ChatGPT mentions 3.2x more brands than it cites. Google cites 2.4x more than it mentions. They disagree 62% of the time.

BrightEdge ran identical prompts across ChatGPT, Google AI Overviews, and Google AI Mode. The three platforms disagreed on which brands to recommend 62% of the time. Only 17% of queries produced the same brand recommendations across all three.

If you are optimizing for "AI search" as if it is one thing, this data says it is at least three different things, and they contradict each other more often than they agree.

BrightEdge published the study analyzing tens of thousands of identical prompts across the three platforms using their AI Catalyst tool. The key numbers:

The most counterintuitive finding is how mentions and citations work on each platform:

ChatGPT names brands 3.2 times more often than it links to them. It says "you should look at Notion" without linking to notion.com. Google AI Overviews does the opposite: it links to 14.30 sources per query but only names 6.02 brands. Google cites more pages than it recommends.

This means two different things are happening. ChatGPT operates as a brand recommender that relies on training data and trust patterns. It knows what Notion is and says so, but does not always retrieve a source page. Google AI Overviews operates as a citation engine that retrieves pages and links them, but the pages it links are not always the brands it would recommend.

If your strategy is "get cited by AI," you need to know which game you are playing. On ChatGPT, the game is being in the model's knowledge. On Google AI Overviews, the game is being in the retrieved source pool.

The disagreement is not uniform across query types. BrightEdge broke down alignment by query intent:

Comparison queries produce the most agreement. When someone asks "X vs Y," all three platforms tend to name the same competitors. "Best" queries produce the least agreement. When someone asks "best tool for X," the three platforms recommend different brands 77% of the time.

This makes sense. Comparison queries have a finite answer set (the two things being compared). "Best" queries are open-ended, and each platform's ranking model fills the blank differently.

Healthcare has the highest disagreement. E-Commerce the lowest. If you are in health or education, assuming your visibility on one platform transfers to another is wrong 60-68% of the time.

If you run a single audit and call it "AI visibility," you are measuring one surface. The calibration study I published last week tested one engine (Perplexity) and found no relationship between the GEO score and citation. That result might be different on Google AI Overviews, where citation behavior is structurally different.

The practical implication: test your visibility on each platform separately. Ask ChatGPT "what are the best tools for [your category]" and see if you are mentioned. Ask Google the same thing and check the AI Overview sources. Ask Perplexity and look at the citations. They will give you different answers 62% of the time, and that 62% is where your gap lives.

If you want to check whether AI crawlers can even reach your site, run a free audit on geoready.dev. It checks 27 AI crawlers and tells you which ones can access your content. The same engine is available as a CLI on GitHub, open source, MIT licensed.

The audit tells you if you are reachable. It does not tell you if you are recommended. These are three different platforms, and being reachable on all three does not mean being recommended on any of them.