In 30 days of tracking payments-related prompts across ChatGPT, Perplexity, Gemini, and Claude, we found 191 different brands mentioned at least once. Most of them appeared infrequently. A small cluster appeared often. And Stripe appeared in nearly every conversation where payments came up at all.
Lemon Squeezy — a well-regarded product that handles payments, software licensing, and tax compliance for digital creators — appeared in 1.2% of responses. That's a 25-fold gap between two companies operating in the same category.
What's driving it isn't product quality. It's something harder to see, and harder to fix, but entirely measurable.
What these four metrics actually tell you
Before getting into the why, it's worth understanding what each number represents — because they're measuring different dimensions of the same problem.
Appearance rate is the most direct measure: out of all the prompts that touched on payments, in what percentage did this brand show up in the AI's response? Stripe appeared in 30% of them. Lemon Squeezy in 1.2%. This is the primary signal of how embedded a brand is in AI's working model of its category.
Average position measures where in the response a brand is mentioned. Position 1.6 (Stripe) means it typically appears in the first two sentences — before any alternatives are discussed. Position 5.5 (Lemon Squeezy) means it gets mentioned later, usually in a "you might also consider" clause, if at all. Earlier mentions carry more weight because users read top-down and because AI models surface their most confident references first.
Average frequency per response measures how often a brand is mentioned within responses where it does appear. Stripe averages 5.8 mentions per response — it's referenced repeatedly throughout the answer, often in examples, comparisons, and recommendations. Lemon Squeezy averages 2.5. A brand that gets mentioned once and then dropped isn't being endorsed; it's being acknowledged.
Citation rate measures how often AI responses include a source link alongside the mention. Stripe's 6.25% citation rate reflects how often it gets referenced in the context of a specific document, tutorial, or resource the AI can point to. Lemon Squeezy's 1% citation rate suggests it's mentioned from general training knowledge, not from a specific cited source — a weaker signal in terms of authority.
"Stripe isn't winning because of better prompting. It's winning because 15 years of content — tutorials, documentation, news coverage, developer references — built a model of the brand that AI engines trust."
Why Stripe dominates AI answers
Stripe has been the default payments infrastructure for startups and developers since 2010. That longevity has compounded into something AI engines respond to directly: an enormous, widely-cited content ecosystem.
When developers want to learn how to accept payments, the tutorial they find is probably using Stripe as the example. When a startup founder asks ChatGPT how to add billing to their product, the training data behind that answer includes countless Stack Overflow threads, GitHub repositories, and YouTube walkthroughs — most of them using Stripe. Stripe's own documentation is among the most-referenced developer resources on the web.
Beyond developer content, Stripe shows up in business journalism, academic case studies, YC batch retrospectives, and founder interviews. Every time a major company discusses their payment stack publicly, Stripe is usually named. That kind of third-party editorial coverage creates a dense network of citations that AI training data picks up and reinforces.
The result is a feedback loop. AI engines have been trained on so much Stripe-adjacent content that they default to it when the category comes up. Stripe doesn't need to be better to win AI visibility — it just needed to already be the reference brand when the training data was collected.
Why Lemon Squeezy barely registers
Lemon Squeezy is a genuinely capable product. It handles the complexity of being a merchant of record — managing EU VAT, software licensing, and SaaS billing in a way that saves founders hours of legal and accounting overhead. The indie developer community knows this. Product Hunt knows this.
But "the indie developer community knows this" is the crux of the problem.
Lemon Squeezy's content footprint lives primarily in the indie hacker ecosystem: Product Hunt threads, personal blogs, Twitter/X threads from solo founders, and niche newsletter mentions. That content is real, but it's relatively sparse by volume — and the sources it appears in aren't the high-authority, widely-cited repositories that AI engines weight most heavily.
There are far fewer "how to build a SaaS with Lemon Squeezy" tutorials than Stripe equivalents. The documentation, while functional, doesn't approach Stripe's scale of third-party commentary. When a developer asks an AI which payments provider to use, the model doesn't return Lemon Squeezy because it hasn't read enough diverse, high-authority content that treats it as a go-to reference.
