← All articles
Strategy· 8 min read

95.7% of AI Citations Come From Sources You Don't Control. Here's How to Find Them.

By Salman Shaikh, Cited

Here's a pattern I keep seeing in GEO audits: a brand rewrites their product pages, adds schema markup, creates an llms.txt file, unblocks every AI crawler — and their AI Visibility Score barely moves. The work addressed the wrong surface area. Source Intelligence — Cited's diagnostic layer that maps which third-party sources drive AI citations for your brand and category — exists because of this exact problem.

Profound analyzed 27 million citations across ChatGPT, Gemini, and Google AI Overviews — research later cited in Microsoft Advertising's "From Discovery to Influence" GEO series. The finding: only 4.3% of AI citations on category-level prompts come from a brand's own website. The other 95.7% come from third-party sources — review sites, publications, YouTube, Reddit, expert roundups, competitor pages.

Knowing the stat and acting on it are two different problems. Most brands hear "95.7% comes from third parties" and think okay, I need more PR. That's not wrong, but it's vague. Which third parties? Which publications actually get cited by AI? Where are your competitors showing up that you aren't?

Why do 95.7% of AI citations come from third-party sources?

When someone types "best niacinamide serum in India" into ChatGPT, the model doesn't search your website. It searches its grounding index — the web pages it retrieved and cached through its grounding (RAG) pipeline. And what it finds there are review roundups, beauty publication articles, Reddit threads, Nykaa listings, and YouTube transcripts.

Your website might have the most detailed niacinamide product page on the internet. Doesn't matter. AI citations on category prompts are an ecosystem game, not an on-site game.

The challenge for brands is that this ecosystem is opaque. You don't know which publications AI actually cites. You don't know which review sites carry weight on which platform. You don't know where your competitors are present and you're absent.

Source Intelligence makes it visible.

What does Source Intelligence do?

Source Intelligence is the second layer inside Cited's platform — the insights engine that sits between raw tracking data and actionable tasks. It captures every citation URL across all five AI platforms — ChatGPT, Gemini, Perplexity, Claude, and Grok — and classifies them into a structured map of your category's citation ecosystem.

Three capabilities matter most.

Priority Targets — the 5 source types that drive AI citations

Source Intelligence classifies every third-party source into 5 types, each with distinct outreach strategies:

Source TypeExamplesStrategic Approach
Review SitesSkinCarisma, IncDecoder, Trustpilot, Google ReviewsEncourage authentic reviews; respond to existing ones
PublishersVogue India, Elle, Cosmopolitan, expert roundupsEarn editorial coverage; contribute expert quotes
CompetitorsCompetitor websites and branded contentMonitor positioning; understand which competitor content AI cites
CommunityReddit, Quora, dermatologist forumsParticipate authentically; don't spam
MarketplaceNykaa, Amazon India, Flipkart, PurplleOptimize product listings with entity-rich descriptions

Each source type surfaces as a card with the number of citations it drives, which AI platforms cite it, and a strategic approach recommendation. This isn't a generic "build more backlinks" suggestion — it's a specific, categorized map of where AI attention lives in your category.

Citation Strengths — sources that already mention you

Not everything is a gap. Some third-party sources already cite your brand. Source Intelligence identifies these with an affinity score — a measure of how consistently a source mentions your brand across prompts and platforms.

This matters because your highest-ROI move isn't always earning new coverage. Sometimes it's deepening existing relationships. A publication that already mentions you in 3 out of 10 relevant articles is easier to convert to 7 out of 10 than a publication that has never heard of you.

Placement Targets — URLs where you're absent but competitors aren't

This is the competitive intelligence angle. Source Intelligence identifies specific URLs — not just domains, but actual pages — where your competitors are cited by AI but you're absent. These are ranked by actionability based on four signals:

  • Platform breadth — how many AI platforms cite this URL
  • Competitor density — how many of your competitors appear here
  • Content accessibility — can you realistically get mentioned on this page
  • Volume — how often this URL appears across prompts

A review roundup on a beauty publication that's cited by 3 out of 5 AI platforms, mentions 4 of your competitors, and doesn't mention you — that's a high-priority placement target. Source Intelligence surfaces it, ranks it, and feeds it directly into Task OS so your team has a specific action item, not a vague directive.

What skincare brands reveal about Source Intelligence in action

Want to know how your brand scores on these same metrics?

We'll run 20 prompts across 3 AI platforms and send your report within 24 hours.

Get a Free AI Visibility Report →

Indian skincare is one of the most competitive AI visibility categories we track. The May 2026 Cited Index covers 17 skincare brands across 4 AI platforms, generating 450 total mentions. The spread is dramatic.

BrandSoV%MentionsAI Visibility Score
Minimalist20.4%9291
The Derma Co15.3%6961
Re'equil10.7%4848
Dot & Key10.2%4646
Deconstruct8.7%3939
Plum6.4%2929
Dr. Sheth's5.8%2626
Foxtale5.6%2525
The Ordinary3.1%1414
Fixderma2.9%1313

Minimalist leads with 20.4% Share of Voice — 92 mentions out of 450. That's nearly double the second-place brand. But the more interesting question isn't who's winning — it's why.

