A founder friend sends you a message out of nowhere: "Hey, random question. I was asking ChatGPT about tools for [exact problem your product solves] and your name didn't come up. Tried a few different ways to phrase it. You definitely should be in those results."

You sit down and try it yourself. "What's the best tool for X?" Nothing. "What should I use for Y?" A list of five competitors. "How does [your brand] compare to [competitor]?" A vague non-answer that reads like it was generated from a product page summary written three years ago.

Your product is solid. Your SEO rankings are decent. You have paying customers who love the thing. And still, when buyers ask the most common question in your category, your name does not appear.

This is the gap we built the AI Visibility Self-Audit to diagnose. In 5 minutes, it tells you your current score, which query types are the problem, and where to start. Here is how it works, what the numbers mean, and what the most common fix actually looks like.

Why LLMs Ignore Most SaaS Brands

The short version: LLMs do not rank by domain authority. They pattern-match by content type.

Google rewards pages that earn links. LLMs learn from pages that answer specific question patterns in specific structures. A brand can have 10,000 inbound links and still be completely invisible to ChatGPT if it has never published content that matches the patterns LLMs use to build recommendations.

46%
of SaaS brands score below 40 on LLM recommendation queries for their own category
73%
of brands cited by Claude have explicit category placement language on their homepage or core pages

The five specific gaps that cause this show up across almost every brand we audit. Most of them take less than 4 hours to fix once you know which one is yours.

1
No schema markup on product pages

LLMs pull structured data at training time. A product page with no JSON-LD schema is harder to categorize and cite correctly. Signal you have this gap: definitional queries ("What is [your brand]?") return vague summaries or nothing, even though your homepage describes the product clearly.

Fix: add a SoftwareApplication or Product JSON-LD block to your homepage and top product pages. Takes under 2 hours for most teams.

2
No SAIO page structure

SAIO (Search AI Optimization) page structure means content written to answer the specific question patterns LLMs use: "what is X," "X vs Y," "tools for [job]," and "how to solve [problem] using X." Most SaaS sites have none of these in a form LLMs can cite reliably. Signal: recommendation queries return competitors but not you, even when your product is a direct match.

Fix: a single 1,500-word "What is [Brand] and who is it for" page often moves recommendation scores by 15 to 25 points in 60 days.

3
Missing or thin JSON-LD structured data

JSON-LD is not just for Google. LLMs use structured data to verify factual claims about your product: pricing, category, integrations, use cases. When that data is absent, LLMs fill the gap with whatever they can infer from marketing copy, which is usually incomplete. Signal: comparison queries describe your product inaccurately or list outdated features.

Fix: the JSON-LD block in the section below. Copy it, fill in your details, add it to your homepage <head>.

4
No llms.txt file

llms.txt is a plain-text file at yoursite.com/llms.txt that tells LLMs how to index your product: category, ICP, primary use case, integrations, and key differentiators in structured prose. It is the equivalent of a robots.txt for AI crawlers, and most SaaS brands do not have one. Signal: LLMs consistently describe your product in a way that feels like a summary of a summary, missing specifics you consider obvious.

Fix: a 200-line llms.txt takes one afternoon to write and has produced measurable citation improvement for brands within 30 days of indexing.

5
No Claude citation footprint

Claude (Anthropic's model) cites sources differently from ChatGPT and Gemini. It weights content that includes specific, verifiable claims: numbers, named integrations, exact use cases, documented methodologies. Brands that optimize for general content readability often score well on ChatGPT queries but score near-zero on Claude. Signal: your audit score is 20+ points lower on Claude than on ChatGPT for the same query types.

Fix: add a "How it works" page with specific numbered steps, named integration partners, and documented workflows. Claude looks for content it can verify, not just content it can summarize.

The question the self-audit answers: which of these five gaps is actually yours? Because most brands have one primary gap and one secondary gap, and fixing the wrong one first is a common waste of effort. Ready to find out which one you have?

