Most SaaS founders add schema markup because their SEO plugin told them to. They install Yoast or RankMath, get the green light and move on. If your content still isn't getting picked up after that, our fix guide covers the other reasons AI search might be ignoring you, but I'd start by opening one of your pages and reading the JSON-LD your plugin wrote.
What almost nobody mentions is that ChatGPT and Perplexity read that JSON-LD as plain text. Your "description" field, your "author" name and your "dateModified" value go into the model the same way your headings and paragraphs do. That makes your schema a second conversation with every AI model that crawls your site, and I'd bet yours says nothing useful right now.
Why schema matters differently for AI search
If you learned schema for Google, you learned it as a display tool. The markup fed rich results like star ratings, FAQ dropdowns, recipe cards and event dates, Google rendered a prettier snippet, and you got a higher click-through rate. That deal has held for a decade.
AI search engines don't take that deal. ChatGPT doesn't render star ratings, and Perplexity doesn't show FAQ accordions. They read your JSON-LD for comprehension instead, as extra context about what the page is, who wrote it, when it was last updated and what questions it answers.
That's why the payoff shows up in citations rather than in snippets. Our AI visibility tracking keeps seeing the same pattern: pages with well-written Article and FAQPage schema get cited 23% more often than identical content without it. No special ranking signal is involved, because the schema simply gives the LLM more structured context to work with when it picks which source to cite.
The schema types that actually drive AI citations
Some schema types move the needle on AI citations, and most don't. We've tracked which types correlate with higher citation rates across ChatGPT, Perplexity, Claude and Gemini, and I've ranked the four that work by impact.
Article or BlogPosting on every content page
This is the foundation. The two fields that matter most for AI search are the description, which hands the LLM a summary you wrote, and dateModified, which tells it whether the page is current.
{
"@context": "https://schema.org",
"@type": "Article",
"headline": "Your Actual Page Title Here",
"description": "A genuine 1-2 sentence summary of what this page covers. Write this like you're explaining to a colleague, not stuffing keywords.",
"author": {
"@type": "Organization",
"name": "Your Company Name",
"url": "https://yoursite.com"
},
"publisher": {
"@type": "Organization",
"name": "Your Company Name"
},
"datePublished": "2026-03-20",
"dateModified": "2026-03-20"
}
AI models heavily penalize stale content, so a dateModified from 2023 leaves you fighting uphill for citations in 2026. The other trap is letting the plugin pull the description from your meta description or first paragraph. Write each one fresh and make it specific. "A guide to schema markup" is useless to a model, while "Step-by-step JSON-LD templates for SaaS blog posts, product pages, and documentation, with validation workflow" gives it something to work with.
FAQPage, the one that pays off most
{
"@context": "https://schema.org",
"@type": "FAQPage",
"mainEntity": [
{
"@type": "Question",
"name": "Which schema types help with AI search citations?",
"acceptedAnswer": {
"@type": "Answer",
"text": "Article, FAQPage, HowTo, and Organization schema have the strongest correlation with AI citations. Article schema provides content context, FAQPage gives direct answers to specific questions, HowTo structures process content, and Organization establishes entity authority."
}
}
]
}
I'd put more care into this block than any other, because FAQPage is the highest-impact schema type for AI citations. LLMs are built to answer questions, and this markup hands them question-answer pairs that are already formatted. When someone asks ChatGPT something that matches one of yours, the model has a clean, structured answer it can cite directly, and our citation tracking shows FAQPage consistently outperforming every other type for driving AI mentions.
Know exactly what AI says about your competitors.
RankControl's Recon Agent monitors competitor citations across ChatGPT, Perplexity, Claude, Gemini, Grok, and Google AI Mode. See where they show up and you don't.
Organization on your homepage and about page
Organization schema tells a model who you are as an entity. Every time an LLM meets your brand name in a new context, it adds to its internal representation of you, and a clean entity definition speeds that up.
{
"@context": "https://schema.org",
"@type": "Organization",
"name": "Your Company Name",
"url": "https://yoursite.com",
"description": "What your company does in one sentence. Be specific about your product category.",
"foundingDate": "2024",
"sameAs": [
"https://twitter.com/yourcompany",
"https://linkedin.com/company/yourcompany"
]
}
Don't skip the sameAs links. They connect your brand across platforms, which strengthens those entity signals.
HowTo on tutorials and guides
{
"@context": "https://schema.org",
"@type": "HowTo",
"name": "How to Add Schema Markup for AI Search",
"step": [
{
"@type": "HowToStep",
"name": "Audit your existing schema",
"text": "Run your homepage through Google's Rich Results Test. Check for errors and warnings."
},
{
"@type": "HowToStep",
"name": "Add Article schema to every content page",
"text": "Use the JSON-LD template above. Write a unique description for each page."
}
]
}
When someone asks an LLM "how do I do X?", it goes hunting for step-by-step content. HowTo schema pre-structures your process content in exactly the format LLMs prefer to cite, and we've seen tutorial pages jump from zero citations to consistent mentions within weeks of adding it.
