Okay, I need to talk about a number that broke my brain this week. And no, it's not my screen time report (though that remains a personal crisis). It's this: 91%.

That's the percentage of URLs cited by AI search engines that appear in only one engine. Not "mostly one engine." Not "a bit skewed toward one engine." Only one. Out of 3.7 million citations tracked across ChatGPT, Perplexity, and Google AI Overviews, the vast majority showed up in a single engine and nowhere else.

At the other end? Just 2.37% of URLs appeared in all three.

I've been staring at those two numbers side by side for days now, and the implication is one of those things that feels obvious once you see it but somehow nobody's saying out loud: there is no such thing as "AI search visibility." There are three completely different visibility games, and most of us have been playing them as if they're one.

Here's the assumption running through basically every AEO conversation right now: do the fundamentals — build topical authority, structure your data, earn editorial credibility — and visibility follows. Treat AI search like Google circa 2015. Get the basics right and the citations come.

New research from Omnia, covering 20,000 prompts across those three engines, suggests this assumption is wrong in the specific way that costs people money and time they don't get back. Which is, you know, the worst way to be wrong.

Because these aren't three versions of the same system. They're three entirely different systems that happen to answer the same questions. It's like discovering that your "universal" phone charger actually only works with one brand, and you've been packing it for trips assuming it would charge everything. Except the trip is your entire search strategy and the phone is your business. (This analogy got away from me, but you see the point.)

The low overlap isn't a bug or a temporary thing that'll sort itself out over brunch. It's structural, and it comes down to the fact that each engine finds information in a fundamentally different way.

Google AI Overviews draws from Google's existing search index. Research consistently shows that the vast majority of sources cited in AI Overviews already rank in the top twenty organic results for the same query. Translation: if you're invisible in traditional Google search, you're invisible in Google's AI answers too. The optimisation work here is basically an extension of the SEO you've been doing for years — topical depth, domain authority, structured data, solid content architecture. Old wine, slightly different bottle.

Perplexity takes a completely different approach. Every single query triggers a real-time web search. No reliance on training data for factual responses, every claim tied to a cited source. Perplexity rewards freshness, source density, and specificity. It actively favours content that's recently published and clearly structured. If your best content went live eighteen months ago and hasn't been touched since, Perplexity will cheerfully bypass it in favour of something newer from a less authoritative source. Recency trumps reputation here, which is either refreshing or terrifying depending on how often you update your content. (For most of us: terrifying.)

ChatGPT operates on yet another model entirely. For the majority of queries, it draws on its training data — a large corpus of text with a fixed cutoff date. Web search only kicks in when the system detects real-time information is needed, or when the user explicitly asks for it. When ChatGPT does cite sources, it favours high-authority, well-established content. When it doesn't activate search — which is roughly two-thirds of the time — your visibility depends entirely on whether your content made it into the training corpus. That's a fundamentally different optimisation problem. You're not competing for a ranking. You're competing for inclusion in the AI's memory, which sounds like the plot of a sci-fi film I'd definitely watch.

This fragmentation creates a practical problem that most teams haven't confronted yet, and it lives inside the very tools designed to help them.

If you're using an AI visibility tool that gives you a single composite score across engines, that number is hiding the most important thing the data has to tell you. A brand that scores well in a blended AEO dashboard could be cited moderately across all three engines — or it could be absolutely dominating one engine and invisible in the other two. Those are wildly different situations, but the composite score makes them look identical. It's the strategic equivalent of averaging your bank balance across all your accounts and concluding you're fine because the number looks reasonable, while one account is overdrawn and another is doing all the heavy lifting.

The scale of this is bigger than most teams expect. Separate research tracking dozens of brands across multiple AI platforms documented citation volume variance of up to 615 times for the same brand between platforms. One study tracking B2B brands across six AI surfaces over 90 days found the per-platform citation profiles were so different they looked like entirely different companies. If your marketing team is monitoring only one engine — or monitoring all three but compressing them into a single number — they're making strategic decisions with a map that doesn't match the territory.

The Omnia data includes a breakdown by content type that helps explain why the overlap is so low — and it's both useful and slightly demoralising.

Guides and tutorials have the highest cross-engine overlap at 2.3%. Blog posts follow at 1.8%. Category pages sit at 1.6%, product pages at 1.2%, and homepages at a heroic 1.1%. The pattern is clear: content that explains, compares, or teaches something travels better across engines than content designed to convert or represent the brand.