The 5-Question Audit (And Why Each Question Matters)

The free AI Visibility Self-Audit is a scored checklist. You answer 5 questions about your current setup and get a score plus a category breakdown. Here is what each question is actually measuring.

1
Do you have JSON-LD structured data on your homepage?
This measures schema gap (Gap 1 and 3 above). If no: your definitional score is almost certainly below 30. The absence of structured data is the single most common finding in the full LLMRadar audit.
2
Does your site have a page that explicitly names your ICP, use case, and top 3 competitors?
This measures SAIO structure gap (Gap 2). LLMs build recommendation lists from content that names categories and comparisons explicitly. If your site has no such page, you are absent from most "tools for X" responses regardless of product quality.
3
Does your site have an llms.txt file?
This measures Gap 4 directly. A yes here typically adds 8 to 15 points to your overall score. If you do not know what this is, your answer is no, and adding one is usually the fastest win available.
4
Have you been cited in at least 3 external "tools for X" lists in the last 12 months?
This measures external citation footprint. External lists are the strongest single signal for LLM recommendation citations. Brands with zero external list citations score an average of 22 points lower on recommendation queries than brands with 3 or more, even controlling for content quality.
5
Does your site have a "How it works" page with numbered steps and specific integration names?
This measures Claude citation readiness (Gap 5). Claude looks for specificity: exact steps, named partners, documented workflows. A "How it works" page in marketing language ("seamless integration with your stack") does not count. Specific named tools and numbered processes do.

Each question produces a sub-score. The sub-scores combine into a total between 0 and 100. The total tells you which band you are in and what to prioritize. Does this make sense as a diagnostic before you spend 4 hours fixing the wrong thing?

You can run the audit now at operatoriq.io/library/ai-visibility-checklist/. It is free, takes 5 minutes, no email required. Come back for the score interpretation below.

What Your Score Actually Means

The three score bands are not arbitrary cutoffs. They correspond to observable differences in how LLMs treat brands in each range when we run the full 40-query scan.

0 to 39
Invisible. Recommendation queries return your competitors but not you. Definitional queries produce vague summaries or nothing. Branded problem queries return zero citations across most models. This is fixable, but requires structural work across at least 3 of the 5 gap categories. The full LLMRadar audit is most useful here because the prioritization matters: fixing gaps in the wrong order slows progress.
40 to 69
Emerging. You appear in some query types but not others. Definitional queries work reasonably well, but recommendation queries are inconsistent and comparison queries often describe you inaccurately. This is the most common band for brands that have invested in content but have not specifically optimized for LLM citation patterns. Usually 1 to 2 targeted fixes move you to the cited band within 60 days.
70 to 100
Cited. Your brand appears in recommendation lists for your category, comparison queries describe you accurately, and branded problem queries produce citations across at least 3 of the 4 major models. The focus at this level shifts from establishing presence to maintaining accuracy as models update and expanding citation coverage to adjacent categories.

Most SaaS brands reading this land in the 20 to 50 range on first audit. That is not a catastrophe. It is a specific, addressable gap with known fixes. The score tells you the size of the gap. The sub-scores tell you which fix to start with.

The 4-Hour Fix

The most common gap in the 0-to-39 band is Gap 3: missing JSON-LD structured data. It is also the fastest to fix. Here is the exact block you need, with the fields you need to customize in brackets.

