ChatGPT cited Facebook zero times. Perplexity cited it 44. That explains the visibility gap.

We reported earlier that Perplexity mentioned one regional brand 17 percent of the time while ChatGPT and Claude each mentioned it 33 percent, using the same 46 prompts in the same week. We pulled the citation URLs each engine returned to find out why. The answer was in the first row of the Perplexity column.

What we extracted

Same full audit: 46 buyer-intent prompts, four engines using live web search, 184 responses total. When an engine answers with live search enabled, it returns the URLs it pulled as citations. We counted how many times each domain appeared across all responses for a given engine.

Gemini routes all its search citations through an internal Google service URL, so domain-level analysis is not available for it. The breakdown below covers ChatGPT, Claude, and Perplexity.

The source mix, by engine

Across 46 prompts, ChatGPT returned 244 total citations, Claude returned 248, and Perplexity returned 370. Here are the domains that most clearly separate them.

Citation counts from the full 46-prompt panel against each engine. Gemini omitted: routes citations through an internal URL that does not expose the underlying domain.

Why this produces a lower mention rate

Facebook community posts about local internet service in a small market mostly discuss national carriers. People complain about Spectrum reliability, ask about Frontier pricing, or compare Xfinity bundles. A regional wireless ISP with limited social footprint rarely surfaces in those discussions. When Perplexity builds an answer from a source pool that includes 44 Facebook citations, the national carriers appear prominently and the regional brand does not.

There are also sources Perplexity does not use that the other engines do. ChatGPT pulled from a local ISP directory (metroconet.com) 22 times. Perplexity pulled from it zero times. That directory has the regional brand listed; Facebook does not. Claude relied heavily on broadbandnow.com (28 citations), which Perplexity ignored entirely.

The net result is predictable. Three of the four engines pulled from sources where the brand had a listing. Perplexity's source pool favored pages where the brand either was not listed or was absent from the conversation.

What the brand's own site did and did not do

The brand had a website and each engine did cite it. Claude cited it most often (22 times), ChatGPT cited it 15 times, and Perplexity cited it 13 times. Having a website helped with all three engines. But a brand's own site being cited does not mean the brand gets mentioned in the answer. An engine can pull a page as context and still write an answer that recommends someone else. The source being included is not the same as the brand being surfaced.

The implication for fixing visibility

If you are trying to improve your brand's visibility on Perplexity specifically, the fix is different from what helps with ChatGPT or Claude. Getting listed on ISP comparison aggregators will help with ChatGPT and Claude because those engines use those aggregators. It will help less with Perplexity, which already pulls from some of those same aggregators but adds a heavy layer of social and video content that those aggregators cannot offset.

For Perplexity, the relevant surface is community content: the places where your category gets discussed in social and community forums. That is a harder, slower fix than getting a listing added to an aggregator database. It also means the work to fix one engine does not automatically fix another.

This is why an audit that reports a single blended visibility score misses the point. The number that matters is the per-engine number, and the fix that matters is the per-engine source gap.