Chat Picker

ChatGPT vs C

ChatGPT vs Claude vs Gemini vs Copilot: Four AI Tools Compared

The four assistants use different plan prices and usage-limit systems, so this guide compares vendor-documented terms before laying out an afternoon test you can run.

Sources checked 3 Oct 2026

ChatGPT, Claude, Gemini and Microsoft Copilot have different documented prices, context claims, feature sets and plan structures, as shown by OpenAI’s ChatGPT pricing page, Anthropic’s pricing page, Google’s US plans page and Microsoft’s individual-plan page. That makes plan fit—not a claimed winner—the right place to start. Chat Picker has not tested the assistants; this page gives you a same-task test you can run in an afternoon.

Know these limits before you start

Confirm prices at checkout. As read on October 1, 2026, OpenAI’s ChatGPT pricing page lists Plus at $20 a month; Google’s US AI plans page lists AI Plus at $4.99 and AI Pro at $19.99 a month. As read on October 3, 2026, Anthropic’s Claude pricing page lists Pro at $20 monthly or $17 monthly with annual billing, with $200 due up front. Read on October 1, 2026, Microsoft’s individual-plan page lists Microsoft 365 Personal at $9.99, Family at $12.99, Premium at $19.99 and Pro at $99.99 a month. Those are Microsoft 365 bundle prices, not quoted standalone Copilot prices.

App context is also plan- and model-specific. Read on October 1, 2026, OpenAI’s pricing page lists 54K for Go and Plus instant models and 256K for reasoning; the corresponding Pro figures are 128K and 400K. Anthropic’s October 3 page says up to 1M on every Claude plan, varying by model. Google’s October 1 limits page lists 32K without an AI plan, 128K for AI Plus, and 1M for AI Pro and Ultra. Microsoft’s consumer page gives no comparable app token figure.

The vendors use different limit systems. OpenAI’s October 1 pricing page gives no exact message counts. Claude’s October 3 page uses a rolling five-hour session window plus weekly limits. Google says complexity, model, feature and chat length affect use, with a refresh every five hours until the weekly limit. Microsoft’s October 1 consumer page describes Personal as having higher-than-free use but gives no numeric request allowance.

Before uploading, check the account you will use. OpenAI’s October 1 pricing page lists training opt-out on Free, Go, Plus and Pro; its October 2 Business overview says OpenAI does not train on Business workspace data. Claude’s October 3 pricing page lists opt-out on Free, Pro and Max and says Team is not trained on by default. The Google plans page and Microsoft consumer page do not state a chat-training setting; mark both “Not verified,” not “off.”

OpenAI warns that ChatGPT can sound confident when wrong; Anthropic says not to rely on Claude as the only source of truth. Google supplies a Sources panel when available, and Microsoft Support tells users to check and improve generated content.

The test plan

Run this as a workflow test, not a general benchmark. Keep inputs low-stakes and comparable; do not use unreviewed output for medical, legal or financial decisions. For every run, use one log row with the date, tool, plan, account type, model label shown, task, requirement result, errors, sources, tools, delay, cap message and notes.

  1. Prepare the input. Choose a low-stakes memo you may share, its source notes with headings, and a spreadsheet of dates, owners and open actions. Give every service this instruction: “Use only the attached files to write a decision brief. Label documented facts and assumptions, list unresolved questions, and cite a source-note heading for each factual claim.” Record upload errors or missing content.

  2. Check account conditions. Open account controls. Record the plan, billing interval, personal or team status, training setting, enabled connectors and model label shown. Write “not found” for an absent control; do not translate that into “off.”

  3. Run the brief. Use the same prompt in a new chat for each service. Check requirement coverage, unsupported claims, source-note use, clarity and correction effort. Mark each required item met, partly met or missed, and quote the specific issue.

  4. Test files and citations. Use the same PDF and spreadsheet. Give: “Extract every date, named owner and open action. Put a source location beside each item and flag contradictions.” Record missing rows, extraction errors and broken references.

  5. Test current-source research. Give: “Research current individual list prices for ChatGPT, Claude, Gemini and Microsoft Copilot. Use first-party vendor pages only; give each page title, direct link, currency, billing period and read date.” If web tools are unavailable, record that. Open every citation and log wrong prices, missing dates and non-vendor sources.

  6. Test workflow friction. Use the memo as a project update. Give: “Turn this update into a checklist grouped by owner. Keep each task traceable to the source and do not assign work to anyone not named.” For a Microsoft 365 workflow, repeat it in Word if available; Microsoft says Personal includes Copilot in Word, Excel, PowerPoint and Outlook desktop apps, as read on October 1, 2026. Record context switching, export steps, source visibility and any cap message. If none appears, write “none observed,” not “unlimited.”

How to read your results

Set must-haves before reviewing output. A missed non-negotiable rules out that configuration for that task; a partial result calls for a targeted retest. Then compare total price, observed limits, visible data controls and required apps or connectors.

Do not import outside scores or turn one run into an overall ranking. If options remain tied, compare operating friction and repeat another task with the same log. Chat Picker contributes documented terms and method, not scores for your run.

Where the plans differ

The Claude vs ChatGPT vs Gemini matrix has rows for free plan, low-cost tier, main paid plan, heavy-use plans, team plan, app context, usage-limit descriptions, training on chats and ads. It does not include Copilot, so use Microsoft’s consumer and business pages for that column. A blank cell marked “Not verified” is unresolved, not zero or “no”; the method page explains the sourcing rule.

Sources

For current prices and limits, see the dated comparison pages.

Compare assistants