We Scored Our Own Site With Our AI Agent Scanner. It Got a C.

Here's exactly how we fixed it — in one afternoon.

I built Pacestack to help brands understand how AI agents see them. So naturally, the first thing I did after launching was scan my own site.

44 out of 100. Grade C.

The tool I built to help other brands optimize for AI agents… was itself poorly optimized for AI agents. The irony wasn't lost on me.

But it also gave me something better than a perfect score would have: a real case study. Here's every fix I made, in order, to take pacestack.io from a 44 to a 96 in a single afternoon.

The Starting Point: 44/100 (Grade C)

The initial scan broke down like this:

• Structured Data: 7/30 — This was the biggest gap. The site had basic meta tags and OpenGraph, but no JSON-LD schemas at all. No Organization schema, no Product schema, no FAQ markup. AI agents trying to understand what Pacestack offers had almost nothing structured to work with.
• AI Accessibility: 22/25 — Solid here. The site was fast, served over HTTPS, had a sitemap, and used semantic HTML. But one AI crawler (cohere-ai) was still blocked in robots.txt.
• MCP & Agent Readiness: 19/20 — Already had an MCP server endpoint and an API, so this category was nearly maxed out.

The diagnosis was clear: structured data was the bottleneck. The site was technically accessible to agents, but it wasn't speaking their language.

Round 1: The Basics (44 → 56)

The first round of fixes targeted the lowest-hanging fruit in structured data.

What I added:

• Organization schema — Told agents who Pacestack is: the name, URL, logo, description, and social profiles. This is the digital equivalent of a business card that agents can actually read.
• WebSite schema — Declared the site's search functionality and primary URL. Basic but necessary.
• Basic OpenGraph improvements — Made sure every page had complete og:title, og:description, og:image, and og:type tags.

What changed: Structured Data jumped from 7/30 to roughly 14/30. The other categories stayed about the same.

New score: 56/100. Still a C, but meaningful progress. The problem was that agents now knew who we were, but still couldn't understand what we sold.

Round 2: Product Data and FAQs (56 → 64)

This is where it got interesting. I needed to help agents understand Pacestack's actual offerings — not just that we exist, but what we sell and how it works.

What I added:

• Product schema for all three pricing tiers — Each product ($49 Agent Readiness Report, $149 Competitor Analysis, $499 Done-For-You Optimization) got complete JSON-LD with name, description, price, currency, availability, SKU, URL, and image. When an AI agent asks "what does Pacestack offer?", this is what it reads.
• SoftwareApplication schema — Described the free scanner tool as a software application with its own URL, description, and pricing (free).
• FAQPage schema — Added three questions that agents commonly need to answer: "What is a free agent readiness scan?", "What is an Agent Readiness Score?", and "What is an MCP server?" When someone asks an AI "how does Pacestack work?", these FAQ answers are what get surfaced.
• BreadcrumbList schema — Gave agents a clear navigation path through the site structure.

What changed: Structured Data climbed from ~14/30 to ~21/30. Still room to grow, but the agent-readable product catalog was now live.

New score: 64/100. Grade B. We crossed the threshold. But I wanted to see how far we could push it.

Round 3: Full Optimization (64 → 96)

The final push targeted every remaining point available.

What I added:

• Complete JSON-LD coverage — Refined all existing schemas for maximum completeness. Every Product schema got brand info, aggregate ratings where applicable, and offer details with price validity dates.
• robots.txt fix — Explicitly allowed cohere-ai (Cohere's crawler), which had been blocked by a legacy rule. This was a 3-point swing from a single line change.
• OpenAPI specification — Published a machine-readable API spec at /api/openapi.json describing the scanner API endpoints, parameters, and response formats. This lets agents understand not just what Pacestack does, but how to interact with it programmatically.
• Services endpoint — Created a public JSON endpoint at /api/services that returns structured product data. Think of it as an API catalog that agents can query directly.

Final score: 96/100. Grade A.

The breakdown:

• Structured Data: 28/30 (up from 7)
• AI Accessibility: 25/25 (perfect)
• MCP & Agent Readiness: 19/20 (near perfect — the remaining point requires a full MCP marketplace listing)

What Actually Moved the Needle

Looking back at the 52-point improvement, here's what mattered most:

JSON-LD structured data was 70% of the gain. Going from zero schemas to complete Organization + Product + FAQ + BreadcrumbList markup was the single biggest factor. If you do nothing else, do this. It took about 30 minutes to implement all the schemas in our Next.js layout file.

Product schema completeness matters more than presence. Having a Product schema with just a name and description is worth maybe 2 points. Having one with name, description, price, currency, availability, SKU, image, brand, and URL is worth 5. The difference between "we sell something" and "here's exactly what we sell, what it costs, and where to buy it" is massive to an agent.

One line in robots.txt cost us 3 points. We had a legacy Disallow rule blocking cohere-ai from a previous configuration. Removing it was the highest-ROI fix of the entire process: 3 points for changing one line.

OpenAPI specs are underrated. Most sites don't publish machine-readable API documentation. If you have any kind of tool or API, even a simple one, publishing an OpenAPI spec signals to agents that your product is designed to be interacted with programmatically. That's exactly what the MCP & Agent Readiness category is looking for.

The 15-Minute Version

If you want to do the fastest possible version of this for your own site, here's the priority order:

1. Add Organization schema (5 min) — Name, URL, logo, description, social profiles. Drop it in your site's <head> as a JSON-LD script tag.
2. Add Product schema for your main offerings (5 min per product) — Name, description, price, currency, availability, image, URL. If you sell things, this is non-negotiable.
3. Check your robots.txt (2 min) — Search for GPTBot, ClaudeBot, anthropic-ai, cohere-ai, PerplexityBot. If any are blocked, ask yourself if that's intentional. Every blocked crawler is a missed discovery opportunity.
4. Add FAQPage schema (5 min) — Pick 3-5 questions your customers actually ask. Mark them up as FAQ schema. Agents will surface these answers when people ask about your product category.

That's 20 minutes of work that could take you from a C to a B or beyond.

Pacestack is the first AI agent optimization platform. Follow @pacestackio for weekly insights on how AI agents see your brand.