We Ran 1,000+ GEO Audits. Here's What Actually Gets Brands Cited by AI.

We built a free GEO audit tool and ran it against 1,000+ domains. The results were striking: only 12% of brands are formally cited by ChatGPT for their target keywords. The other 88% are either vaguely mentioned without attribution or completely invisible to the fastest-growing search channel in history.

This article breaks down what we found — the citation rates, the scoring patterns, and the three specific fixes that separate cited brands from invisible ones. Every data point comes from real audits run through LumenGEO's scoring engine across industries ranging from SaaS to e-commerce to professional services.

Published: March 2026

The dataset: 1,000+ GEO audits across industries

Our dataset spans 1,000+ domains audited through LumenGEO, covering SaaS, e-commerce, professional services, healthcare, fintech, and more — with each audit measuring citation presence, prominence, and quality across ChatGPT and Perplexity.

Every audit follows the same methodology: we query AI search engines with the brand's target keywords, analyze whether the brand appears in the response, measure how prominently it's cited, and score the underlying content for extractability. The result is a GEO Score from 0 to 100 that captures overall AI search visibility.

The brands in this dataset range from seed-stage startups to publicly traded companies. Domain authority spans from DA 8 to DA 92. Annual revenue ranges from pre-revenue to nine figures. This diversity is what makes the patterns so striking — the factors that predict citation are not the ones most marketers assume.

The headline numbers

Across 1,000+ audits, only 12% of brands earn a formal citation from ChatGPT for their target keywords — while 65% are completely invisible.

Here is how the audited domains break down:

• ~12% formally cited — The brand appears by name with a direct link or explicit attribution in the AI response. These brands are actively winning traffic and trust from AI search.
• ~23% mentioned but not cited — The brand name appears somewhere in the response, but without a formal citation link. The AI knows the brand exists but doesn't trust it enough to recommend as a source.
• ~65% completely invisible — No mention, no citation, nothing. For these brands, AI search engines behave as if they don't exist.

Key Finding: Nearly two-thirds of brands have zero presence in AI search results for their own target keywords. This is not a visibility problem — it is an invisibility crisis.

The gap between "mentioned" and "cited" matters more than most people realize. A mention without attribution means the AI absorbed your content during training or retrieval but chose to synthesize it without giving you credit. You informed the answer but received none of the downstream traffic or trust. A formal citation, by contrast, puts your brand in front of the user as a recommended source.

The scoring gap is enormous

The GEO Score difference between cited and non-cited brands tells the real story:

• Average GEO Score for formally cited brands: 62/100
• Average GEO Score for non-cited brands: 14/100

That is a 4.4x scoring gap. Brands earning citations are not marginally better — they are fundamentally different in how their content is structured, validated, and presented to AI models.

Brands in the "mentioned but not cited" middle tier average a GEO Score of around 31/100. They have some of the right signals but not enough to clear the citation threshold. These are often the highest-opportunity brands in our dataset — they are close to being cited and need targeted fixes, not a complete overhaul.

Perplexity cites more generously than ChatGPT

One platform-level finding stood out: Perplexity consistently cites more sources per response than ChatGPT, resulting in a higher overall citation density. For brands that appear in both platforms' retrieval sets, Perplexity is more likely to include a formal citation link.

This aligns with what we've documented in our Perplexity SEO guide — Perplexity's architecture is built around source attribution, with inline citations as a core design element rather than an afterthought. For brands starting their GEO journey, Perplexity may offer the faster path to initial citations.

What cited brands have in common: 3 patterns

The top-cited brands in our dataset share three structural patterns: answer-first content, schema markup, and third-party validation — and all three are more predictive of citation than domain authority alone.

We analyzed the content and technical profiles of the top 12% (formally cited brands) against the bottom 65% (invisible brands) to isolate the strongest differentiators. Three patterns emerged with statistical clarity.

Pattern 1: Answer-first content structure

The single strongest predictor of citation is whether a page leads with a direct, definitive answer to the query it targets.

Cited brands structure their content like reference material. The first paragraph of each section contains a clear, self-contained statement that directly answers a question. No preamble, no throat-clearing, no "In today's fast-paced digital landscape..." — just the answer.

What this looks like in practice:

• Cited brand: "The SR&ED tax credit reimburses 35% of qualifying R&D expenditures for Canadian-controlled private corporations with less than $500K in taxable income."
• Non-cited brand: "The SR&ED program is one of Canada's most important innovation incentives, designed to encourage businesses of all sizes to invest in research and development activities that push the boundaries of what's possible."

