AEO Is Four Separate Games, Not One

We argued in There Is No Universal AI Citation Formula that citation mechanics vary by vertical, so a single optimization checklist fails. A new dataset adds a second axis of fragmentation, and it is sharper than the first.

The signals do not just differ by industry. The platforms cite almost entirely different sources from each other.

The Dataset

BuzzStream, through Vince Nero, ran 595 prompts across four AI surfaces and collected roughly 30,000 citations. The result spanned 16,764 unique URLs and 5,587 domains. The four surfaces were Google AI Mode, Gemini, Google AI Overviews, and ChatGPT.

The question was simple: when you ask the same thing across platforms, do they point you to the same sources?

They do not.

76% on One Platform, 0.8% on All Four

Here is the structural finding. Of every cited source in the study, 76.1% appeared on only one platform. They were never echoed by any of the other three. Only 0.8% of citations, 184 in raw count, were shared across all four surfaces.

Read that again. Less than one percent of what AI cites is common ground. The overwhelming majority is platform-specific.

This kills a comfortable assumption. Most AEO advice treats “get cited by AI” as a single target, as if there were a master list of authoritative sources that every model draws from. There is no master list. Each platform has built its own retrieval reality, and those realities barely intersect.

Even sibling products diverge. Google AI Overviews and Google AI Mode, both Google, both fed by the same index in theory, showed a Jaccard overlap of just 37.4%. Two products from the same company, asked the same questions, agreed on roughly a third of their sources. If Google cannot keep its own two surfaces aligned, no external optimization will align them for you.

What the Shared Core Actually Looks Like

The 0.8% that does appear everywhere is worth studying, because it tells you what universal authority looks like when it exists at all.

Wikipedia made up 35% of that four-platform shared set. But Wikipedia was only 3.8% of total citations. So the encyclopedia is disproportionately the thing all four models trust, yet it is a small slice of the overall citation volume. The shared core is narrow and dominated by reference infrastructure you cannot become.

The more useful number: 96% of the identical cross-platform citations came from blog and content pages, not homepages, not product pages, not category listings. When a source did break through to all four platforms, it was almost always a specific piece of written content answering a specific question. Depth of content, not domain authority alone, is what travels across platforms.

That is the one durable lesson hiding inside the fragmentation. The universal layer is thin, and it is made of substantive content pages, not brand presence.

Why One Formula Cannot Work

If 76% of citations are single-platform and the two Google surfaces only agree a third of the time, then “AI visibility” is not a metric. It is four metrics wearing one name.

A page that ChatGPT loves may be invisible to Gemini. A source AI Overviews surfaces may never appear in AI Mode. Optimizing for an aggregate “AEO score” averages four uncorrelated signals into a number that describes none of them. You would be tuning a blended average that no single user ever experiences, because no user queries all four platforms at once.

This is the same trap as the vertical finding, one level up. Before, the warning was that SaaS advice poisons finance results. Now the warning is that ChatGPT advice tells you nothing about Gemini. The optimization surface is not one game with local rules. It is four games with different boards, and a tiny shared square at the center that you mostly cannot claim.

From Optimization Target to Budget Allocation

The practical shift is from optimization to allocation. You no longer ask “how do I rank in AI.” You ask “which platform matters for my audience, and what share of effort does each one earn.”

That reframes AEO as a portfolio decision. Treat each surface as a separate channel with its own retrieval behavior, its own winners, and its own measurement. Spread effort according to where your buyers actually search, not evenly, and not toward a phantom universal authority that the data says does not exist.

Governance follows the same logic. If leadership asks “are we visible in AI,” the honest answer is a four-cell table, not a single status light. Anyone reporting one AEO number is hiding three quarters of the picture by construction, because three quarters of citations live on exactly one platform.

Do This Now

1. Stop reporting a single AEO score. Build a four-column view: AI Mode, Gemini, AI Overviews, ChatGPT. Track citations per platform separately. One blended number is misinformation.
2. Pick your platforms by audience, not by ego. Find where your buyers actually run their queries. Fund those surfaces. Ignore the rest until they matter.
3. Invest in content pages, not homepages. 96% of cross-platform citations were blog and content pages. The path to the thin universal layer runs through substantive written answers, not brand pages.
4. Re-measure quarterly per platform. Each surface evolves on its own schedule. A win on ChatGPT this quarter says nothing about Gemini next quarter. Track them independently.
5. Set per-platform budgets. Allocate AEO effort the way you allocate ad spend across channels: deliberately, unevenly, and with separate scorecards.

The takeaway is not that AEO is harder. It is that AEO was never one thing. It is four channels, and the firms that budget for them as four will out-position the ones still chasing a single formula.

---

This analysis synthesizes AI Citation Overlap: Do AI Platforms Cite the Same Sites? (BuzzStream, June 2026).

Victorino Group helps brands allocate AEO effort per platform instead of chasing one formula. Let’s talk.