TL;DR
- Pick Perplexity for cited research, ChatGPT for general assistant work.
- Both cost $20 monthly at the tier most buyers need.
- Perplexity routes queries across OpenAI, Anthropic, Google, and NVIDIA models.
- ChatGPT answers without an account; Perplexity requires one.
- ChatGPT Free is unlimited chat; Perplexity Free rations Pro Searches.
Perplexity and ChatGPT both cost $20 a month at the tier most people buy, and they earn that $20 in different ways. Perplexity is the pick when every claim needs a visible source: it puts citations on every answer and routes queries across models from OpenAI, Anthropic, Google, and NVIDIA. ChatGPT is the pick when you want one assistant for writing, coding, images, and research in a single subscription. Pricing and plan details below are current as of September 14, 2026.
What each one is
Perplexity is an answer engine. You ask a question, it searches the web, and it returns a written answer with inline citations pointing at the pages it drew from. The company has grown the product around that loop: Deep Research for long reports, the Computer workspace for building files and apps, and the Comet browser for working with an assistant in the page you are on.
ChatGPT is a general assistant that includes search. The plan matrix spans text chat, voice, image generation, file analysis, coding through Codex, projects, scheduled tasks, and custom GPTs. Web search is one feature among many, and it is available on every plan, free included. How its search behaves next to Google is a study of its own.
Perplexity vs ChatGPT at a glance
| Feature | Perplexity | ChatGPT |
|---|---|---|
| Built for | Cited answers and research | General assistant work |
| Free plan | Account required, then unlimited basic searches, 3 Pro Searches a day, 1 Research query a month | Unlimited text chats with GPT-5.6 Luna, limited uploads, images, and deep research |
| $20 tier | Pro: advanced models, model picker, image and video generation | Plus: GPT-6 Astra and GPT-5.6 reasoning models, projects, tasks, custom GPTs |
| Tier below $20 | None, Free then Pro at $20 | Go at $8 a month, may include ads |
| Top tier | Max at $200/month | Pro at $100, or $200 with new sign-ups paused |
| Models | OpenAI, Anthropic, Google, and NVIDIA models | OpenAI models only |
| Citations | On every answer | On answers that used web search |
| Deep research | Yes, capped by plan | Yes, capped by plan |
| Browser | Comet | Atlas and a built-in browser |
| Annual billing | Yes, Pro and Max | Not offered on individual plans |
| Chats train models | Yes, opt-out in settings | Yes, opt-out in settings |
The differences that decide it
Whose models you get
Perplexity is a router. Its pricing page lists models from OpenAI, Anthropic Claude, Google Gemini, and NVIDIA Nemotron, and Pro subscribers can set a preferred model per query. Max adds a Model Council mode that runs several frontier models against the same question. ChatGPT serves OpenAI’s lineup only: GPT-5.6 Luna on Free, GPT-6 Astra and the GPT-5.6 reasoning models on Plus, and GPT-5.6 Sol Pro on the Pro tiers. If you want Claude or Gemini answering inside the same product, only one of these two offers that. The engines also disagree more than most people expect: the same buying question gets different vendor recommendations depending on which model answers it.
Citations are the default on one of them
Every Perplexity answer carries inline citations, on the free plan included. That is the product’s defining promise, and it is why how Perplexity chooses its citations matters to publishers. ChatGPT cites sources only when it actually runs a web search. An answer written from the model’s own training carries no source list, and there is no setting that forces one. If your work requires checking where a claim came from, that difference shows up dozens of times a day. It also cuts the other way for consistency: ChatGPT’s cited sources shift between identical runs of the same prompt, and on Perplexity, free and Pro answers cite measurably different sources.
The free plans ration different things
ChatGPT Free gives unlimited text chats with GPT-5.6 Luna, subject to abuse guardrails, and rations everything around the chat: uploads, image generation, voice, deep research, and a 27K context window. Perplexity Free inverts that. Basic searches are practically unlimited, but the features that make the product worth paying for are metered hard: 3 Pro Searches a day, 1 Research query a month, no advanced models, no image generation. A free ChatGPT account is a usable daily assistant. A free Perplexity account is a search tool with a taste of the paid product. OpenAI also sells a step in between, ChatGPT Go at $8 a month, which lifts the message, upload and image limits and may show ads. Perplexity has nothing between free and $20. That gap is one reason Perplexity’s share of search traffic is still small next to its mindshare.
What $20 buys
ChatGPT Plus, at $20 a month, deepens your access to one lab’s stack: GPT-6 Astra and the GPT-5.6 reasoning models, expanded deep research and memory, projects, scheduled tasks, and custom GPTs. Perplexity Pro, at the same $20, widens your access instead: the multi-lab model picker, image and video generation, Computer access with 4,000 bonus credits, connectors for 400+ apps, and premium data sources including PitchBook, Statista, and S&P Capital IQ. Those data sources have no equivalent inside ChatGPT’s individual plans. Billing is a real difference too: Perplexity sells annual plans, currently $17 a month for Pro billed annually, while OpenAI’s help center states there is no annual billing for Go, Plus, or Pro.