The first version gives the AI a specific, extractable claim it can cite with confidence. The second version says nothing an AI model can attribute. It's marketing copy, not reference material.

This pattern was consistent across industries. SaaS companies that led product pages with specific capability statements ("processes 10,000 transactions per second with 99.99% uptime") were cited more than competitors with vague positioning ("industry-leading performance at scale").

For a deeper look at how to restructure content for AI extractability, see How to Get Your Brand Cited by ChatGPT.

Pattern 2: Structured data markup

Brands with FAQ schema and HowTo schema on their key pages showed citation rates 40%+ higher than comparable brands without structured data.

This was one of the most actionable findings in the dataset. Structured data — specifically FAQ schema (FAQPage) and HowTo schema — correlated strongly with citation rates even after controlling for domain authority and content quality.

Why? AI search engines use structured data as a shortcut during retrieval. FAQ schema pre-formats content into question-answer pairs that are trivially easy for an AI model to extract and cite. HowTo schema organizes procedural content into numbered steps that map cleanly to the step-by-step format AI responses often use.

The brands in our dataset that implemented both FAQ and HowTo schema on their top pages had an average GEO Score 23 points higher than brands with no structured data. That's the difference between being invisible (Score: 14) and being competitive for citations (Score: 37).

This finding is especially relevant for Google AI Overviews, where structured data plays an even more direct role in determining which sources appear in AI-generated snippets.

Pattern 3: Third-party validation

The third pattern separating cited from invisible brands is the breadth and quality of third-party mentions. Brands that are referenced on trusted external domains — industry publications, review platforms, Wikipedia, authoritative blogs — get cited at significantly higher rates.

This maps directly to what we've documented in our analysis of AI citation signals: brand mentions across the web (r=0.664) are a stronger citation predictor than raw backlink count (r=0.218). AI models aren't just checking your domain authority — they're assessing whether independent sources corroborate your claims.

The implication is clear: GEO is not just an on-site optimization game. Brands that invest in PR, guest publications, review site presence, and industry directory listings create the cross-platform consistency that AI models use to build citation confidence.

Key Finding: The three strongest citation predictors — answer-first content, structured data, and third-party validation — are all within a brand's direct control. None require massive budgets or years of accumulated authority.

The biggest surprise: domain authority matters less than you think

In our dataset, sites with domain authority of 30 regularly outperform DA 70 sites in AI citations when their content is structured for extractability.

This was the finding that challenged our own assumptions. Conventional SEO wisdom says that domain authority is the primary predictor of visibility — high-DA sites outrank low-DA sites for competitive keywords. In AI search, this relationship is significantly weaker.

We found multiple cases where niche sites with DA 25-35 earned consistent citations for competitive keywords while well-known brands with DA 60-80 were invisible. The differentiator was always content structure:

• The high-DA brand had a marketing-oriented homepage with vague positioning, no FAQ schema, and product pages designed to drive conversions rather than answer questions.
• The low-DA site had a comprehensive resource section with direct answers, structured data, comparison tables, and specific data points that AI models could extract and cite with confidence.

This does not mean domain authority is irrelevant. It still influences retrieval — a higher-DA site is more likely to appear in the AI's initial candidate set. But once content is retrieved, the citation decision depends on extractability and specificity, not authority.

Key Finding: Domain authority gets your content into the AI's retrieval pipeline. Content structure determines whether it gets cited. A DA 30 site with answer-first content beats a DA 70 site with marketing copy.

For brands with lower domain authority, this is genuinely good news. You don't need to spend years building backlinks before GEO becomes viable. Restructuring your existing content for extractability can produce citation results within weeks. The LLM optimization guide covers the technical details of how retrieval-augmented generation (RAG) pipelines evaluate content for citation.

The 3 most impactful fixes

Based on our audit data, these three changes produce the largest GEO Score improvements — most brands can implement all three within a week.

We tracked which specific changes correlated with the largest score improvements when brands re-audited after making optimizations. Three fixes consistently moved the needle more than any others.

Fix 1: Add an answer-first paragraph to your key pages

Go to your top 10 pages by traffic. For each one, add a bolded first paragraph that directly and specifically answers the question the page targets. This paragraph should be self-contained — if the AI extracted just this paragraph, it should make sense as a complete answer.

The format is simple:

1. Identify the primary question the page answers (what would someone search to find this page?).
2. Write a 2-3 sentence paragraph that answers that question with specific claims, numbers, or definitions.
3. Bold the entire paragraph and place it immediately after the page's H1 heading.
4. Do not hedge. Do not qualify. State the answer directly.

