Case Study · 2024–2025
← Back to WorkSEO & AI Landing Pages
Building a scalable system to generate high-quality SEO landing pages using AI, combining design templates, structured data, and content workflows to capture organic demand across hundreds of locations.
Role
Product Designer · workflow lead
Team
Growth & SEO marketer, content editor, web engineers
Timeline
2024–2025
Scope
Templates, AI content pipeline, structured data, QA
We were spending heavily on paid acquisition. The opportunity was clear: people searching "RV rental [city]" or "motorhome rental near [destination]" had strong intent, but our organic presence had dwindled and we missed most of these queries. Hundreds of high-intent search terms with no corresponding landing pages.
The challenge: creating hundreds of unique, high-quality landing pages isn't feasible by hand. The pages need to be genuinely useful, not keyword-stuffed templates, while maintaining visual consistency and conversion optimization.
Page template design
Designed a flexible page template with interchangeable content zones: hero with location-specific imagery, listing grid pulling real-time inventory, location guides with local tips, FAQ sections, and trust signals. Each zone could be populated dynamically.
AI content generation
Built an AI-assisted content pipeline that generates location-specific copy (travel tips, destination highlights, seasonal advice) using structured prompts and real local data. Each page gets unique, genuinely useful content, not templated fill-in-the-blank.
Quality control layer
AI generates the draft; humans review the output. I designed a review workflow where content is flagged for tone, accuracy, and brand compliance before publishing. The system catches problems before they go live while maintaining the speed advantage.
This project taught me that AI content generation is a design problem, not just a writing problem. The quality of the output depends entirely on the structure of the input: the template design, the data model, and the prompt architecture. A well-designed system produces content that feels handwritten. A lazy system produces obvious AI slop.
The other lesson: SEO pages need to earn their existence. If the page isn't genuinely useful to someone searching that query, it shouldn't exist. We killed dozens of planned pages that couldn't offer real value beyond keyword matching.
Deciding which pages deserved to exist
With hundreds of possible location pages, the first design decision was what not to build. Working with our growth marketer, I prioritized pages by combining search demand with real inventory: a page for a city only made sense if guests could actually book RVs near it. That rule cut a large share of the keyword list immediately and kept every page tied to something bookable.
Within that list, we started with the highest-intent queries, “RV rental” plus a city or destination, and treated the first batch as an experiment before scaling up.
Designed for people, structured for search engines
- Structured data. Each template outputs FAQ and breadcrumb markup, so search engines understand the page and can show richer results.
- Internal linking. Pages link to nearby cities and popular destinations, forming clusters that help both crawlers and travelers explore.
- Performance. Templates are lightweight and fast on mobile, where most travel searches happen.
- Real inventory. Listing grids pull live availability, so a page is never a dead end.
Keeping AI accurate, and pages distinct
The biggest risk with AI-generated local content is confident nonsense: a campground that doesn’t exist, or a scenic drive in the wrong province. We tackled it at the input rather than the output. Prompts were grounded in structured local data we supplied, and the review step checked facts, not just tone. Anything that couldn’t be verified was cut.
The second risk was sameness. If every page reads the same, search engines treat them as duplicates and visitors bounce. Varying the structure of the content zones, not just the city name, kept pages genuinely distinct.
Designing the workflow, not just the page
Most of the design work here was invisible on the page itself. I mapped the end-to-end workflow with our growth marketer and content editor: who picks the next batch of locations, where the local data comes from, how drafts are generated, who reviews them, and how a page moves from approved to published.
Engineers built the templates and the publishing pipeline from that map, which meant the system matched how the team actually worked instead of forcing a new process on them. The editor’s review queue was designed with them directly; if reviewing a page took too long, the whole system would stall.
We met regularly to look at what was ranking and what wasn’t. Those sessions fed straight back into the template, the prompts and the prioritization rules, so each new batch of pages was better than the last.
Measuring and maintaining
Pages were tracked in Search Console for impressions, rankings and clicks, and in analytics for bookings. Underperforming pages were revised or removed rather than left to decay, consistent with the principle that every page has to earn its place.