Research and search: which AI fits, by plan and price

Start with answers that cite sources. The first two tasks need web search that shows where each answer comes from. Perplexity is the only one of the six whose plan pages say every answer includes citations, and that covers Free, Pro ($20 a month, or $204 a year paid up front) and Max ($2,004 a year, paid up front on the web; a monthly price was not confirmed). ChatGPT documents web search on every plan: Free, Go ($8 a month), Plus ($20) and Pro ($100, $200 or $500). Claude documents web search on every plan: Free, Pro ($20 a month, or $200 a year paid up front) and Max ($100 or $200). For Gemini, the plan pages we checked do not describe search with source links on any plan, so that is not verified. Cursor and GitHub Copilot are coding tools: Cursor's agent can search the web and Copilot Chat can use Bing once you turn it on, but neither documents showing source links.

For reports built from many sources, compare deep research limits. ChatGPT lists limited deep research on Free and Go, expanded on Plus and maximum on the Pro plans. Claude includes Research on Pro and both Max plans; Anthropic documents it as not included on Free. Google lists Deep Research on every Gemini plan: on the free tier it may be unavailable at high demand, Google AI Plus ($4.99 a month) gets more access, Google AI Pro ($19.99 a month, or $199.99 a year) expanded access, and the Ultra plans ($99.99 and $199.99 a month) higher access. Perplexity Free allows one research query a month; Pro and Max have monthly limits. Most of these limits are described in words, not numbers. Cursor and Copilot do not document a research report mode. We have no test kit for reports yet, so open the cited pages yourself before you rely on one.

For summaries of your own reports, check uploads and how much text fits. Claude documents files of up to 500 MB each, up to 20 per chat, and a context of up to 1 million tokens on every individual plan, Free included. Google AI Pro documents up to 1,500 pages of uploads and 1 million tokens; the free Gemini tier lists 32K tokens and does not state its upload limits. ChatGPT Plus lists 256K tokens for its reasoning models and the Pro plans 400K; Free lists 27K for its Instant model. Perplexity Pro documents PDF and CSV uploads with weekly limits and up to 50 files per project, but no context size. Projects keep a report and your notes together: ChatGPT and Gemini (as Gemini Notebook) list them on every plan, Claude Free allows up to five, and Perplexity lists them on Pro and Max.

Weigh the trade-offs. At about $20 a month, ChatGPT Plus, Claude Pro, Google AI Pro and Perplexity Pro all document deep research and file uploads, and all except Google AI Pro document web search with sources on the pages we checked. Claude Pro, Google AI Pro and Perplexity Pro can be paid yearly, up front; ChatGPT has no yearly option. Perplexity Pro and Max let you pick models from OpenAI, Google, Anthropic and others in Search, which helps if you want to ask several models the same question; ChatGPT, Claude and the Gemini app list only their own company's models. OpenAI says ads may appear on ChatGPT Free and Go in the US, and Anthropic says Claude has no ads on any plan; Google and Perplexity make no current statement on the pages we checked. All prices are US list prices before tax.

Then run the test. Our source-check kit gives each assistant the same frozen packet of four documents and asks it to cite a document and paragraph for every claim, so you can see which claims trace back and which do not. The report-summary kit does the same for a summary of a quarterly report. Run both on the free tiers before you pay, and use Find my AI if you are still unsure. How we check facts is on the methodology page.