AI Engines Agree on the Answer. They Don't Agree on Who to Credit.
ChatGPT and Perplexity share only 11% of their cited domains. Even Google's two AI surfaces overlap 13.7%. One visibility score hides the real picture.
AI Engines Agree on the Answer. They Don’t Agree on Who to Credit.
Here’s a finding that should reset how every marketing team reports AI visibility. Google runs two AI surfaces — AI Overviews and AI Mode. They reach the same conclusion about 86% of the time. They cite the same URL only 13.7% of the time (Averi).
Same company. Same index. Same question. Two completely different lists of who deserves credit for the answer.
Now stretch that across the whole field. ChatGPT and Perplexity overlap on just 11% of the domains they cite. If your team tracks AI visibility on one platform and treats the number as “our AI visibility,” you’re reporting on roughly a tenth of the landscape and calling it the whole map.
The answer converges. The citations fragment.
This is the part most people get backwards. We assume that as AI search matures, the platforms will start agreeing — that a “good” source for ChatGPT becomes a good source for everyone, the way a #1 Google ranking used to travel across the web.
The opposite is happening. The answers are converging. Ask four engines for the best project management tool and you’ll get a similar shortlist. But the sources each engine reads to build that shortlist are diverging, and the gap is widening, not closing.
Averi’s benchmark, drawn from 680 million citations across ChatGPT, Google AI Overviews, and Perplexity, put hard numbers on it:
- ChatGPT and Perplexity cite the same domain only 11% of the time
- Google’s own AI Overviews and AI Mode cite the same URL only 13.7% of the time, despite reaching the same conclusion 86% of the time
- Passionfruit’s March 2026 cross-engine check landed in the same range, around 12% overlap across three engines (Passionfruit)
Each engine is building its answer from a different pool. The conclusions rhyme. The bibliographies don’t.
Why the same page wins on one engine and vanishes on another
The divergence isn’t random, and it isn’t about your content being better or worse. It’s structural — baked into how each engine turns a prompt into a set of sources. Nick Lafferty’s breakdown of the asymmetry traces it to a few mechanics:
- Prompt rewriting. ChatGPT fans a single user prompt into roughly 91% unique sub-queries before it retrieves anything. Perplexity stays about 88% literal to what you typed. Different queries pull different sources, before content quality ever enters the picture.
- Source weighting. Each engine leans hard on a different corner of the web. In the Averi data, Wikipedia accounts for 47.9% of ChatGPT’s top citations. Reddit is 46.7% of Perplexity’s. YouTube is the largest single bucket for Google AI Overviews at around 23.3%.
- Co-citation habits. Some engines pair sources together by reflex (ChatGPT cites Edmunds and Kelley Blue Book together about a third of the time in auto queries); others cite in isolation.
- Language and geography. On Gemini, more than a third of cross-language prompt pairs produce zero overlap in the hostnames cited. The same question in two languages can mean two different webs.
So a page optimized into Reddit threads and community discussion can own Perplexity and barely register on ChatGPT, which is reaching for an encyclopedic reference instead. Neither result is a mistake. They’re two different retrieval systems doing exactly what they were built to do.
The numbers that break a single visibility score
Source preferences split by platform, which means a strategy tuned for one engine is actively mistuned for another:
| Platform | Top source by share | Retrieval style |
|---|---|---|
| ChatGPT | Wikipedia (~47.9% of top citations) | Rewrites prompts into many sub-queries; leans encyclopedic |
| Perplexity | Reddit (~46.7%) | Stays literal to the prompt; leans community and forums |
| Google AI Overviews | YouTube (~23.3%) | Pulls video, reference, and forum content into a mixed panel |
| Google AI Mode | Diverges from Overviews | Same conclusions as Overviews, 13.7% URL overlap |
Citation rates swing just as wildly. Superlines tracked how often brands got cited at all across ten platforms between mid-January and mid-February 2026 (AI Search Statistics 2026):
| Platform | Citation rate |
|---|---|
| Grok | 27.01% |
| Perplexity | 13.05% |
| Google AI Mode | 9.09% |
| Gemini | 6.38% |
| Google AI Overview | 2.11% |
| Copilot | 1.27% |
| ChatGPT | 0.59% |
| Claude / Mistral / DeepSeek | ~0% |
Superlines put the headline spread at up to 615x between the most and least citation-heavy platforms. Whether you take that exact multiple or just read down the column, the point holds: the same brand, in the same window, surfaces constantly on one engine and is nearly invisible on the next. A single blended “AI visibility” figure averages those extremes into a number that describes none of them.
What this means for how you measure
If you’ve been chasing one AI visibility score, the score isn’t wrong so much as it’s meaningless. Averaging an engine where you’re at 13% against one where you’re at 0.5% produces a tidy number that hides both the win and the gap. Here’s how to stop fooling yourself.
1. Report per engine, never blended. Track ChatGPT, Google (AI Overviews and AI Mode separately), Perplexity, and Gemini as their own lines. The moment you collapse them into one figure, you lose the only information that’s actionable.
2. Find your worst engine and ask why. The 11% overlap means a strength on one platform tells you almost nothing about another. If you own Perplexity through Reddit presence but barely register on ChatGPT, that’s not noise. It’s a Wikipedia and reference-corpus problem you can actually work on.
3. Match the optimization to the engine’s retrieval. Community proof for the forum-heavy engines. Clean, entity-rich reference content for the encyclopedic ones. Video where the panel pulls video. One content plan can’t serve all four because the four aren’t reading the same web.
4. Watch the gaps move, not just the levels. These pools are diverging over time. A quarterly snapshot misses the drift. The useful signal is whether your per-engine spread is widening or closing.
5. Treat Google as two engines. AI Overviews and AI Mode share an index and still disagree on sources 86% of the time. Lumping them together inside your “Google” number buries a real divergence.
This is also the honest case for why single-platform tracking — or a manual once-a-week check in whichever chatbot you happen to use — can’t tell you where you stand. You’re measuring one ecosystem and inferring nine. The whole reason RivalHound tracks every major engine in parallel is that there’s no shortcut: the platforms genuinely don’t agree, so you have to look at each one.
The takeaway
For two decades, SEO trained us to believe in a single source of truth. Rank well, and you ranked well everywhere that mattered, because everyone was effectively querying the same index. That assumption is dead. We’ve covered how a handful of domains dominate the citation supply and how citations and brand mentions are separate signals. Cross-platform divergence is the third leg: even the sources that do get cited barely overlap from one engine to the next.
The brands that win the next phase of AI search will be the ones that stop asking “what’s my AI visibility?” and start asking “what’s my visibility on each engine, and which one is leaking?” The first question has a comforting single answer. The second one is the only version that’s true.
RivalHound tracks your brand’s visibility across ChatGPT, Google AI, Perplexity, and more. Start monitoring to see where you stand.