Above $20
ChatGPT Pro comes in two tiers, $100 for 5x Plus usage and $200 for 20x, both with GPT-5.6 Sol Pro and a 400K reasoning context. OpenAI paused new sign-ups and upgrades to the $200 tier on 10 September 2026, so $100 is the only Pro tier you can buy today. Perplexity Max is $200 a month, or $167 a month billed annually, and leans into research volume: 10,000 monthly Computer credits, Model Council, and a memory system called Brain in research preview. Both list a $200 ceiling with different things behind the price, usage headroom on ChatGPT and research volume on Perplexity, but only Perplexity is selling its top tier right now.
Where they overlap
Several frequently claimed differences are not differences anymore. Treat these as a tie:
- Deep research: both run agentic long-form research, capped by plan on both sides.
- Browsers: Perplexity ships Comet, OpenAI ships Atlas plus a built-in browser on every ChatGPT plan.
- Image generation: both offer it on paid tiers. ChatGPT Free includes a limited, slower version; Perplexity Free has none.
- APIs: both sell developer access, billed separately from the consumer apps.
- Privacy defaults: both may use consumer conversations for training, and both offer an opt-out in settings.
What a logged-out visitor gets from each
On 14 September 2026 we put three identical prompts through both products from a US connection, logged out of each: a buying query (noise cancelling headphones), a software shortlist (project management for a 10-person team), and a current statistic (ChatGPT’s weekly active users). ChatGPT answered all three. Perplexity answered none of them.
ChatGPT opened the headphones prompt with three shopping cards, each carrying a price and a named retailer: the Sony WH-1000XM6 at $366.36 from Walmart, the Bose QuietComfort Ultra 2nd Gen at $449.00 from Staples, and the AirPods Max 2 at $509.99 from Best Buy. Below them came a ranked breakdown with citation chips on TechRadar and HotHardware. Perplexity took the same prompt, told us it had researched for seven seconds, then replaced the answer with a single line: sign up and repeat your request.
We opened six fresh sessions to be sure. Every one behaved the same way. Perplexity accepted the query, spent six or seven seconds on it, minted a real session URL, and then held the result behind a sign-in box. Whatever it found, a logged-out visitor never sees it.
That is the gap worth knowing if you measure AI search visibility. A citation in Perplexity reaches people who already have an account. A citation in ChatGPT reaches anyone who types the question.
The software prompt ran the same way. ChatGPT produced a five-row cost table for ten seats, recommended Asana at roughly $110 a month, and pointed its citation chips at project-management.com, asana.com and ClickUp’s own help site. Perplexity researched the prompt and returned the same one-line refusal.
The statistics prompt was a freshness check. ChatGPT gave a dated figure, more than 1 billion weekly active users as of September 2026, with 900 million in April 2026 for comparison, sourced to OpenAI. Perplexity refused that one too. Three prompts, one narrow finding: ChatGPT serves people who have never signed up, and Perplexity does not. So we made a free Perplexity account and ran all three again.
The same three prompts, signed in
Signed in, Perplexity answered all three. What changes is not whether you get an answer, it is where the sources sit. Its project management answer listed 10 sources and put a citation chip inside the table, on the row for each tool it named. ChatGPT’s answer to the same prompt carried three, all of them in the prose around its table and none inside it. Both put Asana at the top.
The weekly-users prompt is the one worth dwelling on, because the two answers disagree. ChatGPT dated the 900 million weekly user milestone to April 2026 and cited OpenAI for that number and for every other figure in its answer. Perplexity dated the same milestone to February 2026, put the 1 billion mark at 31 July 2026, and cited a third-party site rather than OpenAI for it. Both answers were checkable. Only one was checkable against the company that published the number.
Which one to pick
- Researchers, analysts, journalists: Perplexity Pro. Citations on every answer, a model picker, and PitchBook, Statista, and S&P Capital IQ access at $20.
- Writers, marketers, developers, generalists: ChatGPT Plus. Reasoning models, projects, scheduled tasks, custom GPTs, and Codex in one subscription.
- Budget of exactly zero: ChatGPT Free is the stronger daily tool and answers without an account. Perplexity Free is the stronger fact-checker once you create one. If you can spend $8, ChatGPT Go raises the free limits and Perplexity offers no equivalent.
- Students: Perplexity’s Education Pro is $10 a month with verification, half the price of either $20 tier.
- Heavy research volume: ChatGPT Pro at $100, since new $200 sign-ups are paused. Perplexity Max at $200 is the only top tier still selling.
If you market to AI engines
The two products are also two distinct distribution channels, and they behave differently enough to track separately. Perplexity rewards pages it can cite cleanly, ChatGPT search has its own ranking behavior worth optimizing for on its own terms, and its sourcing patterns move abruptly, as the Reddit citation collapse showed.
Meanwhile Perplexity’s model routing means Claude answers a share of its queries, so your brand’s AI visibility in one product does not transfer to the other. Track both engines, and expect their answers about your category to disagree.
Verdict
Pick Perplexity when the question ends with “and show me where that came from”. Pick ChatGPT when the question is the start of a working session: drafting, coding, planning, or building. At $20 they are priced identically and aimed at different jobs, so paying for both is not a strange setup. If you only pay for one, pay for the one that matches the majority of your queries, not the one with the better demo.