68% of AI Citations Go to 15 Websites. The Other 32% Is Where You Compete.

Most AI visibility advice assumes the playing field is open -- build enough authority, do the right structural work, and AI models will start recommending you. The field is not open. The AI citation landscape is more concentrated than most practitioners acknowledge, and understanding that concentration is the only way to find where you actually have a shot.

The 68% finding

In our May 2026 investigation of cross-platform citation signals (knowledge file updated 2026-05-17), we documented the 5W PR Citation Source Index -- one of the largest citation studies published to date. The 5W PR agency analyzed 680 million citations across ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews, covering August 2024 through April 2026.

The top-line finding: the top 15 domains absorb 68% of all AI citations across platforms.

Fifteen websites -- Wikipedia, Reddit, major news publications, a handful of established review platforms -- account for nearly seven out of ten AI citations across all major platforms. The remaining 32% is split across the rest of the web.

This isn't a level playing field at the aggregate level, and it cannot be made into one. An SMB cannot outcompete Wikipedia for generic category citations. A landscaping company in Burlington is not going to dislodge "best lawn care" from Angi's main site, a major gardening publication, or the top HomeAdvisor category page. Those slots are locked.

The mistake most businesses make when they start AEO work is trying to compete in the 68%. Writing broad category content, building general topical authority, targeting high-volume queries that Wikipedia or Reddit already own -- none of this moves the needle. The 68% is a closed club. Understanding this stops you from spending a year failing at the wrong problem.

The 32% that's actually available

That remaining 32% is where the work is. And it's not uniformly competitive. The attainable slice of the 32% consists of category-level directories specific to your industry, comparison sites that cover your product or service, and niche expert sources that AI platforms actually use when answering queries with local or specific intent.

These sources are not Wikipedia. They are sites like Houzz for contractors, Avvo for lawyers, Healthgrades for healthcare providers, and category-level pages on G2, Yelp, and BBB. For local service businesses, they include Google Business Profile, HomeAdvisor, Angi, and the regional directories that rank for "[your service] in [your city]." Getting into these sources -- correctly listed, in the right category -- is the actual path to AI citations for businesses that are not themselves among the 15.

The 68% is a constraint. The 32% is a strategy.

Two citation systems running in parallel, not one

The Yext 2026 AI Visibility study (17.2 million citations across ChatGPT, Perplexity, and Gemini, documented in our platform citation behaviors research from session 14-15, 2026-05-12) added a structural nuance that matters here: only 11% of domains cited by ChatGPT are also cited by Perplexity.

These platforms are nearly independent in what they trust. A business that has solved its ChatGPT visibility has not solved Perplexity, and the reverse is equally true. The attainable 32% looks different on each platform.

ChatGPT draws 49% of its citations from third-party directories -- individual business listings on Yelp, BBB, G2, Capterra, and industry-specific registries. Your Yelp page, your BBB entry, your Capterra listing are the documents ChatGPT reads when answering category queries. Getting listed, filling out those profiles with specific service descriptions and geographic coverage, and appearing in the directories your industry uses is ChatGPT's attainable layer.

Gemini draws 52% of its citations from brand-owned websites. Your own site's structure, schema markup, and Google Business Profile carry more relative weight here than on ChatGPT. Gemini is reading your home turf rather than the directory ecosystem. Brand-owned authority is the attainable layer for Gemini.

Perplexity trusts niche expert directories specific to your category -- not the broad directories that work for ChatGPT, but the industry-specific databases relevant to your business type. For SaaS, that means G2 category pages, Capterra, and niche comparison tools. For legal, it means Avvo and Martindale-Hubbell. For home services, it means Houzz and HomeAdvisor in the specific category they index. The platform's live retrieval architecture means these sources are read fresh at query time, not from a training snapshot.

The 11% domain overlap is the important number. Three separate attainable layers, not one. A fix that moves ChatGPT often has limited effect on Perplexity, and vice versa.

What the attainable layer looks like when you measure it

In our May 2026 platform citation investigation (session 16, 2026-05-13), Scout analyzed every source URL Perplexity cited in a live Sourcepull audit -- 102 URLs total across category, brand, and service queries. This is the first time we mapped Perplexity's actual source documents empirically rather than inferring from scores.

Two findings put the attainable layer in concrete terms.

The G2 category page for AEO audit tools appeared in 3 out of 4 category queries. Not individual product listing pages within G2 -- the aggregated category page that lists all tools in the category. Perplexity reads the comparison layer. Being listed in the right G2 category gets a business onto that page. Being absent from the category page means you're out of Perplexity's answer, regardless of how complete your individual G2 product listing is. The category page is the citation surface; the listing is what earns you a spot on it.

A site called aeoaudittool.com -- a small independent comparison directory with no significant domain authority and no press coverage we could identify -- appeared in 3 out of 4 category queries. It appeared at the same citation frequency as a well-funded enterprise competitor. The reason: keyword-exact domain match. Perplexity's live retrieval treats a URL that directly names the category as a high-relevance signal. This site earned a position in the attainable 32% without traditional authority-building -- it got there by clearly signaling category relevance through its domain and content structure.

Both outcomes are attainable for most businesses. The G2 category page is reachable for any business that claims a G2 listing and selects the correct category. Appearing in comparison directories with clear category-intent structure is a strategic choice available to businesses of almost any size or age. Neither requires competing with the 15 domains that own the 68%.

Finding your attainable layer

The practical question, for any business, is not "how do I build AI authority?" It is: which comparison articles, directory category pages, and niche expert sources are the platforms actually reading when someone asks for businesses like mine?

For most service businesses in local markets, the answer is a short list -- five to twelve sources that dominate category queries in your geography and vertical. Appearing in those sources, correctly listed and in the right category, is the mechanism that moves category scores. For SaaS and tool businesses, the list typically centers on G2, Capterra, and whichever niche comparison directories Perplexity has indexed for your specific category. The aeoaudittool.com case suggests these niche comparison directories can outpunch their domain-authority weight when their URL and content directly signal category relevance.

The work is not glamorous. It's identifying which comparison sources exist for your category, verifying you're present in them, and ensuring your business appears on category pages rather than just in individual listings. That's the 32%. That's where citations actually come from for businesses that don't own their own slice of the 68%.

A full Sourcepull audit captures the actual source URLs Perplexity cited for your specific category queries -- the comparison sites and directories that are currently in your attainable layer. The Signal Check at sourcepull.ca gives you the score snapshot first, showing whether category queries are returning any results for your business at all. For most businesses, that's the right starting question before you spend time on the rest.