AI Assistants compared
An AI assistant answers questions, summarizes text, analyzes files and drafts writing. We score the apps and subscriptions for individuals against published criteria: features and documented limits, price, how your chats are handled, and cancellation. We do not measure answer quality ourselves; every score is based on provider information with a source and a check date.
- Rank 1:
85.1 / 100
For people who want many features at a low monthly price.
from $8.00 / month
- Rank 2:
80.7 / 100
For Google users who want to connect Gmail, Drive and Calendar at a low price.
from $4.99 / month
- Rank 3:
76.6 / 100
For people who work with long documents and want to connect services such as email.
from $17.00 / month
All services compared
- Rank 1
ChatGPT
Free planOpenAI's AI assistant with a low-cost Go plan that includes all eight features we check, some with tight limits, per the provider.
Details on ChatGPT
Scores by area
- Features 90.0
- Value for money 100.0
- Privacy & security 76.7
- Ease of use 100.0
- Fairness & support 55.0
Strengths
- Go plan at $8 a month with web search, images, data analysis and voice
- Free plan includes all eight features we check, some limited
- Temporary chat stays out of history and out of training (third-party source)
- Sign-in can be protected with security keys (third-party source)
Weaknesses
- Training on chats is on by default (can be turned off)
- Go plan may include ads, per OpenAI
- Usage limits described only as "more than Free"
- Rank 2
Gemini
Free planGoogle's AI assistant; the Google AI Plus plan costs $4.99 a month and includes 400 GB of Google storage.
Details on Gemini
Scores by area
- Features 83.3
- Value for money 100.0
- Privacy & security 66.7
- Ease of use 100.0
- Fairness & support 55.0
Strengths
- Google AI Plus at $4.99 a month, including 400 GB of storage
- Deep Research, image generation and connected apps in the free plan
- Context window of 128,000 tokens (AI Plus) or 1 million (AI Pro), per Google
- Google Account with passkey or security key
Weaknesses
- Training can only be stopped by turning off Keep Activity, which also stops saving history
- Chats read by human reviewers are kept for up to 3 years
- Usage limits only relative (2x the free limits)
- Rank 3
Claude
Free planAnthropic's AI assistant with a large context window, projects and connectors; it does not generate images.
Details on Claude
Scores by area
- Features 93.3
- Value for money 55.0
- Privacy & security 73.3
- Ease of use 100.0
- Fairness & support 55.0
Strengths
- Context window of up to 1 million tokens, including on Pro, per Anthropic
- Incognito chats; deleted chats removed within 30 days
- Apps for Windows, macOS, iOS and Android
Weaknesses
- No image generation
- Sign-in only via email link or Google account; no built-in two-factor sign-in found
- Usage limits described only as "more usage than Free"
- Rank 4
Microsoft Copilot
Free planMicrosoft's AI assistant; higher usage comes with Microsoft 365 Personal, which includes Office apps and 1 TB of OneDrive.
Details on Microsoft Copilot
Scores by area
- Features 60.0
- Value for money 90.0
- Privacy & security 76.7
- Ease of use 100.0
- Fairness & support 55.0
Strengths
- Chats not used to train foundation models in the Copilot app, per Microsoft
- Microsoft 365 Personal at $9.99 a month, with Office and 1 TB of storage
- Export your full history through the privacy dashboard
Weaknesses
- Deep Research retired in August 2026; advanced research only from Premium
- No context window published
- No temporary chat or deletion period stated
- Rank 5
Mistral Vibe (Le Chat)
Free planAI assistant from French company Mistral AI (called Le Chat until August 2026); data is hosted in the EU by default.
Details on Mistral Vibe (Le Chat)
Scores by area
- Features 40.0
- Value for money 80.0
- Privacy & security 76.7
- Ease of use 100.0
- Fairness & support 75.0
Strengths
- Data hosted in the EU by default, per Mistral
- Free plan with web search, images, voice mode and 100+ connectors
- Temporary chat; two-factor sign-in with an authenticator app
- Refund within 14 days for web subscriptions
Weaknesses
- No desktop app for Windows or macOS (browser and mobile apps)
- No context window published for Pro
- Rank 6
Grok
Free planxAI's AI assistant with access to X and the web; the entry plan costs $10 a month in the App Store.
Details on Grok
Scores by area
- Features 33.3
- Value for money 75.0
- Privacy & security 73.3
- Ease of use 100.0
- Fairness & support 55.0
Strengths
- SuperGrok Lite at $10 a month (US App Store)
- Connectors for Gmail, Google Drive and Outlook for all users
- Analyzes files and runs code, per the help pages
Weaknesses
- No desktop app, only browser and mobile apps
- Training, deletion period and export not verified (x.ai blocks automated access)
- No built-in two-factor sign-in documented
- Rank 7
Perplexity
Free planPerplexity AI's answer engine with cited sources in every answer and access to models from several providers.
Details on Perplexity
Scores by area
- Features 60.0
- Value for money 40.0
- Privacy & security 50.0
- Ease of use 50.0
- Fairness & support 75.0
Strengths
- Answers with cited sources
- Pro includes images, voice mode, Research and connectors
- Published refund windows (24 hours for monthly plans)
Weaknesses
- Free plan without images or connectors, 1 Research query a month
- No export of all chats documented
- Sign-in only via email link or Google/Apple, no passwords
- Rank 8
Meta AI
Free planMeta's free AI assistant in its own app, on the web and in Meta's apps; no paid plan is described.
