If you read any recent digital marketing whitepapers, attend industry events, or read up on GEO/AEO/AI visibility studies, you’ll notice one thing: Perplexity is almost always there.
Now, how many people do you know who actually use Perplexity?
Here’s some Similarweb web-traffic data republished by Momentic, for May 2026:
- ChatGPT: 53.9%
- Gemini: 27.9%
- Claude: 9.2%
- Perplexity: 1.3%
(Yes, that’s web traffic, not user counts, but it’s still a decent proxy.)
So why is Perplexity treated as a tier-one priority in generative engine optimization research? Not because it has a big audience — but because it is uniquely useful to study.
Perplexity is search-native
ChatGPT, Claude, and Gemini began more as general-purpose conversational assistants: they could do most of what they did without ever touching the web. That has changed — web search is now an integrated part of how all of them work — but Perplexity was built differently from the ground up. Web retrieval was always one of its core functions, and that’s one of the reasons it’s especially attractive to GEO researchers.
Perplexity was part of GEO research from the beginning
The academic paper that introduced the term “generative engine optimization” used Perplexity as one of its primary experimental environments. The researchers tested how changes like adding statistics, quotations, citations, and clearer language affected a page’s visibility inside generated answers — and found they could improve visibility by up to 37% on specific criteria.
Because Perplexity was the test environment from the start, later researchers simply kept studying it. Research habits compound: once a platform becomes a standard benchmark, new studies use it because prior studies used it. That makes results comparable and lets you see how the field evolves over time.
Perplexity sends more referral traffic, percentage-wise, than others
In absolute terms, ChatGPT is still the big dog of AI referral traffic, accounting for more than 70% of it. But relative to the size of its user base, Perplexity sends a disproportionate amount. According to SE Ranking’s 2026 US study, citations are so central to Perplexity’s interface that users are more likely to click the links they can open, verify, and explore. And since marketers care deeply about AI referral traffic, that again makes Perplexity more interesting to GEO researchers than its audience size would suggest.
Google’s AI Overviews are a black box
In May 2026, Google said AI Overviews had more than 2.5 billion monthly users, while AI Mode had surpassed 1 billion — and those numbers are only going up. By audience size alone, Google is the most important AI search platform in the world. It is also the most difficult one to study.
AI Overviews operate inside a search engine with more than two decades of indexing infrastructure, ranking signals, link analysis, user data, spam defenses, entity understanding, and established organic results. When a website appears in an AI Overview, it’s hard to say why:
- Is it because it ranked well organically?
- Is it because Google considers the site authoritative on the topic?
- Is it because the content was easy to extract and summarize?
There are many possible factors, and untangling them is much harder in Google than in Perplexity. There’s clearly a relationship between traditional rankings and AI citations, but it’s not one-to-one. Perplexity, too, has hidden retrieval systems, ranking processes, and external indexes — but it’s still a far easier environment for researchers to extract insights from.
ChatGPT ≠ Gemini ≠ Claude ≠ Perplexity ≠ AI Overviews
Every AI platform works differently under the hood, and there’s no universal optimization that works best across all of them. To some degree, you have to decide whether you primarily want to focus on ChatGPT, Gemini, Claude, Perplexity, or AI Overviews. Some best practices apply everywhere, but each platform also differs from the others — and results differ by query, entity, and topic.
Just because something is true for Perplexity doesn’t mean it translates to other AI platforms. But it can still give you valuable data points for building data-driven hypotheses — which researchers and marketers can then put to the test.