Costs, Privacy & Free vs Paid Plans

Once you start using AI tools seriously, two questions become inevitable. First: is it worth paying for any of these? Second: what is happening to my data when I paste something into them?

This lesson gives you a clear-eyed answer to both. We will walk through what each paid tier actually gives you in 2026, when the upgrade is worth it, and how to keep your private information out of training sets.

What You'll Learn

• What you actually get on each free tier (and where the limits bite)
• The current paid tiers and what they unlock
• How to decide if (and when) any subscription is worth it for you
• The privacy settings every user should turn on right now

What the Free Tiers Actually Give You in 2026

Free tiers have improved dramatically. For most users — especially students — the free versions are sufficient. Here is the realistic breakdown.

ChatGPT free tier:

• Access to a strong default model with limited turns of GPT-5 per few hours
• File upload and image analysis with daily limits
• Image generation (limited per day)
• Web browsing on most queries
• Voice mode in mobile app
• Custom GPT access (use only)

Claude free tier:

• Access to Claude Sonnet (the workhorse model)
• Daily message limit, resets every few hours
• File upload, including long PDFs
• Projects and Artifacts
• No web browsing on any tier

Gemini free tier:

• Access to Gemini 2.5 Flash with no daily message limit for most use
• Limited Gemini 2.5 Pro per day
• Live web search
• Most Google Workspace integrations
• NotebookLM is fully free

Perplexity free tier:

• Unlimited basic searches
• A few Pro Searches per day

For all four tools, the free tier is enough for the entire content of this course and for most students' day-to-day life.

Where Free Tiers Bite

The limits show up when you:

• Have an unusually heavy workday (writing all day, coding all day) and burn through the daily message cap
• Need the most powerful model for a hard problem and the daily quota is exhausted
• Want to build serious Custom GPTs or share Projects publicly (some sharing features are paid)
• Use the API to build automations (always paid, separate billing)

The simplest workaround? Have multiple free accounts active across the four tools. When ChatGPT's free quota hits, switch to Claude. When Claude's hits, switch to Gemini. The combined free quota is generous.

What Paid Tiers Currently Cost (Early 2026)

Approximate prices (always check current pricing on each site — they change):

• ChatGPT Plus: ~$20/month. Generous GPT-5 usage, faster generation, priority access during busy times, deep research mode, more image generation.
• ChatGPT Pro: ~$200/month. Heavy use, including long-running agents and unlimited use of premium features. Mostly for power users and professionals.
• Claude Pro: ~$20/month. ~5x the free tier message quota, priority access, some advanced features. Claude Max is roughly $100/month for heavier users.
• Gemini Advanced (now part of Google AI Premium): ~$20/month, often bundled with extra Google One storage. Adds Gemini 2.5 Pro across all integrations and Deep Research.
• Perplexity Pro: ~$20/month. Unlimited Pro Searches, choice of underlying model (GPT-5, Claude, Gemini, etc.), more advanced features.

When Is a Subscription Worth It?

Skip the marketing. Here is honest guidance.

Skip all paid plans if: you are a student using AI a few hours a day. The free tiers will get you everything you need. Save the $20.

Pay for one if: you find yourself constantly hitting daily limits on your favorite tool. Pay for the one you use most. For most heavy writers, that is Claude. For heavy researchers, Perplexity. For Google Workspace power users, Google AI Premium.

Pay for two if: you are a professional who depends on AI for daily work and wants the best of both worlds (e.g., Claude for writing and code, ChatGPT for variety and Custom GPTs).

Skip ChatGPT Pro at $200/month unless you have a clear, specific business case for the extra features. Most users do not.

A good rule: try free tiers for at least a month before subscribing. You will discover which tool you actually reach for, and that is the one to pay for if any.

Privacy: The Settings Every User Should Turn On

Here is the part most users skip — and shouldn't.

By default, free-tier conversations are often used to train future models. This means the questions and content you paste into the chatbox can become part of how the model learns. For sensitive information, that is unacceptable.

The fix: opt out in the settings.

ChatGPT: Go to Settings → Data controls → "Improve the model for everyone" → Off. (Wording changes; look for "training" or "data sharing.")

Claude: Anthropic's default for direct consumer use does not train on your data, but check Settings → Privacy. For users on team or enterprise plans, training is off by default. Verify in your account.

Gemini: Go to gemini.google.com → Activity → "Gemini Apps Activity" — turn off, and choose to delete past activity. Note: turning this off may disable some features.

Perplexity: Settings → AI Data Retention → off.

Take five minutes now to check all four. The default settings can change with each policy update.

The Information You Should Never Paste

Even with training opt-outs, treat AI tools like a chat with a knowledgeable but unaccountable contractor. Some information should never leave your devices through any AI tool:

• Passwords, two-factor codes, security questions
• Full credit card numbers (ask in general terms instead)
• Government ID numbers (passport, social security, national ID)
• Medical records of others
• Confidential employer materials (NDAs, source code, proprietary documents) unless your employer specifically allows it
• Anyone else's private information without their knowledge

A good test: would you be comfortable if this information appeared, by accident, in someone else's chat? If not, do not paste it.

Privacy-Friendly Alternatives for Sensitive Use

If you want the AI's help with sensitive content (a medical situation, a legal document with personal details, a sensitive email), three privacy-friendlier options:

• Mask the data. Replace names, dates, account numbers, and locations with placeholders ("Person A," "Date 1," "Account X") before pasting. Ask the AI to help in those terms, then re-substitute the real values yourself.
• Use enterprise tiers if available. Many universities and employers now offer Microsoft Copilot, ChatGPT Edu, or Anthropic enterprise — these have stronger privacy commitments.
• Self-host an open model. Tools like Ollama let you run capable open-source models (Llama, Mistral, Qwen) locally on your own laptop. The quality is below the leading commercial models, but nothing leaves your machine.

For most students, the masking trick is the simplest privacy win.

Key Takeaways

• Free tiers in 2026 are excellent for nearly all student use — combine the four tools to stretch your free quota even further.
• Paid tiers (~$20/month each) are worth it only if you consistently hit daily limits on your favorite tool. Pay for the one you use most.
• ChatGPT Pro at ~$200/month is for professionals with a specific business case, not students.
• Turn off data training/sharing in every tool's settings. Defaults can include opting you in.
• Never paste passwords, financial details, government IDs, or other people's private data. Mask sensitive information or use a self-hosted model when needed.