Details on Meta AI
Scores by area
- Features 26.7
- Value for money 85.0
- Privacy & security 33.3
- Ease of use 100.0
- Fairness & support 50.0
Strengths
- Free, no subscription needed
- Voice chat, image generation and research reports, per the App Store listing
- Meta account with security key or authenticator app
Weaknesses
- Web search with citations, memory and code execution not documented
- Opt-out from AI training not verified for US users
- No desktop app
Scores by area
| Rank | Service | Features | Value for money | Privacy & security | Ease of use | Fairness & support | Score | Price from |
|---|---|---|---|---|---|---|---|---|
| 1 | ChatGPT | 90.0 | 100.0 | 76.7 | 100.0 | 55.0 | 85.1 | from $8.00 / month |
| 2 | Gemini | 83.3 | 100.0 | 66.7 | 100.0 | 55.0 | 80.7 | from $4.99 / month |
| 3 | Claude | 93.3 | 55.0 | 73.3 | 100.0 | 55.0 | 76.6 | from $17.00 / month |
| 4 | Microsoft Copilot | 60.0 | 90.0 | 76.7 | 100.0 | 55.0 | 72.6 | from $8.33 / month |
| 5 | Mistral Vibe (Le Chat) | 40.0 | 80.0 | 76.7 | 100.0 | 75.0 | 66.6 | from $14.99 / month |
| 6 | Grok | 33.3 | 75.0 | 73.3 | 100.0 | 55.0 | 59.6 | from $10.00 / month |
| 7 | Perplexity | 60.0 | 40.0 | 50.0 | 50.0 | 75.0 | 55.3 | from $20.00 / month |
| 8 | Meta AI | 26.7 | 85.0 | 33.3 | 100.0 | 50.0 | 50.5 | free |
Criteria and weights
Every score relies on a source. Group score = average of its criteria, total = weighted sum (0–100).
With AI assistants, what the service can do and how much it can process at once matters most, so features count 35% instead of 30%. Platforms are part of features here, so usability counts 10% instead of 15%.
Features Weight: 35 %
Capabilities
- 10: All eight features
- 8: Seven features
- 6: Six features
- 4: Five features
- 2: Three or four features
- 0: Two features or fewer, or not verified
Context window
- 10: At least 500,000 tokens
- 7: At least 128,000 tokens
- 4: At least 32,000 tokens
- 0: Less, or not stated by the provider
Platforms
- 10: Browser and all four apps
- 7: One of the five missing
- 4: Two of the five missing
- 0: Three or more missing, or not verified
Value for money Weight: 20 %
Free plan
- 10: At least seven features
- 7: Five or six features
- 4: Three or four features
- 2: Two features or fewer
- 0: No free plan, or not verified
Price
- 10: Free or up to 8.00
- 8: Up to 12.00
- 6: Up to 16.00
- 4: Up to 20.00
- 2: Up to 25.00
- 0: Over 25.00, or not verified
Privacy & security Weight: 20 %
Training on your chats
- 10: No training on chats unless you explicitly opt in
- 7: Training on by default, can be turned off in settings in all plans, history kept
- 4: Opt-out only with a catch (history is lost, form only, or paid plans only)
- 0: No opt-out, or not verified
History and deletion
- 10: Temporary chat and deletion within 30 days
- 6: Only one of the two
- 3: Chats can be deleted, but no temporary chat and no stated deletion period
- 0: Not verified
Account protection
- 10: Yes, authenticator app, passkey or security key
- 5: Only SMS code, email link, or sign-in through another account (e.g. Google, Apple)
- 0: Not verified
Ease of use Weight: 10 %
Language
- 10: Interface and help center
- 5: Interface only
- 0: No interface in the market language, or not verified
Chat export
- 10: Yes, all chats at once
- 5: Single chats only, or only on request to the provider
- 0: No export documented
Fairness & support Weight: 15 %
Cancellation and term
- 10: Monthly, cancel online and a published refund policy (or free with no contract)
- 6: Monthly and cancel online, no refunds
- 3: Cancel online, but annual term only
- 0: Not verified
Published usage limits
- 10: Concrete numbers for the main feature
- 5: Relative statements only (e.g. "5x the free plan") or numbers for secondary features only
- 0: No information, or not verified
Frequently asked questions
What is an AI assistant?
A chat service built on a language model. It answers questions, summarizes, translates, drafts text and, depending on the plan, analyzes files, images or web pages. Answers can contain errors, so check important facts at the source.
Are my chats used for training?
That depends on the provider and your settings. Many services use chats for training by default and let you turn this off in the privacy settings. If in doubt, do not enter confidential information.
What is a context window?
The amount of text the model takes into account at once in a chat, measured in tokens (roughly three quarters of a word). A larger window helps with long documents and long conversations.
Is the free plan enough?
For occasional questions, often yes. Paid plans usually offer higher usage limits, more capable models and features such as deep research. Not every provider publishes its limits as fixed numbers.
How do you rate answer quality?
We don't. Quality depends heavily on the task and changes with every model version. We only score verifiable information such as features, limits, prices and privacy rules.