Perplexity vs ChatGPT Search vs Gemini: what the paid plans actually include

The short answer

The free tiers of all three products already do cited web search; the paid tiers decide whether the product is a research tool or a demo. Perplexity's published pricing makes the split explicit: the Free plan gets citations in every answer but only limited web answers, limited file uploads and no advanced-model access, while Pro at $20/month adds preferred-model selection, expanded Computer access and deep research.

OpenAI and Google price the same rung at $20 and $19.99 respectively. ChatGPT Plus buys reasoning models, expanded deep research and larger memory and context windows; Google AI Pro buys frontier-model access (Gemini 4 Argon), Deep Research and expanded AI Mode in Search — plus 5 TB of storage and a bundle of non-AI perks that the other two vendors do not attempt.

What the free tiers include

Perplexity's Free plan is a search product with the research features rationed: the company's plan guide describes "practically unlimited basic searches" and a "very limited" amount of Pro Searches, with Perplexity choosing the model per query and no access to advanced models or image generation. The pricing page's comparison matrix grades the free column as "Limited" for web answers, file uploads and asset generation against "10x" on Pro, and shows no Computer credits and no preferred-model selection at all.

ChatGPT's Free plan is broader on conversation and narrower on everything expensive: unlimited text chats with GPT-5.6 Luna (subject to abuse guardrails), but limited messages with uploads, limited and slower image generation, limited voice chats, limited deep research, limited memory and context, and limited Codex access. Web search itself is not a paid feature — OpenAI's help centre lists it as available on Free, Go, Plus, Pro, Business, Enterprise and Edu, and even to signed-out users, subject to plan limits. One commercial caveat: OpenAI's release notes record ads rolling out to Free and Go plans (Australia, New Zealand and Canada from April 2026), while Plus and above stay ad-free.

Google frames Gemini's free app as the baseline rather than publishing a feature list for it: the Google AI plans page defines Plus, Pro and Ultra as "2x", "4x" and "up to 20x" access "in comparison to non-AI subscribers", and notes that the 15 GB of storage every Google Account includes counts inside each plan's total. The free tier's exact feature set is therefore the residual of what the paid cards claim.

Prices and plan line-ups

The three ladders are directly comparable at the $20 rung, and diverge above and below it. Perplexity and Google both sell a cheaper entry tier above free (Google AI Plus at $4.99); OpenAI slots an ad-supported Go tier between Free and Plus, whose US list price was not present in the page copy we fetched (see "What we could not verify").

Two billing asymmetries matter at checkout. OpenAI's help centre states plainly that ChatGPT Plus has no annual option — monthly only — while Perplexity's annual billing cuts Pro to $17/month and Max to $167/month, and Google's Ultra is quoted "from" $99.99 because the 20x configuration costs $199.99.

Citation behaviour

Citations are where the products differ most in what they promise. Perplexity's pricing page lists "Citations in every answer" as a Free-plan feature and its FAQ states that every answer includes inline citations showing where each claim comes from; Pro adds depth, with the Pro help article promising 10x as many citations per answer and, for paid tiers, premium data sources such as PitchBook, Statista and S&P Capital IQ.

OpenAI's language is deliberately weaker: responses that use web search "may include citations", a citation opens its source, and a Sources control lists cited sources and other relevant links. The same article warns that search results and citations can be incomplete, outdated or incorrect, and tells users to open the cited source when accuracy matters.

Google sits between the two. Gemini Apps show a Sources button at the bottom of a response or in-line, opening a side panel of related links, and cite direct quotes and web images specifically — but the help page also states that not all responses include sources, so an uncited Gemini answer is normal behaviour, not a failure.

Model selection

Perplexity is the only vendor that publishes user model selection as a paid feature. On Free, "we pick the best model to use for your query"; Pro and Max unlock a picker that Perplexity's help centre lists as Sonar, GPT-5.2, Claude Sonnet 4.6 and Gemini 3.1 Pro for referenced answers, with Sonnet 4.6 Thinking and Grok 4 for complex analysis (and o3-Pro and Claude 4.5 Opus named for Max users). Its "Best" mode, which auto-selects a model per query, carries no quota limits on any plan.

ChatGPT gates models by tier instead of offering cross-vendor choice: the published pricing matrix gives Free only GPT-5.6 Luna (plus limited GPT-5.6 Terra in Work and Codex on desktop), gives Plus GPT-6 Astra, GPT-5.6 Sol and Terra with expanded GPT-5 Thinking Mini, and reserves GPT-5.6 Sol Pro and unlimited tiers for Pro. Context scales the same way — 27K on Free, 54K on Go and Plus, 128K on Pro for the instant model.

Google publishes no user-facing model picker on its plans page; it gates model access by tier. Plus gets Gemini 3.6 Flash and Deep Research; Pro adds Gemini 4 Argon, expanded Gemini 3.1 Pro, Deep Search on Gemini 3 Pro and expanded AI Mode in Search; Ultra raises those ceilings again and adds Gemini Agent (US only, English only).

Deep research and agentic modes

All three sell a slow, multi-source research mode and meter it by tier: ChatGPT's matrix grades deep research Limited (Free), expanded (Plus) and maximum (Pro), and Google lists Deep Research as a Plus benefit with higher access on Pro and Ultra. Perplexity's equivalent split runs through credits. Ask searches — including web search — consume no credits on any plan; Computer, the agent that does multi-step work across web, files and connectors, does.

Perplexity's help centre prices that work at 100 credits = $1 and publishes the allowances: consumer Pro carries no monthly allocation (a one-time 4,000-credit signup bonus that expires in 30 days), while Max starts at 10,000 credits per month plus a one-time 35,000. In practice a light Computer task runs 100–350 credits and a "mega" project 2,400–9,800, so Pro's bonus is a trial allowance and Max is the sustained-work plan.

OpenAI meters the adjacent capability, Codex, the same way — Limited on Free and Go, yes on Plus, expanded on Pro — and its release notes position the $100 Pro tier specifically around longer, high-intensity Codex sessions. Storage follows the same ladder: 500 MB on Free, 4 GB on Go, 20 GB on Plus and 100 GB on Pro.

Who should pay for which

The honest reading of the three plan pages is that none of the $20 tiers is a bad buy; they are different bundles around the same core. The expensive mistake is buying a top tier for a capability the $20 tier already includes — Perplexity's own Max page is a volume upgrade, and OpenAI's Pro tiers are explicitly framed around usage intensity.

What we could not verify

• ChatGPT Go's US list price. The pricing page we fetched lists Go in the line-up and feature matrix, but the price element is client-rendered and absent from the copy. Not verified
• Numeric free-tier caps. All three vendors publish qualitative limits ("limited", "very limited", access multipliers) rather than message or query counts for free users.
• Whether Gemini's free app includes Deep Research at all. Google lists Deep Research as a Plus-and-above benefit without stating the free baseline explicitly. Not verified

A note on method: Google's plans and help pages were fetched live on 19 October 2026. Perplexity's and OpenAI's live pages block automated fetching, so those sources were read through Internet Archive snapshots dated between 27 August and 16 October 2026 (listed per source below); prices move, and the snapshot date is the date of the evidence.