All Four AI Platforms Cited the Same Source Only 1.7% of the Time

If you've been treating AI visibility as a single channel -- fix your directories, update your schema, and assume the work carries across all platforms -- a September 2026 study should change that assumption.

Orbit Media Studios analyzed 13,184 citations across ChatGPT, Claude, Gemini, and Perplexity using 1,792 query-domain combinations (data through August 23, 2026). Their finding: all four AI engines cited the same domain for the same question in only 30 cases out of 1,792 -- a four-way agreement rate of **1.7%.**

That number isn't a quirk of methodology. It's consistent with what we've been documenting in our own citation behavior research. And it has direct implications for how any serious fix plan should be structured.

Why 1.7% Isn't Surprising If You Know How These Platforms Work

This finding didn't come out of nowhere. We've been tracking platform citation divergence since the first multi-source studies appeared in early 2026.

In our June 2026 investigation of cross-platform citation behavior -- drawing from three independent studies totaling over 680 million analyzed citations -- we documented that only 11% of domains cited by ChatGPT also appear in Perplexity's citation pool for the same queries. That was the two-platform figure, and it already told a stark story: fixes that work on one platform transfer to the other less than 90% of the time.

The Orbit Media four-platform figure takes the same logic further. Two engines that barely agree on sources (11% overlap) plus two more engines with their own distinct source architectures produces predictably low consensus. Each platform has a fundamentally different relationship with the web:

- **Perplexity** runs a real-time web search for every query, drawing from multiple live search APIs. Its citation behavior reflects what's currently indexed and crawlable. - **ChatGPT** answers primarily from training data (parametric memory). It retrieves from the web selectively, and its citation pool reflects what was well-represented in its training corpus -- third-party directories, earned media, sites indexed heavily before the training cutoff. - **Gemini** trusts the business's own website at high rates. In Steady Demand's AI Citation Ledger study (1,487 queries, 50 US metros, 10 service categories, July 2026), 59.9% of Gemini's local service citations came from the business website itself -- not from directories or third-party sources. - **Claude** operates differently from all three. The Orbit Media research distinguishes between citations (AI links to a source) and mentions (AI names a brand without sourcing it). Claude names brands without providing source links at a high rate relative to ChatGPT and Perplexity. If you're tracking whether AI platforms "cite" your business, Claude may be talking about you without creating a traceable citation.

These are not parallel systems pointed at the same source pool. They are structurally independent, each optimized for a different kind of retrieval.

What The Local Services Data Shows

The Steady Demand Citation Ledger, which focused specifically on local service queries across US metros, provides an additional cross-engine figure that's more granular than the Orbit Media cross-category data.

For local service queries, Gemini and ChatGPT cited the same domain only **8% of the time** -- the lowest cross-engine overlap figure in our vault for same-query cross-engine comparison.

That 8% figure is lower than the 11% ChatGPT/Perplexity overlap from the broader category research. It's lower because local service queries are more geographically and categorically specific -- reducing the chance that two platforms, pulling from different source architectures, happen to land on the same business for the same hyperlocal query.

There's a practical consequence here that Steady Demand's data surfaces directly. Gemini citations for local services were stochastic: the same business appeared in only 7% of repeated searches for the same query. Google's traditional local pack returns the same businesses roughly 90% of the time for the same query. AI-generated local search results are operating in a different regime entirely -- a snapshot of your Gemini presence today is a low-reliability indicator of your sustained visibility.

What Perplexity's Citation Volume Means

The Orbit Media study also puts a specific number on one of the most practically relevant differences between these platforms: Perplexity cites approximately **19.2 sources per answer**.

We've documented this before -- Perplexity's platform architecture requires citation in every response because every response is built from live web retrieval. But the Orbit Media figure, drawn from 13,184 actual citations, confirms the volume and contrasts it explicitly with ChatGPT's more selective citation behavior (when ChatGPT names a brand, it typically links it -- but it names brands much less frequently than Perplexity does).

For a local business trying to understand where to focus, this has a practical read: Perplexity is the platform most likely to cite your business if your content is well-structured, crawlable, and uses vocabulary that matches how customers describe their need. The platform cites actively and frequently. The question is whether your site and directory presence is showing up in its retrieval layer.

Gemini is the platform where your own website quality matters most. The Steady Demand data (59.9% business website citations for local services) makes that case more strongly than any prior dataset in the vault. If a client's Gemini scores are weak, the starting point is site architecture and semantic clarity -- not directories.

ChatGPT is the platform that rewards historical footprint. Its training data skews toward content published between 2023 and 2025. Businesses with thin pre-2025 digital presence face a different kind of gap than businesses that need to fix their current web presence.

The Fix Plan Consequence

If the platforms are almost entirely independent in their citation behavior, the implication is that a fix plan targeting all four needs to be structured as four separate problems, not one problem solved four ways.

This is what our multi-platform audit architecture is designed to surface. The platform divergence data from Averi (680 million citations), Whitehat SEO (118,000 responses), and Passionfruit -- all independently arriving at roughly the same 11% ChatGPT/Perplexity overlap -- confirmed that single-platform auditing misses the majority of a business's AI citation landscape. The Orbit Media four-way figure (1.7%) extends that logic across all four major engines.

A business that audits only on ChatGPT is working with less than 2% of the source overlap it shares with the other three platforms. The 98% that remains is not unreachable -- it's just invisible from a single-platform view.

We built Signal Check as a free first step to surface these per-platform gaps without requiring a full audit. If you've been monitoring AI visibility through a single platform or a tool that synthesizes across platforms without surfacing the per-engine breakdown, the 1.7% agreement rate is a reasonable prompt to look at the numbers by platform rather than as an aggregate.

The divergence isn't a bug in AI search. It's a structural feature of how these systems retrieve information. The fix plans are different because the problems are different.