Schema for AI search: Turning entity gaps into organic growth

My last article presented a framework for using schema markup to reverse-engineer your entity gaps. That framework prioritized building a knowledge graph using custom JSON-LD and schema.org entities alongside vector embeddings and agents.

Building the graph is a theoretical approach to entity resolution. The goal was to train Gemini, Claude, and ChatGPT to better understand two separate university brands. 

• Partner A is a private, multi-campus international university with graduate business programs. The brand is unique in its approach to international learning and its model, which expands to four countries. Their AI footprint and organic presence were underdeveloped, which resulted in lower AI visibility.
• Partner B is an online nursing program for a regional college. The program long dominated organic traffic and drove roughly 80% of the lead volume in its online portfolio. The program saw a drop in organic clicks and an increase in competitors with similar offerings. 

Each program started with a 1% share of voice in AI answers against competitors. That made the entity work with a lower barrier to entry. 

Start with the target entity model and reverse engineer the gaps

For both partners, we built an audit that analyzes ArcherEDU’s custom JSON-LD and compares it against live schema and content. We then used the results to close the entity gaps we found.

Through this analysis, we identified weak spots that undermined our partners’ brand positioning and program offerings with vague language and framing. Our goal was to determine which entities are legible, ambiguous, and unverifiable. 

This audit compares vector embeddings of our existing site with our vectorized ideal data set. We then score the corpus, our existing embeddings, and schema markup.

When comparing our partner entities against the custom JSON-LD of 85 entities, Partner A’s coverage included more than 50 entities. A third of those entities were flagged as ambiguous or unverifiable. More than 20 entities were missing from program pages and supporting content.

For Partner B, we identified 58 entities, with more than half ambiguous. We found 27 unverifiable entities across the nursing program. Both partners observed parity across entity types, with the curriculum, faculty, admissions, and financial aid clusters having the largest data gaps. 

We also mapped these entities back to where they live. Are they on a program page, or do they live on a blog/ancillary page?

We used this to prioritize gaps and inform our content strategy. In higher education, our program page is the money asset.

For Partner A, we addressed a schema-level issue involving incorrect formatting or missing mentions of credits, GPA, modality, and other program identifiers, which would include it in conversations against competitors. The incorrect schema created vague connections to requirements that their prospective students cared about.

For Partner B, we addressed a content coverage gap around career outcomes, admissions information, and curriculum. 

Close what you can, stay honest about what you can’t

We shared these results with our partners. Instead of rushing to draft content, we used these gaps to drive discovery and collaboration.

In an ideal world, our entity coverage would be perfect, LLMs and search engines would understand us, and prospective students would get every answer that satisfies their intent. In reality, some of these gaps may reflect regulatory issues, a lack of competitive pricing, or other institutional risks. 

We tackled what we could with our partners.

For Partner A, the schema and language around their program offerings aligned with international standards, but their primary audience is based in the U.S. For example, we converted duration to ISO 8601, and the page language reinforced that timeline. Their program pages list “good academic record” as a requirement, but we updated that to a numeric GPA. We fixed over a dozen entities by updating schema and supporting language with specificity. 

For Partner B, we updated our program page around the degree naming convention, admissions information, and program outcomes. We then refreshed existing blog pages and interlinked to our program page to reinforce those entities. We also reinforced the program page and blog articles with a robust schema offering. 

For both partners, we left out different entities that posed different challenges.

For Partner A, we didn’t address the application deadline because of their rolling starts across multiple campuses, which added complexity and could create friction for prospective students. For Partner B, we didn’t focus heavily on faculty due to availability issues.

Each partner had administrative issues they needed to work out. We focused on what we could rather than provide inaccurate information. 

Measure the results and the organic-growth payoff

Across both programs, they struggle with AI visibility among competitors, with Partner A sitting at the bottom of a 10-institution set on AI share of voice with 1% compared to 22% for the leader. Partner B ranks last among six national competitors, with 2% compared to 29% for the top competitor.  

While both programs trail competitors, initial post-entity growth shows clear traction.

Partner A went from 24,000 citations in January 2026 to 42,000 in July, a 75% increase over six months. That visibility is significant, especially given higher ed seasonality.

Partner A also saw significant improvement in lead quality. Enrollment growth is flat year over year, but lead volume dropped this year. Their admissions team is closing more opportunities because students are better informed of program information and requirements. Their lead-to-payment rate improved by 20%, and their application-to-payment rate rose 26%. 

Partner B reversed the decline in lead volume from 2024-2025. In 2026, they’ve seen an 18% increase in organic lead volume from 2025 to 2026.

Partner B also saw a 22% increase in application volume. From an AI visibility standpoint, AI citations moved similarly, up about 11% year over year and roughly 77% above the prior six months. Share of voice also gained a percentage point against top competitors. 

What to expect

Treat entity work like infrastructure rather than a GEO hack. Our results show significant improvement in conversion quality with a limited lift in AI search visibility. 

Markup doesn’t guarantee growth. But when paired with a content strategy, entity reconciliation can improve conversion quality and help users get the answers they need.

Our results also show us that this is a moving target. Entity reconciliation should be a routine part of your workflow. Content and facts can drift over time. Make sure your information stays fresh and relevant to your core audience. 

Turning entity gaps into organic growth

Our audit yielded two different tactical outcomes. One fixed ambiguous schema entities, and the other refreshed pages and supporting blog content around entities. Both improved organic conversion, while the partner with the supporting content strategy saw a slight rise in market share. 

The three buckets for entity reconciliation (legible, ambiguous, and unverifiable) are universal across programs, products, and services. Reverse-engineer it against an ideal entity model to determine which is which. 

Partner A Data

Partner B Data