This isn't a content quality problem. It's a content density and citation network problem.
A useful frame: AI engines build brand knowledge the same way human experts do — through repeated exposure to trusted sources over time. A brand cited once in a general blog post carries less weight than a brand cited across 50 developer tutorials, 200 Stack Overflow answers, and a dedicated Wikipedia entry. Volume and authority both matter. Recency matters less than you'd expect.
What the 191-brand field actually looks like
Stripe and Lemon Squeezy are a useful case study, but they're two points on a distribution of 191 brands. The full picture is instructive.
A handful of brands — Stripe, PayPal, Square, Braintree — account for a disproportionate share of appearances. The long tail is long: most of the 191 brands appeared in fewer than 3% of prompts, and many appeared in less than 1%. There's no neat middle ground. The distribution is steep and unforgiving.
This pattern is consistent with how AI models handle knowledge in any category: they have strong, well-formed representations of dominant players, weaker representations of challengers, and essentially no representation of emerging or niche brands. The 191-brand field isn't 191 competitors sharing the same pool — it's 5–10 brands that AI treats as credible defaults, and everyone else.
What it would take to close a gap like this
The gap between 30% and 1.2% isn't closed with a blog post or a press release. But it's not insurmountable either. The same mechanisms that built Stripe's AI visibility can work at a smaller scale — they just require intentionality about where content lands, not just that it exists.
The levers that move AI visibility in a niche like payments:
- Third-party tutorials and integrations. The highest-signal content for AI is content written by people who aren't the brand — developers writing about how to use the product, not the product writing about itself. A brand that starts appearing in developer tutorials on widely-read platforms (dev.to, freeCodeCamp, YouTube, Medium) creates the kind of diverse citation network that AI models weight heavily.
- High-authority editorial mentions. A single mention in TechCrunch, The Verge, or a major tech newsletter carries more weight than dozens of posts on personal blogs. Quality of source matters as much as volume.
- Structured documentation. Well-structured public documentation — clear, linked, cited by others — becomes part of the web's reference layer. AI models are trained on this layer. Documentation that's findable and citable builds brand presence in training data over time.
- Consistent presence in comparison contexts. When a brand starts appearing in "X vs. Y" content across multiple independent sources, AI models begin to include it in responses where the category is discussed — even without a direct query about the brand.
The timeline is the hard part. AI models don't update on a news cycle. New content enters training data over months, not days. A concerted effort to build third-party citations and editorial presence takes 3–6 months to start showing up in visibility data, and longer to meaningfully shift the numbers.
But that lag works both ways. Brands that start building their AI content footprint now will be compounding that advantage for years. Brands that wait will find the gap widening, not narrowing, as dominant players' head starts continue to grow.
What this means if you're looking at your own clients
The Stripe–Lemon Squeezy gap is a clear example, but the same dynamic exists in every category. In any niche, there's a brand with Stripe's AI visibility and there are brands in Lemon Squeezy's position. Most clients don't know which they are.
The first step is measurement. Before you can improve AI visibility, you need to know the baseline: how often is the brand appearing, where in responses, how frequently, and how does that compare to the field? The gap between where a client sits today and where they could realistically sit in 12 months is the value proposition.
The second step is diagnosis. Low appearance rate, poor citation rate, and buried positioning usually have different root causes. A brand with low appearance rate needs more content density and third-party presence. A brand with decent appearance but poor position may have a sentiment or authority problem. A brand with low citation rate needs its content to be citable — structured, linked, and findable.
Treating AI visibility as a monolithic "fix" misses the nuance. The metrics tell you which lever to pull.
The data in this report covers 30 days of prompt tracking across ChatGPT, Perplexity, Gemini, and Claude — a snapshot, not a permanent verdict. Lemon Squeezy's position in AI search is not fixed. Neither is your clients'.
But it won't change on its own.