But — and this is important — even guides, the best-performing content type, only appear across all three engines 2.3% of the time. The correct takeaway is not "publish more guides and win everywhere." It's more modest: utility-focused content has a slightly better chance of crossing engine boundaries, but the chance is still small enough to make you want to lie down.

Even the internet's most-cited domains barely travel. Wikipedia appears over 16,000 times in the dataset, but only 1.3% of those appearances are universal across engines. Reddit appears over 14,000 times with a cross-engine overlap of 0.1%. Zero point one. If Wikipedia and Reddit can't crack cross-engine portability, the rest of us probably shouldn't be building strategy around the assumption that we will.

One common instinct is that this fragmentation should narrow for high-intent, commercial queries. When someone asks an AI system for the best project management tool or the best accountancy software, the pool of credible sources feels smaller. Surely the engines agree more when there are fewer reasonable answers?

The data says no. Commercial prompts show 2.4% universal overlap. Informational prompts show 2.0%. The difference is negligible. Each engine's own retrieval logic — what sources it trusts, what formats it prefers, how it weights freshness versus authority — does most of the work regardless of query type.

This matters for anyone assuming their commercial content will naturally surface across engines because the competitive field is smaller. It won't. The engines are making independent judgements about which sources to trust, and they disagree on commercial queries just as enthusiastically as they disagree on everything else.

If a single AEO strategy is too blunt to be useful, what replaces it? Not three entirely separate strategies — that way lies burnout and a very tired content team. It's a base layer of fundamentals with an engine-specific layer on top.

The base layer stays the same. Structured data, topical authority, earned media coverage, editorial credibility, and well-organised content architecture. These are the retrieval fundamentals — the table stakes that get your content into any engine's consideration set. Nothing in this new data changes that. If your content isn't findable, structured, and credible, no engine-specific tactic will compensate. (This is the "eat your vegetables" portion of the newsletter. Sorry.)

The engine-specific layer is where the new work lives. For Google AI Overviews, the priority is conventional SEO strength — ranking well in organic search is still the primary path to citation. For Perplexity, the priorities are content freshness, clear source attribution within your content, and a publishing cadence that keeps your material current. For ChatGPT, the priority is building the kind of broad brand authority and web presence that makes it into training data — consistent mentions across authoritative third-party sources, strong knowledge base presence, and the kind of brand footprint that an AI would encounter repeatedly during training.

Measurement needs to change too. Three metrics matter more than any composite score. First, presence: the percentage of your tracked prompts where your domain appears in any engine. Second, portability: the percentage of your cited URLs that appear in all three. Third, concentration: the percentage of your citations that come from one engine. That last one is the one that'll make you uncomfortable, because it tells you where your dashboard has been giving you false confidence.

Practical move: Take twenty prompts that matter to your business — the questions your customers actually ask when they're researching, evaluating, or deciding. Run them through ChatGPT, Perplexity, and Google AI Mode. Don't use the same browser session. Note where you appear, where you don't, and what type of content is being cited instead of yours. This takes less than an hour and will tell you more about your real AI visibility than any dashboard. From there, the diagnosis becomes specific: invisible in Perplexity means a freshness gap; missing from ChatGPT means a brand authority gap; absent from Google's AI surfaces means your foundational SEO needs work. Each diagnosis leads to different actions — which is exactly why the blended approach doesn't work.

AI search visibility is real, it's growing, and it matters. But it is not one thing. It's three different distribution systems that sometimes overlap and usually don't. The sooner teams stop treating AEO as a single discipline and start treating it as three distinct problems with a shared foundation, the sooner the investment starts producing results that hold up across the places where their customers are actually looking.

The strategic question isn't "how do we rank in AI search?" anymore. It's "which engines matter most for our customers, and are we actually visible in those specific engines?"

That's a narrower question. It's also a more honest one. And in my experience, the honest questions are the ones that eventually save you money — even if they're slightly less fun at the strategy offsite.

Jo Lambadjieva is an entrepreneur and AI expert in the e-commerce industry. She is the founder and CEO of Amazing Wave, an agency specializing in AI-driven solutions for e-commerce businesses. With over 13 years of experience in digital marketing, agency work, and e-commerce, Joanna has established herself as a thought leader in integrating AI technologies for business growth.

😍It was great, all I needed

🙂Good, not great

🙄It sucked