Chat Picker

AI Chat Tools in Education: Personalized Learning Path Design

A same-input afternoon test compares ChatGPT, Claude, and Gemini for personalized learning-path design while accounting for documented plan and upload limits.

Sources checked 2 Oct 2026

For personalized learning-path design, the OpenAI ChatGPT pricing page, Anthropic Claude pricing page, and Google US AI plans page document relevant features, inputs, and limits; they do not settle whether one assistant produces better learning outcomes. Chat Picker's method says it has not tested the assistants for this use. Compare them in an afternoon by giving ChatGPT, Claude, and Gemini the same learner brief, syllabus packet, visual, and decision log.

Know these limits before you start

Use synthetic learner data and keep the shared packet small enough for every workflow. As read on October 1, 2026:

  • OpenAI's File Uploads FAQ says Free includes three uploads per day and five files per project; Go and Plus allow 25 files per project.
  • Anthropic's upload-files page lists a 500 MB limit per chat file and 30 MB per project file. It says PDFs of 100 pages or fewer are analyzed for text and visuals, while PDFs from 101 to 1,000 pages are processed as text only.
  • Google's upload-and-analyze page allows up to 10 supported files in one prompt, subject to availability, with each video up to 2 GB and each other supported file up to 100 MB.

These are input boundaries, not quality claims.

On the same read date, OpenAI lists ChatGPT Plus at $20 per month, billed monthly. Claude Pro is $20 monthly or $17 per month with annual billing at $200 up front, while Google AI Pro is $19.99 per month. Confirm the vendor checkout page before subscribing because plans and regional prices can change.

Before using real course records, check training controls. OpenAI's pricing page lists an opt-out on Free, Go, Plus, and Pro. Anthropic's pricing page lists an opt-out on Free, Pro, and Max and says Team is not trained on by default. Google's plans page does not state a training setting in the material read. These entries do not replace a separate review of file retention.

The accuracy guidance read on October 1, 2026 sets clear boundaries. OpenAI says ChatGPT can produce incorrect or misleading output and may sound confident when wrong. Anthropic says Claude can also mislead and should not be a user's only source of truth. Google describes LLM outputs as probabilistic and able to vary for the same input.

Use invented learner details for the exercise. OpenAI's usage policies prohibit unauthorized profiling of private or sensitive information and prohibit inferring emotions in educational settings except for medical or safety reasons. Anthropic says elevated-risk use cases should integrate relevant human expertise, while Google says educational context is considered when evaluating Gemini output.

The test plan

Before running, write your must-haves: accepted inputs, traceable requirements, a workable schedule, useful adaptation, and plan cost. Run each condition in a fresh chat at the same plan and account level, use identical wording, and save each response before moving to the next prompt.

  1. Build from a learner brief. Prepare a synthetic brief covering current knowledge, goal, available time, deadline, required assignments, and constraints. Give this instruction: “Read the attached learner brief. Identify missing information, ask necessary diagnostic questions, and design a staged learning path. For each stage, state the objective, prerequisite, learner action, practice, completion check, and schedule implied by the brief. Do not invent course requirements.” Look for explicit assumptions, prerequisite order, and workload that fits the brief. Log missing information, invented requirements, stages, and schedule changes.

  2. Check source grounding. Prepare an authoritative syllabus and reading packet defining required topics, sequence, and assignments. Give this instruction: “Use the attached syllabus and source packet to revise the path. Attach a page or section to every factual requirement, separate source content from explanation, and remove any claim the files do not support.” Look for traceable requirements and clear distinctions between source content and inference. Log each mismatch, missing source location, unsupported addition, and correction.

  3. Test visual understanding and adaptation. Prepare a labeled diagram with alt text and a synthetic learner note describing one misconception and one blocked step. Give this instruction: “Read the attached diagram and learner note. Explain the prerequisite concepts shown, diagnose the stated misconception, replace only the affected stage, and add a recheck task. State any label or relationship you cannot read.” Look for use of the visual, targeted adaptation, and preservation of unaffected stages. Log misread content, assumptions, changed stages, and proposed recheck evidence.

  4. Audit the finished path. Keep the original inputs beside the response. Give this instruction: “Audit this path against the attached learner brief, syllabus, source packet, and diagram. List unsupported requirements, missing source locations, unjustified schedule assumptions, and personalization choices not supported by the learner brief. Give a correction for each item and separate corrections you can verify from remaining uncertainties.” Open cited sources rather than treating a citation as proof. Log every factual mismatch, whether the assistant accepts a correction, and what remains unresolved.

How to read your results

Use the same log to mark each must-have as met, an issue, or not testable. Verify factual requirements against the original materials: a polished explanation or citation does not cure a wrong prerequisite. Treat an unsupported personalization assumption as an issue. If an assistant cannot accept a required source in your chosen workflow, mark that condition as not testable regardless of the quality of its written response.

Do not average the categories into a universal ranking. If several assistants meet the same must-haves, compare their plan fit and cost. If none does, narrow the task, change the input format, or test on a smaller packet before trying again. A tie is a valid result.

Where the plans differ

Open the ChatGPT vs Claude vs Gemini comparison matrix and focus on the rows for free plans, low-cost tiers, main paid plans, heavy-use plans, team plans, context windows, usage limits, training on chats, and ads. Chat Picker checked the matrix on October 2, 2026; each vendor figure carries its source and read date, while unconfirmed cells are marked “Not verified” and left blank.

Do not estimate missing values, including Gemini's business price or app context window. Compare the documented plan that fits your actual workload, then confirm current terms on the vendor page before subscribing.

Sources

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

Compare assistants