There is no single 'AI visibility.' Four engines describe your brand — and they disagree

Ask your team for the brand’s AI visibility and you will probably get a number. One number, tracked over time, going up or down. It feels like a rank. It is not. It is an average — and averaging is exactly the wrong operation for what is actually happening.

Your brand is not being described by an AI. It is being described by at least four: ChatGPT, Google AI Overviews, Google AI Mode, and Gemini. They read different sources, they name different brands, and they tell different stories about you. A single score blends those four stories into one figure that matches none of them. You optimize against the average, and you miss the engine that is quietly dropping you.

This is not a hunch. In the largest study of its kind to date, Semrush analyzed 126 million real US AI search prompts across 22 industries and the four engines, from January to April 2026. The headline finding is not that any one brand won. It is that the four engines barely agree with each other.

The number that hides everything

Start with how aligned each engine’s mentions are with its citations — the brands it talks about versus the sources it quotes. On Google AI Overviews, the overlap is 64%. On Gemini, it is 30%. That is a 34-point spread between two of the four engines answering for your brand right now. A blended score sits somewhere in the middle and describes neither.

The divergence shows up everywhere you look in the data. Only three brands — YouTube, Amazon, and Facebook — appear in every engine’s top ten; after that, the rankings split sharply. The two closest engines, ChatGPT and Gemini, still share only 78 of their top-100 most-mentioned brands. Citation overlap runs below 50% across every pair of engines. And the same brand scores wildly differently depending on who you ask: Notion’s AI Visibility ranges from 58 to 72 across the four platforms.

One score cannot carry that. When you compress a 58-to-72 reality into a single “65,” you have thrown away the only information that would tell you where to act.

This matters more each quarter because the audience is no longer early adopters. Pew Research Center’s Americans and AI 2026 survey found 49% of US adults now use AI chatbots, up from 33% in 2024, with 44% using ChatGPT specifically and 60% reading the AI summaries at the top of search results. The conversation about your brand is happening across these surfaces, mostly without you in the room.

Four engines, four brands

The reason the engines disagree is that they eat differently. ChatGPT is encyclopedic and community-led: it reaches for Wikipedia and Reddit first and cites an average of 15.4 sources per answer. Gemini is Google-anchored and commerce-leaning, with the smallest source pool of the four — just 3.3 sources per answer. That is more than a fourfold difference in how many doors are open for you to be quoted through, for the exact same question.

Google AI Overviews sits in its own world again. Because it answers Google Search queries rather than conversations, its average prompt is 24.6 characters against roughly 58 for the other three — short, keyword-led, definitional. It cites YouTube above any other source by a wide margin. Google AI Mode leans social and local, surfacing Yelp into its top tier where no other engine does.

So a brand built entirely on community discussion over-indexes on ChatGPT and may be near-invisible on Gemini. A brand built on video over-indexes on AI Overviews. There is no content type that wins everywhere — which is precisely why a single number is the wrong instrument. You need four, because you are competing in four different rooms with four different door policies.

Why prompting cannot fix it

The instinct, once you see the gap, is to try to talk your way into the engines — better prompts, an FAQ written for the model, a page that addresses the AI directly. It does not work, and the reason is structural: you cannot prompt your way to AI visibility. You do not control how these four engines are built or what they retrieve. You control the sources they quote — and that is upstream of all of them at once.

This is the lever the Semrush data keeps pointing at. The brands described consistently across engines are not the ones with the cleverest owned content; they are the ones whose third-party footprint tells one coherent story. Patagonia is the cleanest example: it holds an overall AI Visibility score of 79, steady at 79 or 80 every single month of the study, scoring 87 on AI Overviews, 78 on Gemini, and 76 on both ChatGPT and AI Mode. It earns that not by publishing more, but because six specialist outdoor-gear review sites describe it consistently — together accounting for more than 65,000 brand mentions, more than the entire tier-one press combined.

Teams that treat AI visibility as a content campaign miss this entirely. Teams that treat it as an operations problem win: Semrush’s own survey of 481 marketers found that 81% of those with fully integrated SEO and AI-search workflows report more traffic or leads from AI platforms, against just 36% of teams running them separately.

Mentioned, cited, and still wrong

There is a worse failure than a low score, and the average hides it best of all: being described incorrectly. The engines do not only decide whether to mention you. They decide what you are. And when the signal about you is thin, they fill the gap with whatever the broader web implies — mispositioning you, stripping your context, or conflating you with a competitor or a namesake.

This is why being mentioned and being cited are not the same win, and why readable is not the same as cited. A mention names you inside the answer; a citation quotes your site as a source. They move independently. Reference platforms live at the extreme of that gap — AI cites Wikipedia 4.3 times, and IMDb 3.9 times, for every single time it names the brand. They have become infrastructure for the answers AI gives about everyone else. Your own brand sits somewhere on that spectrum, and a blended visibility score tells you nothing about where.

The cost of getting this wrong compounds with the traffic. AI-referred traffic to US retail sites rose 1,324% between October 2024 and May 2026, and in travel it climbed 2,215% over the same window, per Semrush. A brand that AI describes wrong is not failing to grow a small channel — it is mis-positioned in front of the fastest-growing one.

What to measure instead

Replace the one number with four, and add a second axis. For each engine — ChatGPT, AI Overviews, AI Mode, Gemini — track two things: how often you are mentioned, and how often you are cited. That is eight data points where most brands have one, and it is the difference between “our score dipped” and “Gemini stopped quoting the three review sites that carry us.”

Then calibrate ambition to your category rather than to the field. Brand concentration — the share of top mentions held by the top three brands — runs from 82.9% in News & Media down to 41.4% in Finance. In a concentrated category, breaking the top three is a multi-year project and positions four to ten are the realistic target; in a flat one, focused work can move you several ranks in a quarter. Benchmarking a finance brand against YouTube’s numbers, as a single cross-industry score invites you to, is comparing two different games.

None of this is a prompt you can write this afternoon. It is the slower, more durable work Citable runs for clients: measure each engine honestly, find the third-party sources that form your category’s citation core, and make sure that what they say about you is accurate, current, and consistent — so all four engines resolve you as the same company, doing the same thing, worth quoting. That is the narrative layer, and it is the only part of AI visibility you actually own.

Start by auditing how the four engines describe you today. One number cannot tell you. Four can.