This single change moved brands from "invisible" to "mentioned" in our dataset more than any other individual fix. It works because it gives AI models exactly what they need: a clean, extractable claim they can attribute to your domain.

Fix 2: Implement FAQ schema on your top 10 pages

Add FAQPage structured data to your 10 highest-traffic pages. Each page should include 3-5 question-answer pairs that address the most common questions related to that page's topic.

The questions should be the actual queries your audience uses — check Google Search Console, People Also Ask boxes, and ChatGPT's own follow-up suggestions for your keywords. The answers should be direct (2-4 sentences each) and include specific data points where possible.

FAQ schema is high-leverage because it requires minimal content creation (you're marking up content that should already exist on the page) and produces disproportionate citation gains. The 40%+ citation rate improvement we observed makes this arguably the highest-ROI technical GEO fix available.

For implementation details specific to Google's AI features, see our Google AI Overviews optimization guide.

Fix 3: Get mentioned on 3-5 authoritative sources in your niche

Identify the top 5 authoritative domains in your vertical and get your brand mentioned on each one. This could mean guest articles, product reviews, directory listings, expert roundups, or earned press coverage.

The goal is not backlinks (though those help). The goal is brand mentions on domains that AI models already trust. When multiple independent sources reference your brand in the context of a specific topic, AI models gain confidence in citing you for queries related to that topic.

Start with the low-hanging fruit:

• Claim and complete your profiles on relevant review sites (G2, Capterra, Trustpilot, industry-specific directories).
• Contribute expert commentary to industry publications that cover your space.
• Update or create your Wikipedia page if your brand meets notability guidelines.
• Publish original research that other sites will reference and link to.

This fix takes longer than the first two (weeks rather than days), but it builds the kind of durable citation authority that compounds over time. Our AI citation signals analysis provides the full data on how brand mentions correlate with citation rates.

What these findings mean for your GEO strategy

The audit data points to a clear conclusion: AI citation is not a mystery — it's a structural problem with structural solutions.

The brands earning citations are not doing anything exotic. They are writing clearly, structuring their content for extraction, using schema markup, and building consistent brand presence across trusted domains. These are all achievable within a standard content marketing workflow.

The brands that are invisible are typically making one or more of these mistakes:

• Leading with marketing copy instead of answers. AI models cannot cite "We're the industry-leading solution." They can cite "Our platform processes 50,000 invoices per month with a 0.3% error rate."
• Blocking AI crawlers. A surprising number of domains in our dataset had robots.txt rules that blocked GPTBot or PerplexityBot, making citation impossible regardless of content quality.
• Treating content as a conversion tool rather than a reference resource. Pages optimized purely for human conversion (short copy, big CTAs, minimal detail) give AI models nothing to extract. The highest-cited pages serve a dual purpose: they answer questions comprehensively AND convert visitors.
• Ignoring structured data. FAQ schema and HowTo schema take hours to implement and produce measurable citation improvements within weeks.

The path from invisible to cited is not a complete content overhaul. For most brands, it is three targeted fixes applied to their 10 most important pages. The data from our 1,000+ audits makes that case clearly.

Frequently asked questions

What counts as a "formal citation" vs. a "mention" in your data?

A formal citation means the AI response includes the brand name along with a direct link to the brand's website, or explicitly attributes a specific claim to the brand as its source. A mention means the brand name appears in the response text but without attribution or a link — the AI is aware of the brand but doesn't position it as a recommended source. The distinction matters because citations drive referral traffic and trust, while unattributed mentions provide neither.

How often should I re-audit my GEO Score?

We recommend re-auditing every 4-6 weeks, or immediately after implementing significant content changes. AI models update their retrieval indexes regularly, and competitive citation landscapes shift as other brands optimize their content. A quarterly audit is the minimum to track trends; monthly is ideal for brands actively working on their GEO strategy.

Do these findings apply to all industries equally?

The three core patterns — answer-first content, structured data, and third-party validation — held across every industry in our dataset. However, citation rates vary by vertical. B2B SaaS and fintech brands see higher baseline citation rates (likely due to more structured content practices), while local services and e-commerce brands tend to score lower. The fixes are the same regardless of industry; the starting points differ.

Is a GEO Score of 62 considered "good"?

In the current landscape, a GEO Score of 62 places a brand in the top 12% of all domains we've audited. Most brands score between 10 and 40. A score above 50 indicates consistent citation presence across platforms, and a score above 70 suggests strong competitive positioning in AI search. For benchmarks by industry and score tier, see What is a GEO Score?.