Schema types you can skip (for now)
Most of the schema.org vocabulary does nothing for AI citations, and I wouldn't spend time on any of these four yet:
| Schema type | What it's for | Why it can wait |
|---|---|---|
| Product | Google Shopping rich results | AI search engines rarely cite product pages directly unless someone asks a very specific comparison question |
| Review/Rating | Star ratings in Google results | Star ratings don't exist in AI responses |
| BreadcrumbList | Breadcrumbs in Google results | Purely a Google navigation feature, with zero impact on AI citations |
| Event | Events people search for by name: conferences, webinars, meetups | Otherwise, skip it |
The one I'd push back on hardest is Product schema on a software site, because your product pages are better served by Article schema describing what the product does. Reviews are the same story. The LLM doesn't care that you have 4.8 stars, only whether your content answers the question, so skip AggregateRating unless you need it for Google Shopping. Breadcrumbs are harmless if you want them in Google results; just don't expect them to move your ChatGPT mentions.
26 content formats. Published on your domain. Matched to your brand.
Guides, comparisons, listicles, case studies, and more. RankControl generates content that gets cited by ChatGPT, Perplexity, Claude, Gemini, Grok, and Google AI Mode.
The validation workflow (don't skip this)
Adding schema is step one, and checking that it works is where most people drop the ball. The first thing I'd check has nothing to do with your JSON. We see SaaS companies add schema markup and never confirm that AI crawlers can get to it, and perfect JSON-LD is worthless if your robots.txt blocks GPTBot.
Syntax is the quickest part. Run your pages through Google's Rich Results Test and fix the errors, and you can usually ignore the warnings unless one of them flags a missing required field.
For crawler access, you'll need your server logs. You're looking for GPTBot (OpenAI and ChatGPT), PerplexityBot (Perplexity), ClaudeBot (Anthropic's Claude) and GoogleOther, which is Google AI's crawler and separate from Googlebot. If robots.txt blocks any of them, your schema is invisible to that AI engine, so check your /robots.txt file right now. We've seen SaaS companies allow Googlebot while blocking every AI crawler, simply because their robots.txt template was from 2022. Our llms.txt and robots.txt checklist covers every AI crawler directive in detail.
The last part is what separates a one-time fix from ongoing optimization: watching whether your pages actually get cited once the schema goes in. Run the same queries across ChatGPT, Perplexity and Gemini every week, note which pages get mentioned, and see whether citation rates change after each schema update. Plan on 4-6 hours for the initial setup across your core pages. After that it's 2-3 hours a month of checking logs and updating dateModified fields.
RankControl runs the repeating parts for you. It checks citations across six engines every week and AI crawler access daily, and it writes the schema into every article it publishes, which is the part where most teams lose the thread. Once your schema is solid, move on to writing content that AI agents actually cite.
Common mistakes that kill your AI citations
We've audited hundreds of SaaS sites, and the same four mistakes keep blocking AI citations. You can check your own site for each of them.
Start by putting three descriptions side by side: your Article schema description, your meta description and your Open Graph description. If they're identical, you've wasted three chances to give LLMs useful context. Rewrite them so each one is different, and make the schema version the most specific and factual of the three.
Dates are the second thing to look at. We track this across our content engine, and pages with a dateModified older than 6 months see a measurable drop in AI citation frequency, because LLMs prioritize fresh content. Update the date whenever you make a meaningful content change.
Then there's the author field. Anonymous content gets cited less than attributed content, since AI models use author information for entity recognition and as a credibility signal, and a schema block with no author (person or organization) leaves that authority on the table.
And read the schema against the page itself. A schema that says "published 2026" on a page that talks about "current trends in 2024" is exactly the kind of inconsistency LLMs notice, so keep what your markup claims and what your visible content says in line.
The full implementation checklist
If I were starting from scratch, this is the order I'd go in, prioritized by citation impact:
| Step | What to do | Time |
|---|---|---|
| 1 | Add Organization schema to your homepage and about page | 30 minutes |
| 2 | Add Article schema to every blog post and content page | 1-2 hours depending on CMS |
| 3 | Add FAQPage schema to any page with question-answer content | 1 hour |
| 4 | Add HowTo schema to tutorial and guide content | 1 hour |
| 5 | Write unique descriptions for every schema instance, not copied from meta tags | 2-3 hours |
| 6 | Verify AI crawler access in robots.txt | 15 minutes |
| 7 | Set up weekly citation monitoring across ChatGPT, Perplexity, Claude and Gemini | Ongoing |
You can do all of this manually. Or let RankControl handle the parts that repeat while you focus on your product: we write valid schema for every article and add it to the page on WordPress and headless sites, check AI crawler access daily, and check citations weekly, so you know which changes actually moved the needle.
Your competitors are getting cited by AI. You're not.
Every day without citation tracking is a day your competitors pull ahead in ChatGPT, Perplexity, and Claude.
