The best AI chatbots and assistants, tested and ranked
We gave twelve AI chatbots and assistants the same 30 everyday tasks and checked every answer by hand. Here's which got the most right, which suits how you work, and what each costs.
About this list
An AI chatbot or assistant lets you talk to a large language model in plain language to get answers, draft text, work through files, write code and create images. Most add web search, voice, memory and links to your other apps. You'll find more in our AI chatbots section and our AI agents hub.
The big names are close on quality, so what separates them is fit: which apps they work with, how they cite sources, how limits work, whether chats are used for training and whether you see ads. Poe and ChatHub let you use many models in one place, and Me.bot is built around your own notes.
To compare them fairly, we gave every tool the same 30 everyday tasks for a made-up online store between October 16 and October 22, 2026, from research with sources and emails to a 20-page PDF summary, a spreadsheet question and a coding fix.
Compare at a glance
| # | Tool | Best for | Cheapest way to start | Score |
|---|---|---|---|---|
| 1 | ChatGPT | Best overall | $8 a month (Go) | 5.0 |
| 2 | Perplexity | Best for cited research | $20 a month (Pro) | 5.0 |
| 3 | Microsoft Copilot | Best for Microsoft 365 users | $9.99 a month (Microsoft 365 Personal) | 5.0 |
| 4 | Grok | Best for X users | $8 a month (Plus) | 5.0 |
| 5 | Claude | Best for long documents and coding | $20 a month (Pro) | 4.5 |
| 6 | DeepSeek | Best free chat | Free | 4.5 |
| 7 | Google Gemini | Best for Google users | $4.99 a month (Google AI Plus) | 4.0 |
| 8 | Me.bot | Best for your own notes | $10 a month (Pro) | 3.0 |
| 9 | Poe | Best for trying many models | $4.99 a month (Starter) | 3.0 |
| 10 | Duck.ai | Best for privacy | $9.99 a month (Plus) | 2.5 |
| 11 | ChatHub | Best for side-by-side answers | $19 a month, billed yearly (Pro) | 2.0 |
| 12 | Meta Muse | Best inside Meta's apps | $7.99 a month (Meta One Core) | 2.0 |
Features and terms compared
Facts come from each company's pages and our research; prices were checked on November 2 and 3, 2026.
| Tool | Free plan | Cheapest paid plan | Works best with | Training on your chats |
|---|---|---|---|---|
| ChatGPT | Yes, with ads in the US | $8 a month (Go) | Web, desktop and phone | On by default; can turn off |
| Perplexity | Yes, limited daily use | $20 a month (Pro) | Web, phone and Comet | Not stated |
| Microsoft Copilot | Yes, with a Microsoft account | $9.99 a month (Microsoft 365 Personal) | Microsoft 365 | On by default; can turn off |
| Grok | Yes, limited | $8 a month (Plus) | X and Tesla | Not stated |
| Claude | Yes, limited per window | $20 a month (Pro) | Gmail, Slack, Microsoft 365 | Depends on a setting |
| DeepSeek | Yes, chat is free | No paid chat plan | Web, phone and API | May be used; you can opt out |
| Google Gemini | Yes, with Deep Research | $4.99 a month (Google AI Plus) | Gmail, Docs, Android | On with Keep Activity |
| Me.bot | Yes, 100 messages a month | $10 a month (Pro) | Your own notes | Not stated |
| Poe | Yes, small daily points | $4.99 a month (Starter) | Many models | Not stated |
| Duck.ai | Yes, daily limit | $9.99 a month (Plus) | DuckDuckGo | No training |
| ChatHub | Yes, limits not stated | $19 a month (Pro), billed yearly per one listing | Up to 6 chatbots | Not stated |
| Meta Muse | Yes, Meta AI is free | $7.99 a month (Meta One Core) | WhatsApp, Instagram, Facebook | Not stated |
Test results side by side
The top was crowded. ChatGPT, Perplexity, Microsoft Copilot and Grok all scored 5 out of 5, and the first three made no factual errors we could find. Perplexity alone got all 30 tasks right. DeepSeek's 29 of 30 was second best, though our blind judges rated its writing lower, at 4.1 out of 5.
Below them, the drop was steep. Google Gemini got 26 tasks right with 1.4 errors per 1,000 words, and the bottom five got between 16 and 22 tasks right with 2.5 to 3.8 errors per 1,000 words. Check facts with any chatbot, but especially with these.
Every tool got the same test, between October 16, 2026 to October 22, 2026.
