How

Best AI Chatbot: Features, Plans, Pricing, and Limits

A side-by-side guide to the features, plans, prices, limits, and data controls documented for ChatGPT, Claude, and Gemini, including differences between free and paid tiers.

Sources checked 2 Oct 2026

For readers comparing the best ai chatbot options, vendor pages settle only the practical baseline: [OpenAI's Plus Help page], [Anthropic's pricing page], and [Google's US AI plans page] document free access, paid consumer tiers, and plan-specific features and limits. They do not show which assistant will handle your work best. Chat Picker has not tested ChatGPT, Claude, or Gemini for quality, speed, accuracy, reliability, or feature performance, and does not rank them.

What the vendors document

All amounts are U.S. dollars. The figures below were as read on October 1, 2026. Confirm the current terms before subscribing.

Assistant	Free access	Paid plans and limits
ChatGPT	Free is the comparison baseline for Plus.	Plus is $20 per month and provides more messages and broader model and tool access than Free. It lists voice conversations, image generation, file uploads and analysis, and Deep Research tools where available. Message caps can vary with system conditions, and API usage is billed separately. OpenAI's Plus Help page
Claude	The $0 plan includes chat on web, desktop, and mobile, plus web search, file creation, and code execution. Anthropic's pricing page	Anthropic's US Pro Help page lists $20 per month, while its general pricing page displayed $34 monthly or $28 per month with annual billing and $340 upfront. The pages therefore do not settle one universal price; check your location at checkout. Pro limits reset every five hours and also have a weekly limit. Max offers 5x or 20x Pro usage per five-hour session. Anthropic's Pro Help page
Gemini	The $0 plan requires a Google Account and lists image generation and editing, Gemini Live, and 15 GB of storage.	Google AI Plus is $4.99 per month with 2x usage and 400 GB; AI Pro is $19.99 with 4x usage and 5 TB; Ultra is $99.99 with 5x or $199.99 with 20x AI Pro limits. Limits refresh every five hours until the weekly cap, and AI credits can extend them. Google's US AI plans page

Training controls also differ. OpenAI says users can opt out of having conversations used for training; Anthropic lists an opt-out for Free, Pro, and Max and says Team is not trained on by default. Google's plans page does not state a training setting, so Chat Picker leaves it unverified. A missing statement does not prove that no control exists elsewhere. See OpenAI's Plus Help page, Anthropic's pricing page, and Google's US AI plans page.

All three acknowledge inaccurate output. OpenAI's accuracy guidance says ChatGPT can be incorrect or misleading and may sound confident when wrong. Anthropic's accuracy guidance recommends checking cited sources and the original pages for context. Google's Gemini guidelines say outputs reflect training-data limits and may contain overgeneralizations.

For elevated-risk uses, OpenAI's usage policies prohibit tailored advice requiring a license without appropriate professional involvement and automated high-stakes decisions in sensitive areas without human review. Anthropic's Usage Policy requires qualified professional review for covered advice, recommendations, and subjective decisions before release. Google's guidelines say Gemini should not produce factually inaccurate outputs that could cause significant harm to health, safety, or finances.

What the documentation cannot tell you

A plan page can tell you what a vendor includes, but not how an assistant will handle your files, questions, terminology, or workflow. It cannot establish whether code runs correctly, translations preserve meaning, or an answer remains faithful through a long conversation.

A context-window figure is a capacity ceiling, not proof of reliable recall. Likewise, a monthly subscription price does not reveal the cost of one completed task. You need a controlled trial using material you can check yourself.

How to check it yourself

Run the same prompts in fresh chats, then verify the outputs independently.

Reasoning and price assumptions. Give each assistant: “Compare ChatGPT Plus at $20 per month, Claude Pro at the US help-page price of $20 per month, and Google AI Pro at $19.99 per month. Calculate each plan's difference from $20, state the currency and billing assumptions, and explain why price alone does not establish value.” Check the arithmetic and assumptions against the linked OpenAI, Anthropic, and Google pages. Record errors and unsupported conclusions.
Coding. Paste: “Write a Python function named group_anagrams(words) that groups exact anagrams, preserves the first-seen group and word order, handles an empty list, and includes tests for ['listen', 'silent', 'enlist', 'rat', 'tar', 'rat'].” Run the supplied tests and your own cases where tools are available. Look for working code, clear assumptions, and edge-case handling. Record pass or fail and any tool failure.
Translation. Paste: “Translate this into Japanese and then back into English: ‘The $20 invoice must be paid by October 1, 2026. Keep the term purchase order in English and explain why.’ Preserve the amount and date.” Look for omissions, additions, mistranslations, and terminology changes. Record every discrepancy you can verify.
Context and memory. Send: “Store this test brief for later: the launch markets are Canada and Ireland; the budget must not be exceeded; the supplier marked prohibited must remain excluded; US-only data residency is a risk. Reply only: Brief stored.” Then ask, “List every constraint from my earlier brief.” Record omitted, changed, or invented constraints.
Images, vision, and voice. Ask: “Create a flat vector illustration of a blue train at an empty station, with no text or brand names. Inspect the result and list every visible element and rendering error.” If voice is available, ask it to read: “The museum opens on Saturday morning and closes at dusk.” Inspect the actual output, record rendering or pronunciation errors, and write “not available” when the plan does not expose the feature.
Cost and caps. Paste the same test brief and request a concise checklist that preserves every constraint. Note the plan, any usage counter before and after, cap warnings, and token or charge information shown. For API use, compare input, cached-input, output, and tool charges separately using OpenAI's API pricing, Anthropic's Claude Platform pricing, and Google's Gemini API pricing. Record missing data as “not shown,” rather than estimating it.

Which rows of the comparison matter for the best AI chatbot

Open Chat Picker's Claude vs ChatGPT vs Gemini comparison matrix and check the rows for free access, low-cost tiers, main paid plans, heavy-use plans, team plans, app context windows, usage-limit wording, training on chats, and ads. Where present, feature rows for uploads, images, voice, search, code execution, memory, and storage matter too. Add API pricing only if you intend to build an integration.

Chat Picker leaves unconfirmed entries blank and marked “Not verified”; each confirmed figure carries its vendor source and read date. Its sourcing method explains the boundary between documented plan facts and testing you run yourself.