JSON-LD for SaaS homepage (paste into <head>)
<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "SoftwareApplication",
  "name": "[Your Brand Name]",
  "description": "[One sentence: what it does, for whom, primary outcome]",
  "applicationCategory": "[Your software category, e.g. 'ProjectManagementApplication']",
  "operatingSystem": "Web",
  "url": "https://[yourdomain].com",
  "offers": {
    "@type": "Offer",
    "price": "[Your starting price or 0 for free tier]",
    "priceCurrency": "USD"
  },
  "publisher": {
    "@type": "Organization",
    "name": "[Your Company Name]",
    "url": "https://[yourdomain].com"
  },
  "featureList": [
    "[Feature 1 in plain language]",
    "[Feature 2 in plain language]",
    "[Feature 3 in plain language]"
  ],
  "audience": {
    "@type": "Audience",
    "audienceType": "[Your ICP: e.g. 'B2B SaaS founders', 'ecommerce operators']"
  }
}
</script>

The four fields that matter most for LLM citation are description, applicationCategory, featureList, and audience. The description should name your category explicitly (not just your outcome). The audience should name a specific ICP, not a demographic. LLMs use these fields to answer definitional and recommendation queries about you.

After adding this block, the typical timeline to see citation improvement is 30 to 60 days for models that crawl regularly (Perplexity is fastest) and 60 to 90 days for models with less frequent index updates (ChatGPT and Claude tend to move slower). The fix is permanent unless you change your product category significantly.

Is JSON-LD your primary gap? The self-audit score will tell you. If you scored 0 to 39, it almost certainly is. If you scored 40 to 69, the gap is more likely SAIO structure or external citations, and the JSON-LD fix alone will not move you to the cited band.

Why We Built This

The LLMRadar audit started as a manual process. We were running 40 queries by hand across four models, recording results in a spreadsheet, and writing fix lists that took a full day to produce. For brands that could spend $197 and wait 2 hours, that process worked well. The fix lists were specific, the prioritization was clear, and the results were measurable.

The problem was that the first question every founder asked before buying was some version of: "Is it even worth auditing? How bad is my score?" And we could not answer that without running the full scan.

The self-audit is the triage layer. It tells you whether you have a structural problem worth diagnosing in detail, and if so, which category of fix to prioritize. For about 30% of brands, the self-audit surfaces a clear primary gap and the fix is obvious enough that they can act on it without the full audit. For the other 70%, the self-audit score confirms there is a problem worth investigating and gives them the specific sub-score breakdown that tells them what the full audit will find.

We built the free version because we kept answering the same five triage questions manually. Now the tool does it. If you want the specific numbered fix list with every item mapped to a query result, the full LLMRadar audit is the next step. But start with the free one. It takes 5 minutes and you will know your score before you finish reading this sentence if you go now.

How to Use the Audit

Here is the practical sequence that produces the most useful result.

  1. Run the self-audit first. Go to operatoriq.io/library/ai-visibility-checklist/ and answer the 5 questions honestly. "I think we might have JSON-LD somewhere" counts as no. "We added schema markup last year but I'm not sure if it's still live" counts as check first, then answer.
  2. Note your sub-scores, not just your total. A total of 45 with a sub-score of 8/20 on schema and 18/20 on external citations tells you something very different from a total of 45 with the reverse breakdown. The sub-scores point to the specific fix.
  3. Match your primary gap to the 5-gap list above. If your schema sub-score is below 10, start with the JSON-LD block above. If your SAIO structure sub-score is below 10, the "What is [Brand]" page is your first priority. If your external citation sub-score is below 10, one guest post or directory listing in the right context will move your score more than any on-site change.
  4. If you scored below 50 and the gaps feel unclear, the full LLMRadar audit gives you the specific numbered fix list. 40 queries, 4 models, PDF report in 2 hours. Every item in the fix list maps to a specific query result so you know exactly what is driving each recommendation.

The self-audit is the right starting point. If your score is 70 or above, you probably do not have a structural problem and the paid audit is unlikely to surface high-priority changes. If your score is below 70, you now know which of the 5 gaps to investigate, and you have the JSON-LD template to fix the most common one today.

Free Tool
AI Visibility Self-Audit: Score your brand in 5 minutes
5 questions. A score from 0 to 100. Sub-scores across 4 LLM query categories. No email required. Takes less time than reading this paragraph again.
Run the free audit →