Perplexity AI spent two years building something genuinely new: a search engine that answered questions instead of listing links. That head start gave it a loyal user base, press coverage, and venture capital attention. Then OpenAI showed up with a bigger distribution machine.
ChatGPT Search Enters the Room
OpenAI’s search feature, now integrated directly into ChatGPT, does not require a separate app, a separate account, or a habit change. For the hundreds of millions of people already using ChatGPT, it is simply there – a tab, a toggle, a natural extension of the interface they already trust. That kind of passive availability is something Perplexity, with its dedicated app and separate sign-up flow, cannot replicate through product design alone.
The rollout accelerated through late 2024 and into 2025, with OpenAI expanding search access to free-tier users and integrating real-time web results more deeply into the default chat experience. The company has also been working with publishers on data-sharing arrangements, attempting to address one of the early criticisms that AI search tools scraped content without compensation or credit. Whether those arrangements fully satisfy publishers is a separate argument, but the effort signals that OpenAI is treating search as infrastructure, not a side feature.
Perplexity built its reputation on something specific: a clean, citation-forward answer engine that showed users exactly where its information came from. That transparency was a genuine differentiator when the alternative was either a raw Google results page or a hallucination-prone chatbot. The problem is that ChatGPT’s search mode now surfaces citations too. The UI is different, the depth varies, but the core promise – “here is an answer and here is where it came from” – no longer belongs exclusively to Perplexity.
What Perplexity still holds is a user base that sought it out deliberately. These are people who downloaded a separate app, created an account, and built a search habit around a product that was not already on their phone. That self-selected audience tends to be more engaged and more likely to pay for a Pro subscription. But engagement metrics and conversion rates do not protect a company when a larger competitor starts absorbing the casual discovery traffic that would otherwise have expanded Perplexity’s funnel.
Where the Pressure Actually Shows Up
The competitive squeeze is not happening in a single dramatic moment. It is playing out in the quiet arithmetic of user acquisition. Every person who tries AI-powered search for the first time through ChatGPT is a person who may never need to discover Perplexity. That top-of-funnel erosion is harder to measure than a product feature comparison, but it is where early-mover advantages actually disappear.
Perplexity has been expanding aggressively to compensate. The company launched Perplexity Assistant, an agent-style feature aimed at handling multi-step tasks beyond simple queries. It has pursued partnerships with device manufacturers, including a reported deal to pre-install its app on certain Android phones in markets where Google’s default status is being legally challenged. These are smart moves, but they are moves made under pressure – attempts to find distribution channels that do not run through OpenAI’s existing user base.
Monetization is another friction point. Perplexity has been experimenting with sponsored follow-up questions, a format that places brand content within the answer flow rather than as sidebar ads. The company framed this as a native advertising model suited to conversational search. Critics called it a blurring of the line between editorial answers and paid promotion. OpenAI, backed by Microsoft’s advertising infrastructure and its own enterprise revenue, does not face the same urgency to monetize search traffic quickly, which gives it more flexibility to keep the experience clean while Perplexity balances the books.
There is also the question of trust signals. Google, for all its flaws, carries decades of brand conditioning. When people think of “looking something up,” they think of Google. Perplexity was trying to claim the next mental category: “getting an actual answer.” ChatGPT, at this point, is arguably closer to that second category in public perception than Perplexity is, and OpenAI did not have to build search from scratch to get there. It built a chat product that became synonymous with AI capability, then added search to it.
The structural problem for any AI search startup competing with OpenAI is that OpenAI’s core product – conversational AI – is itself the interface through which search happens. Perplexity built a search-first product and added conversation. OpenAI built a conversation-first product and added search. At a time when users increasingly want both in the same session, the direction of integration matters. OpenAI’s approach means search feels like an enhancement. Perplexity’s approach means conversation sometimes feels like an add-on to what is still, at heart, a query-answer engine. That is a positioning gap that product updates alone may not close. Anthropic is navigating a similar dynamic in the enterprise space, where being first does not guarantee staying first when a better-resourced competitor reframes the category.
What Perplexity Still Has
Speed and focus remain real advantages. Perplexity’s interface is still faster for pure research tasks, and its answer quality on specific factual queries is consistently competitive. The Pro tier, which includes access to multiple underlying models and deeper research modes, has genuine appeal for users who want a dedicated research tool rather than a general-purpose AI assistant that also searches. That niche is real and it is willing to pay.
But niches require defending, and Perplexity is now spending capital and management attention on distribution deals, advertising experiments, and feature races that were not part of its original pitch. The company that promised to replace the search results page is now running the same plays as everyone else trying to grow against a larger incumbent – which is precisely the position it was supposed to put Google in.