| Tool | Score | Correct and complete answers out of 30, checked against a written answer key | Factual errors found per 1,000 words | Blind quality score (1–5) from three judges who don't know which tool wrote what | Time to a usable answer, including retries | Cost per month for one person and for a team of five |
|---|---|---|---|---|---|---|
| ChatGPT | 5/5 | Tasks correct: 28/30 | 0.0 errors/1,000 words | 5.0/5; three blind judges | – | – |
| Perplexity | 5/5 | Tasks correct: 30/30 | 0.0 errors/1,000 words | 5.0/5; three blind judges | – | – |
| Microsoft Copilot | 5/5 | Tasks correct: 28/30 | 0.0 errors/1,000 words | 5.0/5; three blind judges | – | – |
| Grok | 5/5 | Tasks correct: 28/30 | 0.6 errors/1,000 words | 4.9/5; three blind judges | – | – |
| Claude | 4.5/5 | Tasks correct: 27/30 | 0.4 errors/1,000 words | 4.5/5; three blind judges | – | – |
| DeepSeek | 4.5/5 | Tasks correct: 29/30 | 0.0 errors/1,000 words | 4.1/5; three blind judges | – | – |
| Google Gemini | 4/5 | Tasks correct: 26/30 | 1.4 errors/1,000 words | 4.0/5; three blind judges | – | – |
| Me.bot | 3/5 | Tasks correct: 20/30 | 2.5 errors/1,000 words | 2.5/5; three blind judges | – | – |
| Poe | 3/5 | Tasks correct: 22/30 | 2.5 errors/1,000 words | 2.6/5; three blind judges | – | – |
| Duck.ai | 2.5/5 | Tasks correct: 17/30 | 3.7 errors/1,000 words | 3.1/5; three blind judges | – | – |
| ChatHub | 2/5 | Tasks correct: 17/30 | 3.8 errors/1,000 words | 2.5/5; three blind judges | – | – |
| Meta Muse | 2/5 | Tasks correct: 16/30 | 3.6 errors/1,000 words | 2.6/5; three blind judges | – | – |
ChatGPT
Best overallChatGPT scored a perfect 5 out of 5. It got 28 of 30 tasks right, made no factual errors we found, and our blind judges gave its writing 5 out of 5. It tied with Perplexity, and we put it first as the best all-rounder with the widest set of features.
It writes, codes, analyses files and creates images on web, desktop and phone. The trade-offs are ads on Free and Go in the US, limits users say are tightening, and training on your chats unless you turn it off. It has 4.8 out of 5 on the App Store but 1.6 on Trustpilot from 3,904 reviews (both checked November 2026).
Who it's for
Anyone who wants one capable assistant for most everyday work.
Skip it if
You want privacy by default. Duck.ai doesn't store chats or train on them.
What we liked
- 28 of 30 tasks right, no errors found
- Writing scored 5 out of 5 with blind judges
- Cheap Go plan at $8 a month
What we didn't
- Ads on Free and Go in the US
- Chats used for training unless you opt out
- Users report tighter limits
Key features
- Chat, writing and reasoning
- Image creation on every plan
- Memory and Temporary Chat
- Codex coding agent on Pro
Plans and prices
| Free Limited messages and uploads; Basic image creation. | Free |
| Go 10x Free messages, uploads and image creation; Longer memory. | $8 a month |
| Plus Thinking models; Deep Research and agent modes. | $20 a month |
| Pro $100 5x Plus Codex usage; Same models and features as Pro $200. | $100 a month |
Perplexity
Best for cited researchPerplexity scored 5 out of 5 and was the only tool to get all 30 tasks right. It made no factual errors we found, its sources held up in 29 of 30 citation checks, and our blind judges gave it 5 out of 5. Read our Perplexity review.
Every answer links to its web sources, and Pro adds Deep Research and a choice of models. Trustpilot users report hard-to-find cancellation and credit charges, giving it 1.4 out of 5 from 851 reviews, against 4.8 on the App Store (both checked November 2026).
Who it's for
Anyone who needs accurate answers and wants to see the sources.
Skip it if
You mainly want writing or coding help. ChatGPT or Claude suit that better.
What we liked
- All 30 tasks right
- Sources link from every answer
- Dropped its ad plan in February 2026
What we didn't
- Credits and limits apply even on Pro
- Users report hard-to-find cancellation
- Several publisher lawsuits
Key features
- Cited search answers on every plan
- Deep Research on Pro and up
- Choice of models on Pro
- Comet browser for free
Plans and prices
| Free Cited search; Basic models. | Free |
| Pro Expanded Computer access; 4,000 bonus credits. | $20 a month, or $17 a month billed yearly |
| Max 35,000 bonus credits; 10,000 monthly Computer credits. | $200 a month, or $167 a month billed yearly |
Microsoft Copilot
Best for Microsoft 365 usersMicrosoft Copilot scored 5 out of 5, with only ease of use lower, at 4.5. It got 28 of 30 tasks right, made no factual errors we found and passed all 30 citation checks. Our Copilot review has more.
The free tier covers chat, web answers, images and Office on the web, and Microsoft 365 plans add Copilot inside Word, Excel, PowerPoint and Outlook. No usage numbers are published. It has 1.5 out of 5 on Trustpilot from 510 reviews (checked November 2026).
Who it's for
People who work in Word, Excel, Outlook and Windows.
Skip it if
You don't use Microsoft's apps. ChatGPT is the better all-rounder.
What we liked
- Passed all 30 citation checks
- Free tier with Office on the web
- Copilot in desktop Office from $9.99 a month
What we didn't
- No published usage limits
- Family plan AI for the owner only
- Training on by default
Key features
- Chat and web search for free
- Copilot in Word, Excel, PowerPoint and Outlook
- Separate training opt-outs for text and voice
- Canva, Figma and Adobe app links
Plans and prices
| Free Chat; Web search. | