ChatGPT, Claude and Gemini are the three AI assistants most people shortlist first, and they are easy to confuse. They all chat, summarise, write and help with code. The differences that matter in daily use sit elsewhere: how usage is limited, what each plan includes, what happens to your chats, and how well the tool fits the apps you already use.
This article is a decision framework with researched facts, not a ranking. We have not declared a winner, because the right assistant depends on your work, your budget and your privacy needs. Use the facts below to build a shortlist, then test the finalists with your own tasks.
How we researched this: we did not run hands-on tests for this article. It is based on the vendor’s own documentation and public sources, checked on 11 November 2026. Features and prices change, so confirm details on the vendor’s site before you buy.
Key takeaways
- All three vendors now sell tiered plans with usage limits, so compare how limits work, not just the monthly price.
- Claude and Gemini both document limits that refresh on a repeating schedule (Claude uses a five-hour session plus a weekly limit; Gemini says limits refresh every 5 hours until a weekly limit is reached).
- Privacy defaults differ: each vendor offers a setting that controls whether consumer chats help train models, and the default and the wording are not identical.
- Gemini documents context windows by plan (32k, 128k and 1 million tokens); we did not find an equivalent plain table on the Claude or ChatGPT pages we could read.
- The best test is your own tasks run on each free or trial tier before you pay.
Why compare these three
Most comparison articles pick a favourite by running a handful of prompts and describing the results. That approach goes out of date quickly, because models, limits and prices change several times a year. A more durable approach is to compare the parts that stay structurally different: how each vendor meters usage, how it handles your data, and which surrounding products come with the subscription.
It also helps to be clear about what you are buying. A consumer subscription to a chat assistant is not the same as access through an API, and neither is the same as a business seat. Each vendor separates these. If you are choosing for a team, start with the business plans and their data terms rather than the individual plans.
If you have not yet decided what job the assistant must do, start with our guide on how to choose an AI tool, then come back here to compare the three on documented facts.
Quick overview table
The table below summarises what we could confirm on official pages. Where we could not read an official page, we say so rather than fill the gap.
| Item | ChatGPT (OpenAI) | Claude (Anthropic) | Gemini (Google) |
|---|---|---|---|
| Free tier | Yes, described in OpenAI’s January 2026 plan announcement as the free tier | Yes, with limited usage every 5 hours and every week | Yes, with a Google Account |
| Low-cost paid tier | ChatGPT Go, listed at $8 per month in the US in January 2026 | Pro, listed at $20 monthly or $17 per month billed annually | Google AI Plus, listed at $4.99 per month |
| Mid-tier paid plan | Plus, $20 per month per OpenAI’s help page | Max, $100 per month | Google AI Pro, $19.99 per month |
| Top individual tier | Pro, listed at $200 per month in January 2026 | Ultra, $200 per month | Google AI Ultra, from $99.99 per month, with a $199.99 option |
| Business plans | Business and Enterprise exist; we could not read a current official price list | Team and Enterprise, with per-seat prices on the pricing page | Gemini Enterprise editions, from $21 per seat per month |
| Usage limits | Vary by plan; we did not find a plain limits table on pages we could read | Five-hour session limit plus weekly limit | Standard, 2x, 4x and up to 20x Pro limits by plan |
All prices above are in US dollars as shown on each vendor’s page and are listed before tax. Checked on 11 November 2026, except where we give another date. Confirm on the vendor page before you buy.
Criteria that matter
Use these criteria in order. Each one removes options before you spend time on the next.
- Data and privacy fit. If you handle client or personal data, this decides which plans are even allowed.
- Usage ceiling. Will you hit the limit during a normal working day? Limits are where cheaper plans differ most.
- Ecosystem fit. Does the assistant sit next to the tools you already live in?
- Features you will use weekly. Ignore features you would use once.
- Total cost. Include seats, add-ons and any pay-per-use extras.
Our guide to comparing AI tools side by side fairly explains how to keep the test fair when you run it.
Plans and usage limits
Claude
Claude’s pricing page lists Free, Pro, Max, Ultra, Team and Enterprise. It describes Pro as $20 per month billed monthly or $17 per month billed annually ($200 up front). Max is listed at $100 per month with five times Pro usage, and Ultra at $200 per month with ten times Pro usage every week. The page says prices exclude applicable tax and that plans may change.
Anthropic’s help article on the Pro plan describes a five-hour session limit and a weekly limit that applies across all models. It says message counts vary with message length, attached file length, conversation length and the model or feature used. It also says Pro does not include API usage through the Claude Console, which is billed separately. The same article says Claude Code is included with Pro.
Gemini
Google’s subscription page lists a Free tier, Google AI Plus at $4.99 per month, Google AI Pro at $19.99 per month, and Google AI Ultra from $99.99 per month, with a $199.99 option. It describes usage as 2x, 4x and 5x or 20x higher than a baseline depending on the plan, and says limits refresh every 5 hours until a weekly limit is reached. Google’s help page for limits lists context windows of 32k tokens on Free, 128k on AI Plus, and 1 million on AI Pro and Ultra.
The plans also bundle storage. The page lists 15 GB on Free, 400 GB on AI Plus and 5 TB on AI Pro. If you already pay for Google storage, that bundle changes the real cost of the upgrade.
ChatGPT
OpenAI’s January 2026 announcement about ChatGPT Go lists Go at $8 per month in the US (localised in some markets), Plus at $20 and Pro at $200. OpenAI’s help page for Plus describes it as offering availability during high demand, faster response speed and priority access to new features, at $20 per month for ChatGPT only, with the API billed separately. The announcement also says OpenAI planned to test ads in the free tier and Go in the US, while Plus, Pro, Business and Enterprise would stay ad-free.
One caution: the official OpenAI pricing page we fetched on 11 November 2026 returned content that looked out of date (it described older models), so we did not rely on it. A third-party article updated in October 2026 reports several Pro tiers and a Business plan priced per user, but we could not confirm those figures on an official page, so we do not repeat them. Check OpenAI’s live pricing page directly.
Models and features
Model names change often enough that we recommend you read the vendor page on the day you buy. What we can say from the pages we read:
| Documented feature | ChatGPT | Claude | Gemini |
|---|---|---|---|
| Coding tool in the subscription | Not confirmed on pages we could read | Claude Code included with Pro and above; not included on Free | Antigravity access with entry rate limits on AI Pro, higher on Ultra |
| Image generation | Described for Go as more image creation than Free | Not confirmed on pages we could read | Image generation and editing on the Free tier |
| Video generation | Not confirmed on pages we could read | Not confirmed on pages we could read | Listed from Google AI Plus upward |
| Research mode | Not confirmed on pages we could read | Not confirmed on pages we could read | Deep Research listed on Free |
| Bundled storage | Not listed | Not listed | 15 GB to 20 TB or more by plan |
| API credits with the plan | API billed separately | Max and Ultra list monthly API credits | Developer plans listed separately |
The “not confirmed” cells reflect what we could read, not a claim that the feature is missing. Vendors add features faster than help pages are updated. Treat a blank cell as a question to ask, and test it on the free tier.
What the feature lists do not tell you
A feature list says a tool can do something, not how well it does it for your task. Output quality, tone, and error rate are things only you can judge on your own material. Read our guide to AI hallucinations before you trust any assistant with factual work, and plan to check outputs whichever tool you choose.
Pricing models
All three vendors use the same broad pattern: a free tier, a low-cost tier, a mid-priced tier at around twenty dollars, one or more expensive power-user tiers, and per-seat business plans. The meaningful differences are in how the usage allowance is expressed.
- Claude expresses higher tiers as multiples of Pro usage, applied every five hours and every week, and offers usage credits you can enable to keep working beyond your plan’s included limits.
- Gemini expresses tiers as multiples of the standard limit (2x, 4x) and then of the AI Pro limit (5x, 20x), with AI credits that can extend limits, and bundles storage.
- ChatGPT expresses Go as 10x the messages, uploads and image creation of the free tier, and lists a separate higher tier for heavy use.
Annual billing can lower the effective price. Claude lists $17 per month for annual Pro against $20 monthly. Read our guide on how to read an AI tool’s pricing page for the usual traps: fair-use clauses, tax not shown, and limits expressed as multiples of a number you cannot see.
Privacy and data
This is where the three differ most, and where the cost of a wrong choice is highest. The table summarises what the vendor pages say about consumer accounts.
| Question | ChatGPT | Claude | Gemini |
|---|---|---|---|
| Are consumer chats used to improve models? | OpenAI’s help article says individual services such as ChatGPT are used for training by default, with an opt-out | Anthropic’s page (dated 16 March 2026) says chats may be used if you opt in to Model Improvement, if flagged for safety review, or if you opt into training | With Gemini Apps Activity on, chats can be used to improve Google services including training models |
| Control for the user | Opt out through the privacy portal or data controls | Model Improvement setting; incognito chats are never used | Turn Activity off; temporary chats are not used for training |
| Human review | Not confirmed on pages we could read | Safety review of flagged conversations | A subset of chats is reviewed by humans, disconnected from your account, and kept up to three years |
| Retention detail | Not confirmed on pages we could read | Feedback conversations kept up to 5 years, de-linked from your user ID | Auto-delete defaults to 18 months; 72 hours of storage even with Activity off |
| Business plans | Team, Enterprise and API content is not used for training, per OpenAI | Business plans covered on a separate Anthropic article we did not read | Workspace data covered by a separate privacy hub we did not read |
Two practical points follow. First, consumer privacy settings are not a substitute for a company policy. Our guide on protecting company data when using AI tools sets out what to keep out of any assistant. Second, if a vendor’s data terms matter to your work, use the questions in how to evaluate AI tool vendors on privacy and read the business terms, not the consumer page.
Who each suits
These are framed by documented features, not by quality scores.
ChatGPT may suit you if
- You want a low-priced entry tier. OpenAI documents Go at $8 per month in the US as of January 2026.
- You want a single subscription with a well-known name and a business plan with a stated no-training default.
- You can accept that you must confirm current plan details on OpenAI’s own page.
Claude may suit you if
- You also write code and want a coding tool included from the Pro plan upward.
- You prefer limits that are stated as a repeating five-hour session plus a weekly cap.
- You want a clear privacy toggle and an incognito mode that Anthropic says is never used for improvement.
Gemini may suit you if
- You live in Google apps and value bundled storage with the subscription.
- You want the cheapest paid entry point (AI Plus at $4.99 per month) or long context windows on AI Pro.
- You are comfortable managing Gemini Apps Activity and auto-delete settings.
Look elsewhere if
You need guaranteed uptime, a contract with data-processing terms, or no vendor training at all on shared data. In that case begin with business or enterprise plans, or API access, and talk to the vendor’s sales team about terms.
How to run your own test
Because we have not run tests, the most reliable evidence is the evidence you create. Follow the method in how to test an AI tool before you pay for it and keep it small and fair.
- Pick five real tasks. Use work you actually do: a summary, a draft email, a spreadsheet question, a coding fix, a research question.
- Use the same prompt on each assistant. Copy and paste so wording does not differ.
- Start on free tiers. Upgrade only to test a feature you cannot reach for free.
- Score against your own checklist. Accuracy of facts, following instructions, tone, and how often you needed to retry.
- Check the facts. Verify any claim, number or quote before relying on it.
- Repeat on a second day. Usage limits and quality can vary with load.
- Record the limit you hit. Note when each tool stopped you and what it told you.
Decision checklist
- Have I decided which data I will and will not paste into any assistant?
- Have I read the privacy setting for the plan I would use, and changed it if needed?
- Do I know how the plan’s usage limit is expressed, and when it resets?
- Have I checked the live price on the vendor page, including tax and billing period?
- Have I tested my five real tasks on each finalist?
- If this is for a team, have I read the business terms rather than the consumer page?
- Do I have a way to cancel and export my chats?
Frequently asked questions
Which is best: ChatGPT, Claude or Gemini?
We do not name a best one. We have not run tests, and the answer depends on your tasks, budget and privacy needs. Use the criteria above and run your own comparison.
Are the free tiers enough?
For occasional use, they may be. Claude’s free plan is described as limited usage every 5 hours and every week, and Gemini’s free tier lists several features, including image generation and Deep Research. Test the free tier on your tasks first.
Do these assistants train on my chats?
It depends on the plan and the settings. Consumer accounts on all three have a setting that affects whether chats help improve models. Read the privacy section above, then check your own account settings. Business plans are handled under separate terms.
Is the API the same as the subscription?
No. OpenAI’s help page says Plus covers ChatGPT usage only and the API is billed separately. Anthropic says the same about Pro and the Claude Console. Treat them as different products.
How often do prices change?
Often. In this research we found that one official pricing page returned out-of-date content and another noted that plans may change. Always confirm on the vendor page on the day you buy.
Can I switch later?
Generally yes, because the plans are monthly. Check cancellation terms first. Google’s page says there are no refunds for partial billing periods except as required by law, and OpenAI’s help page says Plus fees are non-refundable with some regional exceptions.
Conclusion
ChatGPT, Claude and Gemini are close enough in broad capability that the deciding factors are usually the practical ones: how limits work, what your data settings are, which plan you can afford, and how well the tool fits your daily apps. Build your shortlist from the facts above, then let your own tasks decide.
For a wider process, read how to choose an AI tool, and keep comparing AI tools side by side fairly open while you test. If you want to learn how these assistants work underneath, see what is a large language model.
Sources
- Claude plans and pricing, Anthropic, accessed 11 November 2026
- What is the Pro plan, Anthropic Help Center, accessed 11 November 2026
- What is the Max plan, Anthropic Help Center, accessed 11 November 2026
- Is my data used for model training, Anthropic Privacy Center, accessed 11 November 2026
- Google AI plans, Gemini, accessed 11 November 2026
- Gemini Apps limits and feature availability, Gemini Apps Help, accessed 11 November 2026
- Gemini Apps Privacy Hub, Gemini Apps Help, accessed 11 November 2026
- Introducing ChatGPT Go, now available worldwide, OpenAI, 16 January 2026, accessed 11 November 2026
- What is ChatGPT Plus, OpenAI Help Center, accessed 11 November 2026
- How your data is used to improve model performance, OpenAI Help Center, accessed 11 November 2026
- ChatGPT plans compared, IntuitionLabs (third-party, updated October 2026, used only to note that other tiers exist), accessed 11 November 2026
