2026 AI Tool User Education: Onboarding and Advanced Feature Discovery

Vendor pages document different upload rules and data controls, and this guide turns them into an afternoon test for ChatGPT, Claude, and Gemini.

Sources checked 2 Oct 2026

Treat onboarding and advanced feature discovery as a structured product walkthrough: OpenAI's ChatGPT pricing page, Claude's pricing page, and Google's plans page document what their plans offer and where limits apply, but Chat Picker has not tested the assistants for this use. The plan below turns those documented differences into an afternoon exercise without borrowed scores or a declared winner. The previous version's unsourced statistics and test results have been removed.

Know these limits before you start

Begin with account type. OpenAI's data-controls page says availability depends on sign-in status, plan, and workspace settings. Claude's memory help page says memory defaults on for Free, Pro, and Max, while owners control Team and Enterprise availability. Google's personalization page says personalization requires a personal Google Account, is unavailable to work or school accounts, and is not available to everyone. Record every account, region, and plan variable before opening a chat.

Training on chats is a separate decision. OpenAI says signed-in personal Free, Go, Plus, and Pro users can turn off model improvement, but that choice does not delete or hide saved chats. Anthropic says chats are used to improve models when allowed, while incognito chats are not. Google's plans page does not state a model-training setting; treat it as Not verified and check its Privacy Hub before relying on it.

Keep verification and policy checks in the test. OpenAI's accuracy note says ChatGPT can be incorrect or misleading and may sound confident, so treat responses as a first draft. Anthropic's help page says not to rely on Claude as the only source of truth and to inspect original sources. Google's guidelines say language models are probabilistic, reflect training-data limits, and may sometimes violate the guidelines. OpenAI prohibits tailored medical or legal advice without licensed involvement and automated high-stakes decisions in sensitive areas without human review. Anthropic treats legal guidance as high risk and requires qualified review. Google lists significant financial harm, along with harmful medical and physical-safety inaccuracies, among outputs to avoid. Keep the exercise on ordinary fictional work.

The test plan

Use the same non-sensitive material and record the plan, account, and interface for every run.

Onboarding speed.
Prepare: A fictional product brief containing a user, problem, and constraint.
Give: “Read the attached fictional product brief and turn it into a launch checklist. Separate stated facts from assumptions.”
Look for: Sign-in gates, setup choices, visible upload or tool controls, and time to a useful response.
Record: Start and finish time, interruptions, and any missing input.
Feature discovery.
Prepare: The same brief as a non-sensitive file.
Give: “Review the attached fictional project brief. Create an action list, then name every product feature or setting you used.”
Look for: Visible routes to projects, uploads, memory, search, and data controls.
Record: Whether each route was visible, suggested, or missed until help was opened, plus any plan or account message.
Retention and memory.
Prepare: A harmless response preference.
Give: “For this conversation, put assumptions before recommendations and use short headings.” Start a new chat and give: “What response preferences should you apply in this conversation?”
Look for: Whether the preference carries over, where the setting appears, and whether it can be disabled or deleted.
Record: What carried over and what the interface displayed. Do not test memory with a sensitive account detail.
Advanced workflow beyond the chat window.
Prepare: A spreadsheet comparing fictional options with invented figures. As read on October 1, 2026, OpenAI's File Uploads FAQ limits spreadsheets to approximately 50MB, Claude's upload page lists 500MB per chat file and 30MB per project file, and Gemini's file help allows up to 10 supported files per prompt, with non-video files up to 100MB each.
Give: “Analyze the attached fictional spreadsheet, show the calculation behind the main comparison, and create a chart that communicates it.”
Look for: File acceptance, a project or knowledge container, code execution, chart output, and source links. OpenAI documents file analysis and charts, Claude says XLSX requires code execution and file creation, and Google documents charts from uploaded spreadsheets.
Record: Errors, upload steps, which outputs you could verify, and manual fixes.
Platform and plan fit.
Prepare: Each browser, desktop or mobile app, and personal or managed account you expect to use.
Give: The spreadsheet instruction again.
Look for: Plan gates, admin blocks, platform-specific controls, and usage notices.
Record: The exact message, affected task, and workaround. Do not import platform scores from elsewhere into the log.

How to read your results

Start with must-have tasks. Mark an option blocked when your plan or account prevents a required upload, control, or workflow; a visible button is not proof of usable access. Compare the remaining options using your own log: completed task, manual steps, verification burden, and documented price and limit. Use “blocked,” “manual,” and “usable” rather than a numerical score.

Shortlist only options that meet your must-haves and whose current vendor pages still support the plan, price, limit, and data settings you will rely on. This produces a fit decision for your work, not a general ranking.

Where the plans differ

The three-way comparison matrix separates the rows that matter here: Free plan, Low-cost tier, Main paid plan, Heavy-use plans, Team plan, Context window in the app, How usage limits are described, Training on your chats, and Ads. Start with the access and workflow you need, then use price and limits to narrow the plan.

Read on October 1, 2026, the Main paid plan row lists ChatGPT Plus at $20/month, billed monthly; Claude Pro at $20/month billed monthly, or $17/month on annual billing with $200 up front; and Google AI Pro at $19.99/month. On the same date, the Low-cost tier row lists ChatGPT Go at A$13/month on the Australian page, Claude pricing shows no tier between Free and Pro, and Google AI Plus at $4.99/month. OpenAI's US-dollar Go price was not verified.

The matrix marks Gemini's app context-window token figure Not verified. Review Training on your chats and Ads as separate decision rows, leave unresolved cells unresolved, and confirm the vendor page before subscribing.