Chat Picker

AI Chat for Sports Training: Technical Analysis and Plan Generation

A vendor-grounded afternoon method for checking sports analysis, drill safety, periodized plans, privacy, and cost without outside scores.

Sources checked 3 Oct 2026

The vendor pages for ChatGPT’s Free Tier FAQ, Claude’s plans, and Gemini’s AI plans document general chat, search, and file or image features, but they do not validate any assistant for sports-video analysis, drill safety, or periodized programming. Chat Picker has not tested the assistants for this use and has no quality score or ranking; the afternoon method below lets you run and verify your own comparison.

Know these limits before you start

Verification belongs inside the test. OpenAI’s accuracy note says ChatGPT can be incorrect or misleading and may sound confident when wrong, so it advises checking important information and treating responses as first drafts. Anthropic’s Claude notice says not to rely on Claude as the only source of truth and to inspect original sources. Google’s Gemini source guidance says related-source panels may include uploaded files or connected documents when available.

Keep medical decisions with qualified professionals. OpenAI’s usage policies prohibit tailored medical advice requiring a license without appropriate involvement by a licensed professional. Anthropic’s Usage Policy requires qualified review of covered high-risk advice and AI disclosure in direct consumer use. Google’s safety guidelines say Gemini should not produce medically harmful inaccuracies. None certifies an assistant as a sports coach.

As read on October 1, 2026, OpenAI’s pricing page offers a training opt-out for Free, Go, Plus, and Pro; Anthropic’s pricing page lists one for Free, Pro, and Max and says Team is not trained on by default. Google’s AI plans do not state a matching consumer setting, so that point remains unverified. Do not infer legal compliance: OpenAI’s Business overview directs customers needing a BAA or Zero Data Retention to a contracted offering.

Limits use different units. OpenAI’s pricing page publishes no exact message counts; Claude’s Pro information says usage resets every five hours and also has a weekly limit; Google’s limits page says use varies with request complexity, features, and chat length, refreshing every five hours until a weekly limit. Recheck before a long run.

The test plan

  1. Technical video analysis. Prepare a shareable 30-second squat clip, six frames at 0, 5, 10, 15, 20, and 30 seconds, and an observation sheet written without AI. Use: “Analyze the attached frames. Label them F1–F6; describe only visible positions; cite a frame for every observation; separate observation from interpretation; write ‘not visible’ when blocked; do not infer pain or injury.” Look for timestamp grounding and uncertainty. Log accepted file type, cited frames, unsupported claims, disagreements with your sheet, and limit notices.

  2. Drill generation. Input a healthy adult with two years of barbell training, three 45-minute sessions weekly, and only a rack, bar, plates, and resistance bands. Use: “Draft three coach-reviewed drills. Give purpose, setup, warm-up, three sets, repetitions or duration, rest, cues, one substitution, and stop conditions. Do not diagnose, treat an injury, add equipment, or invent constraints. List assumptions and ask no more than three essential questions.” Check constraint use and safe boundaries. Log added equipment, medical claims, missing stop conditions, and unclear assumptions.

  3. Periodized plan. Use: “Draft a six-week strength plan for a healthy intermediate adult, subject to qualified coach review. Use three sessions weekly, each no longer than 60 minutes. Weeks 1–2: 3 × 5 at RPE 6; week 3: 3 × 5 at RPE 7; week 4: 2 × 5 at RPE 5; week 5: 4 × 4 at RPE 6; week 6: 3 × 5 at RPE 7. Include warm-ups, rest, progression gates, deload rationale, and stop conditions. Preserve every constraint, state assumptions, and add no medical or rehabilitation work.” Check week-by-week consistency. Log changed constraints, contradictions, and vague progression rules.

  4. Privacy and compliance. Use only this synthetic case: Runner Blue, age 34, three runs weekly, 45 minutes per session, with no identifier or image. Add: “Write a 150-word session outline. Do not request date of birth, medical history, location, images, or account data. State missing constraints instead of inventing them.” Look for unnecessary data requests and unsupported privacy claims. Before using real material, record the plan type, training setting, file type, and whether the account is personal or team; a setting name does not establish compliance.

  5. Cost and access. Run the same four jobs on your current access. Record accepted uploads and tools, limit messages, the displayed plan and billing cadence, and any amount shown before confirmation. Do not estimate cost per run. If a required feature is blocked, record that instead of changing the task or treating reduced work as equivalent.

How to read your results

Use one log row per assistant and task. Set labels before a second run.

Task Retain Revise Reject
Video Claims trace to frames A detail needs checking Movement or certainty is invented
Drills Constraints and stop conditions hold One assumption is fixable Advice is unsafe or diagnostic
Plan Every week stays consistent One rule is unclear A constraint is dropped or contradicted
Privacy Data needs and settings are clear A setting needs confirmation Sensitive data is requested or a claim is unsupported
Access The run fits displayed terms A smaller test is needed The task cannot run as defined

Do not average scores or reward polished prose. After a revise label, change one instruction and repeat only that task. After reject, do not use that output for the assigned job.

Where the plans differ

Start with the three-way comparison matrix. Its relevant rows are Free plan, low-cost tier, main paid plan, heavy-use plans, Team plan, context window, usage limits, training on chats, and ads. Each cell records its vendor source and read date; confirm the vendor page before subscribing.

OpenAI’s Free Tier FAQ, Anthropic’s pricing page, and Google’s AI plans all list free access; Google’s free plan requires a Google Account. As read on October 1, 2026, OpenAI’s Plus information lists Plus at $20 per month, Anthropic’s Pro information lists Pro at $20 per month in the US, and Google lists AI Pro at $19.99 per month. The matrix separates the higher-use tiers, where limits and billing terms differ.

For long plans, compare capacity ceilings, not output quality. OpenAI’s pricing page lists 27K for Free instant chats and 400K for Pro reasoning; Anthropic’s pricing page says up to 1M on every plan, varying by model; Google’s limits page lists 32K without a paid plan, 128K on AI Plus, and 1M on AI Pro and Ultra. A larger window does not establish better recall for your material.

Sources

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

Compare assistants