§01tl;dr
We ran 480 identical buyer-intent queries across Claude, ChatGPT, Perplexity, and Gemini — three runs each, every day, for 30 days. 43,200 total responses. The citation behavior across models is radically different. Treating them as one channel is leaving citations on the table.
A brand cited by Claude has a 91% chance of also appearing in ChatGPT responses. But a brand cited by Perplexity only has a 54% chance of appearing in Claude. The overlap is smaller than you’d expect.
§02methodology
480 queries. 16 B2B categories, 30 queries each. Categories chosen to represent a mix of competitive density — from “best CRM software” (high competition, well-known brands) to “best CDP for mid-market” (low citation, generic landscape).
Same prompt, same system instructions, same temperature (default for each model). We extracted all brand and URL mentions using a combination of regex and Claude-3-Haiku for ambiguous cases. “Cited” = mentioned by name in a way that implied recommendation, not just passing reference.
§03claude: the specialist
Claude cites fewer brands than any other model — 2.8 per response on average — but its selections are the most technically accurate. In dev tool categories, Claude almost always names the genuine category leader, not the most-advertised one.
What gets you cited by Claude: specificity and documentation quality. Claude appears to weight pages that contain detailed how-it-works explanations, API references, and concrete use-case examples more than any other model. Vague positioning doesn’t get through.
If you’re targeting Claude specifically: invest in explanatory depth. The brands Claude cites most — Stripe, Linear, Vercel, Supabase — all have documentation that reads like it was written to teach, not to sell.
§04chatgpt: brand recognition wins
ChatGPT’s citation behavior is the most conservative. It defaults to well-known brand names, even in categories where those brands have been surpassed. In AI tooling queries from Q1 2026, ChatGPT still regularly cited products that had been discontinued or pivoted, because the brand name was familiar from its training data.
This has an important implication: getting cited by ChatGPT requires brand presence in its training data, not just current content quality. Reddit threads, press coverage, and analyst mentions from 12–24 months ago matter more for ChatGPT than they do for Perplexity.
ChatGPT also shows the strongest “Matthew effect” — already-cited brands get cited more. Category leaders get mentioned in 78% of responses; everyone else competes for the remaining mentions.
§05perplexity: cite everything
Perplexity is a fundamentally different beast. It’s a live search engine first, and its citation behavior reflects that. Average response cites 7.2 brands — more than double Gemini and 2.5× ChatGPT. Almost every brand with a clean, crawlable site will appear somewhere in Perplexity responses for their category.
What separates frequently cited from occasionally cited on Perplexity: URL structure and page titles. Perplexity appears to weight page title relevance heavily. A page titled “Stripe vs PayPal — Pricing, Features & Integration Comparison” outperforms “Stripe Pricing” on every comparison query we tested.
The other Perplexity-specific signal: Reddit. Perplexity cites Reddit threads in 38% of buyer-intent queries. If your brand isn’t appearing authentically in community discussions, you’re invisible to 38% of Perplexity’s citation surface.
§06gemini: the wildcard
Gemini has the highest run-to-run variance of any model we tested. The same query on consecutive days can return completely different brand sets. In our 30-day dataset, 44% of Gemini-cited brands appeared in fewer than half of the repeat runs for that query.
That variance is actually an opportunity for less-established brands. Because Gemini doesn’t converge on category leaders the way Claude and ChatGPT do, smaller brands show up significantly more often — especially when they have recent content (Gemini appears to weight freshness highly).
If you’re an early-stage company, Gemini is where to start. It’s the most forgiving model and the most likely to cite you unprompted.
§07per-model strategy
The mistake most brands make: optimizing for one model as if they were all the same. They’re not.
Claude: go deep on explanatory content.
ChatGPT: build brand surface area across the web.
Perplexity: nail URL structure and show up on Reddit.
Gemini: publish frequently, prioritize freshness.
The good news: the things that help you on Claude (technical depth, specificity) overlap substantially with what works on Perplexity and ChatGPT. The baseline is the same — be genuinely worth citing. The model-specific stuff is the 20% on top.
Run your domain across all four and you’ll immediately see where your gaps are by model.