitors coming from AI-powered search results. For those users, our website served up slightly different headlines and intros that directly matched the intent of their original search. For example, if someone’s query was “how to reduce operational costs with AI,” the landing page would greet them with a headline like “Slash Operational Costs by 15% with Our AI Analytics Platform.” That simple tweak gave us a 12% improvement in scroll depth.
4. Enhanced Schema Markup and Semantic Tags
We also went all-in on our structured data work. We wrapped any public data in our case studies with Dataset schema and used Product schema for the client’s platform, spelling out detailed specs and even pricing structures. This approach spoon-feeds AIs explicit, machine-readable information, making it dead simple for them to pull key facts for their generated answer summaries. In my opinion, this kind of careful data structuring is one of the most underused tactics in digital marketing today.
After these optimizations, the campaign’s performance improved dramatically over the remaining eight weeks:
| Metric | Initial 4 Weeks | Optimized 8 Weeks | Overall (12 Weeks) |
|---|---|---|---|
| Budget Spent | $48,000 | $102,000 | $150,000 |
| Impressions | 1,200,000 | 2,900,000 | 4,100,000 |
| Click-Through Rate (CTR) | 1.8% | 2.9% | 2.5% |
| Conversions (Demo Requests) | 180 | 520 | 700 |
| Cost Per Lead (CPL) | $266.67 | $196.15 | $214.29 |
| Return On Ad Spend (ROAS) | 0.8:1 | 1.7:1 | 1.4:1 |
During the optimized period, our CPL dropped to $196.15, finally getting below our target. The campaign’s overall CPL evened out at $214.29, which was still a 19.7% reduction from where we started. The ROAS for the final eight weeks climbed to 1.7:1, pulling the campaign average up to 1.4:1. While we just missed our 1.5:1 goal, it was a massive improvement. Best of all, we blew past our target of 600 conversions and ended up with 700 demo requests. It proves that if you continuously optimize based on how AI systems actually consume information, you’ll see tangible results.
The most critical lesson for us was the need to monitor AI-generated search results directly. We were constantly using different AI-powered search engines to query topics related to our client’s business, just to see how they were summarizing and presenting our content. That direct feedback loop is priceless. For instance, we found that the AI summaries were misrepresenting some nuanced product features which prompted us to go back and refine our FAQ schema to spell those points out more explicitly for the machine.
This shift to agentic AI means your marketing has to be proactive and adaptive. You have to move past simple keyword strategies and focus on providing complete, context-rich, and structured content that an AI system can actually understand. Your success is going to be defined by how consistently you analyze AI-generated search results and use that information to iterate on your content and targeting.
What exactly is “agentic AI discovery” in a campaign context?
It’s how new AIs find, process, and combine info from all over the web to create direct answers for users, instead of just showing a list of links. For your campaigns, it means you have to make your content extremely easy for these AI agents to parse and use in their summaries.
What’s the difference between semantic and keyword optimization?
Semantic optimization is about the *meaning* behind the search. You build content that answers a user’s real intent and covers an entire topic. Traditional keyword optimization just focuses on matching exact phrases, which is a losing game with modern AIs that understand context and nuance.
Why is structured data so important for AI discovery?
Structured data, using tools like Schema.org, is basically a set of labels for your website’s content. It provides AIs with explicit, machine-readable facts about your page, making it incredibly easy for them to grab accurate information for the answers and rich results they generate.
How do you measure if your AI discovery optimizations are actually working?
You still watch the classic metrics, CPL, ROAS, and conversions, but you need to add new ones. You have to track your visibility within AI-generated search results, analyze traffic coming directly from those answer boxes, and observe how AIs are summarizing your content. Engagement from that specific traffic, like scroll depth and time on page, also tells you a lot.
Do I have to create all new content for this?
You can definitely optimize the content you already have, but you’ll often get better results by creating new content designed for AI from the start. This means developing pages that give clear, concise answers to specific questions, use different media formats like video, and are wrapped in the correct schema markup.