The pattern: coverage breadth separates leaders from the rest

When I look at Source Intelligence data across skincare brands, a clear pattern emerges. The brands at the top of the table share one trait: they appear across all 5 source types. Review sites cite them. Publications write about them. They show up in Reddit discussions. Their marketplace listings are rich and well-optimized. Even competitor comparison pages mention them.

Brands in the bottom half of the table? They're typically absent from 3 or more source types. Their website might be technically solid — good GEO Score, clean crawl radar, schema markup in place. But their third-party footprint is thin. AI crawlers find plenty of content about the category, but not enough content about the brand in the sources AI trusts most.

This is the 3-Layer AI Visibility Stack in action. Layer 1 — Discoverability — is a solved problem for most of these brands. Their sites are crawlable. Layer 2 — Citability — varies, but many have decent content. Where they diverge is Layer 3 — Authority — and authority in AI search is almost entirely a function of third-party coverage.

Platform bias makes source mapping even more critical

Here's a detail that makes Source Intelligence more valuable than generic PR outreach: different AI platforms cite different sources.

Dot & Key scores 88 on Claude but only 8 on ChatGPT — an 80-point gap. Dr. Sheth's scores 60 on ChatGPT and 0 on Claude. Re'equil scores 84 on Gemini but 20 on Perplexity.

These aren't random fluctuations. Each AI platform has a distinct grounding (RAG) pipeline. ChatGPT Search prioritizes different publications than Perplexity's index. Claude's retrieval stack surfaces different community threads than Gemini's. A brand that's well-covered in sources ChatGPT trusts but absent from Perplexity's preferred sources will have a massive platform disparity — and no amount of on-site optimization will close it.

Source Intelligence maps this per-platform. It shows you not just which sources are missing, but which sources matter on which platform — so you can prioritize outreach that moves visibility on the platforms where you're weakest.

How this maps to the 3-Layer Stack

If you've read the 3-Layer AI Visibility Stack framework, you know that most brands fail at the layer they don't measure. Source Intelligence is primarily a Layer 3 — Authority diagnostic tool.

Stack LayerWhat It MeasuresSource Intelligence Role
Layer 1 — DiscoverabilityCan AI crawlers find your content?Minimal — use GEO Score and Crawl Radar
Layer 2 — CitabilityWhen AI finds you, does it cite you?Content Format Analysis shows which formats get cited
Layer 3 — AuthorityWhen AI cites sources, is your brand present?Core function — Priority Targets, Citation Strengths, Placement Targets

Most GEO tools stop at Layer 1. They'll tell you if your robots.txt blocks GPTBot. They'll check your schema markup. They'll scan your sitemap. That's important work — but it's the 4.3%.

Source Intelligence operates in the 95.7%. It's the diagnostic layer that tells you which third-party sources you need to appear in to move from "AI can find me" to "AI recommends me."

How to improve your third-party AI citations

Knowing your source gaps is step one. Acting on them is where the score actually moves. Here's how to prioritize based on what Source Intelligence surfaces.

Start with Citation Strengths. Identify publications and review sites that already mention your brand. These are your warmest leads. Reach out with updated data, new product launches, or expert commentary. Converting a partial mention into a featured position is faster than earning coverage from scratch.

Attack Placement Targets by actionability score. Don't go after every gap at once. Source Intelligence ranks targets by how realistic it is to get mentioned — a review roundup that's been updated in the last 90 days and already includes 4 competitors is more actionable than a 2-year-old listicle that hasn't been touched. Start with high-actionability, high-platform-breadth targets.

Cover all 5 source types. If you're present on review sites but absent from community threads, that's a structural gap. AI platforms pull from diverse source types — Reddit presence matters on Perplexity, YouTube presence matters on ChatGPT, marketplace listings matter on Gemini. Aim for coverage across all five types, not just the ones that feel familiar.

Match outreach to platform weakness. If your platform bias data shows you're invisible on Perplexity but strong on ChatGPT, focus outreach on sources that Perplexity's RAG pipeline prefers. Source Intelligence shows you which sources drive citations on which platforms — use that to close specific platform gaps rather than spraying effort everywhere.

Feed everything into Task OS. Every placement target and citation gap converts directly into a task inside Cited — specific, actionable, and ranked by estimated impact. This isn't a report you read and forget. It's a system you execute against.

Rewriting your website is Layer 1 work. It's necessary. But if 95.7% of AI citations on category-level prompts come from sources you don't control — and they do — then the real competitive advantage is knowing which sources matter, where your competitors appear, and where you're missing.

That's what Source Intelligence was built to reveal.

If you want to see how your brand appears across AI platforms — and get prioritized recommendations on what to fix first — request a free AI visibility audit. It covers 20 prompts across ChatGPT, Gemini, and Perplexity, with mention rate, positioning analysis, and 3 prioritized recommendations. Delivered within 24 hours, no commitment required.

S

Salman Shaikh

Former SEO nerd. Recovering big-tech PM. Currently losing sleep over whether your brand exists in an AI answer — and building tools to find out. Cited is the company. The AI Shelf is the newsletter. The obsession is real.

Free Report

See how your brand appears in AI answers — free

We'll run 20 prompts across 3 AI platforms and send your AI Visibility Report within 24 hours.

Get Your Free Report →