If you are relying on just one search engine to track your AI visibility, you don't have the full picture.
Kevin Indig's AI Halftime Report for the first half of 2026, published this week via Search Engine Journal, contains a statistic that should change how every brand thinks about AI search measurement: 91% of AI citations appear in only one of ChatGPT, Perplexity, or Google AI Overviews.
That means fewer than 1 in 10 citations are shared across engines. If your tool tracks ChatGPT and your customer asks Perplexity, you have no idea what they are seeing.
The market is fragmenting, fast
The same report tracked market share shifts from July 2025 to July 2026.
ChatGPT's share of the AI search market fell from 78% to 56%. That is still the largest single player by a wide margin, and ChatGPT now logs over 900 million weekly active users. But the dominance is eroding.
Gemini climbed from 15% to 30%. Claude grew from 2% to 10%. Google AI Overviews now reach over 2.5 billion users monthly, appearing on close to half of all searches and 14% of shopping queries, according to IAB data.
This is not a winner-take-all market anymore. It is a multi-engine market where different platforms serve different audiences, different query types, and different stages of the buyer journey.
Why citations do not overlap
There is no single index behind AI search. Each engine builds its own retrieval pipeline, trains on different data, weighs sources differently, and applies different reasoning before generating an answer.
When ChatGPT cites a source, it does so based on its own real-time retrieval and training corpus. Perplexity uses a different retrieval system with different source preferences. Google AI Overviews draws from its web index with its own ranking signals. Claude has its own approach entirely.
The result is that the same query, asked across four engines, can produce four completely different sets of cited sources. And Indig's data shows this happens 91% of the time.
This is not a bug. It is the architecture of AI search. Every engine is a walled garden with its own view of the web.
What this means for your measurement strategy
If you are using a single-engine rank tracker, you are playing a game where you can only see one of the boards.
Imagine running traditional SEO for years and only checking Google, ignoring Bing entirely. That was always a reasonable shortcut because Google had 90%+ market share. But AI search share is splitting four or five ways, and the fragmentation is accelerating. The Google-only shortcut does not work here.
The practical implication is straightforward. If you want to know how your brand appears in AI search, you need to check every engine your customers actually use. Not just the one that is easiest to track.
The consumer behavior problem
Here is where it gets urgent. Indig's report also found that roughly 75% of consumers pick whatever appears at the top of an AI-generated shortlist.
That means the engine your customer happens to use determines which brands they see, and most people go with the first recommendation. You do not get to choose which engine your customers use. They choose, and the citation data shows their choice changes the answer.
If your brand is visible in ChatGPT but invisible in Perplexity, and your customer uses Perplexity, you are not in the conversation. You do not get a chance to make your case. The AI made a different recommendation, and the consumer went with it.
The cost of single-engine tracking
Most brands are not even tracking AI visibility at all. The IAB reported this week that only 16% of brands currently measure their AI search visibility. Of those that do, many rely on a single platform, usually ChatGPT, because it has the largest user base.
But Indig's 91% finding makes the single-platform approach mathematically indefensible. If you track only ChatGPT, you capture at most the citations that appear in ChatGPT. You miss every citation that appears exclusively in Perplexity, Gemini, Claude, or Google AI Overviews. Given the overlap rate, that is the majority of your total AI visibility.
You are measuring a slice and calling it the whole pie.
What GA4 referral data actually tells you (and what it doesn't)
There is a practical way to see which AI engines are sending you traffic today: Google Analytics 4 referral data. When someone clicks a link from ChatGPT or Perplexity and lands on your site, GA4 records the source. You can set up a custom channel group with a regex filter to separate AI referral traffic from everything else.
This is genuinely useful. If you see that ChatGPT sends you 87% of your AI referral traffic, and those visitors spend more time on your site and convert at higher rates, that tells you something real: your buyers are on ChatGPT. You should prioritize optimizing for ChatGPT's retrieval signals. That is a smart, data-driven decision.
But GA4 referral data has a critical blind spot. It only records clicks. It does not record citations.
Say a customer asks Perplexity for a recommendation. Perplexity cites your competitor. The customer clicks the competitor's link and never visits your site. GA4 shows zero Perplexity traffic. You might look at that empty data and conclude Perplexity does not matter for your business. But what actually happened is Perplexity recommended your competitor and you lost the customer. You just cannot see it because there was no click to measure.
GA4 answers the question: "Which engines do people click through from?" A multi-engine citation audit answers a different question: "Which engines are citing me at all?" You need both.
The practical strategy is to combine the two data sources. Use GA4 to find which engines send you engaged visitors. Those are the engines your actual customers use, and you should optimize hardest for them. Then use a multi-engine audit to find where you are cited but getting no clicks. Those are untapped opportunities or, worse, silent losses where a competitor is winning the citation and you don't even know it.
This is not theoretical. AI referral traffic grew 527% year-over-year according to reporting from Dana DiTomaso of Kick Point, and it consistently outperforms average engagement metrics. The visitors who do click through from AI engines are high-intent. Knowing where they come from is valuable. Knowing where they do not come from, because a competitor got the citation instead, is arguably more valuable.
What a real audit looks like
A credible AI visibility audit needs to do three things.
First, it needs to query multiple engines. At minimum: ChatGPT, Perplexity, Google AI Overviews, and increasingly Gemini and Claude. Each engine is a separate measurement surface.
Second, it needs to use consistent, documented query construction. The IAB framework, published August 3 as part of Project Eidos, explicitly calls out query construction as a disclosure requirement. If you cannot explain how queries were formed, your results are not reproducible, and reproducibility is the line between what the IAB calls directional data and decision-grade data.
Third, it needs to report per-engine results, not a blended score. A single number that averages visibility across engines hides the fragmentation. You need to see where you are strong, where you are weak, and where you are absent entirely.
The bottom line
AI search is not consolidating. It is fragmenting. ChatGPT lost 22 percentage points of market share in one year, and three other engines gained ground. Citations barely overlap across platforms. Consumers pick from whatever shortlist the AI generates, and different engines generate different shortlists.
If you are measuring one engine, you have a directional read on a fraction of your visibility. If you are measuring none, you are flying blind entirely.
The fix is not complicated. Audit across engines. Use transparent, documented queries. Report per-platform results. That is the difference between thinking you know how AI sees your brand and actually knowing.