How we went from "what is AEO?" to building a dedicated tool and running full implementations - and what surprised us along the way.
Six months ago, a client asked us a question we couldn't fully answer:
"Is our website visible to ChatGPT?"
We knew SEO. We knew structured data. But AI visibility as a measurable, optimizable thing? We had to figure that out. Fast.
What followed was months of research, testing, breaking things, and eventually building our own tooling to do this properly. This is what we learned.
The problem nobody is talking about (yet)
65% of Google searches now end without a click. Users get answers directly from AI. ChatGPT, Perplexity, Google AI Overviews, Gemini - they're all pulling information from websites and presenting it as direct answers.
If your client's website isn't structured in a way these systems can parse, understand, and trust - it doesn't exist in that answer.
This is not theoretical. We tested dozens of websites across different niches. The pattern was brutal and consistent: technically solid websites, good SEO, great design - completely invisible to AI engines because they lacked the right structured signals.
What "AI visibility" actually means technically
Here's where most content on AEO gets vague. Let us be specific.
AI engines don't crawl your site the way Google's traditional spider does. They rely heavily on:
1. Structured data (JSON-LD)
Not just having schema - having the right schema, correctly nested, with complete properties. We've seen websites with Organization schema that's missing sameAs, contactPoint, or logo. To an AI engine, that's an incomplete entity. Incomplete entities don't get cited.
The schema types that matter most, in order:
- Organization - who you are
- Service / Product - what you do
- FAQPage - direct answers to common questions
- HowTo - process-oriented content
- Article / BlogPosting - for content pages
2. Entity clarity
AI systems build an understanding of who you are by connecting signals across the web. Your website says one thing. LinkedIn says something slightly different. Your Google Business profile has an outdated description. Crunchbase doesn't exist.
To an AI, you're a blurry entity. Blurry entities don't get recommended.
The fix sounds simple but requires actual work: consistent name, description, and URLs across every platform, connected via sameAs in your Organization schema.
3. Content structure
This is the one that surprises most developers we talk to. The way content is written - not just tagged - matters enormously.
AI engines look for what we call "answer-ready paragraphs": 40-60 word blocks that directly respond to an implicit question. Think of every H2 on your page as a question. The paragraph immediately below it should answer that question completely, in plain language.
Most websites don't do this. Most content writers don't think this way. It's a structural problem, not a content quality problem.
4. Bot access
We've checked robots.txt files across hundreds of sites now. A significant percentage are blocking AI crawlers - sometimes intentionally, often accidentally. GPTBot, PerplexityBot, Google-Extended, ClaudeBot - if these are blocked, no amount of schema or content optimization will help.
The platform problem
Here's something that doesn't get discussed enough: AEO implementation is platform-specific, and the differences are significant.
Next.js gives you the most control. generateMetadata, JSON-LD in layout files, server-side rendering - you can implement everything properly. It's also where we do our most thorough implementations.
WordPress is manageable but messy. Yoast and RankMath handle some schema automatically, but they rarely go deep enough. Custom functions.php additions are usually necessary, and the output can conflict with plugin-generated schema.
Framer is where it gets interesting. There's a 5,000 character limit on custom code injection. Most complete Organization + Service + FAQ schema blocks exceed this. The workaround - code overrides on individual components - works but requires knowing the platform well. We've built specific processes just for Framer implementations.
Each platform needs a different implementation approach. Copy-pasting the same schema solution doesn't work.
How we built our testing process
Manual AI visibility testing is essential and underrated. Automated tools can check for schema presence and validity - but they can't tell you whether ChatGPT actually cites your client's site when someone asks a relevant question.
Our testing protocol: prepare 8-12 queries relevant to the business, test each across ChatGPT, Perplexity, Google AI Overviews, and Gemini, document citations and positions, compare against competitors.
The results are often revealing. We've had clients who ranked #1 on Google for their primary keywords - and had zero AI visibility. We've also found the opposite: smaller sites with excellent schema implementation appearing consistently in AI answers.
The correlation between proper structured data + entity clarity + answer-ready content and AI citation rate is real and measurable.
Why we built AEO Checker
Doing this manually across dozens of client sites wasn't scalable. We needed a way to quickly establish a baseline, identify gaps, and track improvement.
So we built AEO Checker - a tool that analyzes a URL across four categories: schema markup, content structure, AI accessibility (bot access, crawlability), and citability signals. It outputs a score with specific, prioritized recommendations.
It's not a replacement for manual testing. But it's the starting point for every audit we do, and it lets us show clients a clear before/after when implementation is complete.
What we'd tell ourselves six months ago
A few honest observations after doing this in production:
AEO is not SEO with a different name. The optimization targets are different. The success metrics are different. A site can have a perfect Lighthouse score and excellent keyword rankings and still be invisible to AI. Treat it as a separate discipline.
Schema errors are common and costly. A single malformed JSON-LD block can invalidate an entire schema implementation. Validation with Google's Rich Results Test and Schema.org Validator isn't optional - it's part of every deployment.
The field is moving fast. What works today in Perplexity may weight differently in six months. Building monitoring into every client engagement - not just one-time implementations - is the right model.
Content structure is the hardest part to sell, but it matters. Developers can implement schema in an afternoon. Restructuring content for AI readability is an ongoing process that requires buy-in from content teams. The technical side is the easier conversation.
Where this is going
AEO is roughly where SEO was in 2010. Most businesses don't know they need it yet. The ones who move early will have a significant advantage - not just in AI visibility, but in overall search as AI-generated answers continue to absorb a larger share of search traffic.
For developers building client sites: this is a service worth understanding and offering. The technical implementation isn't out of reach - but it requires platform-specific knowledge and a systematic approach that most clients can't do themselves.
If you're working through AEO implementation challenges or want to discuss the technical approach for a specific platform, we're happy to talk. This is what we do now.
We're Inity Agency - a technical implementation agency specializing in AEO. Our AEO Checker tool is free to use for baseline analysis.
Top comments (0)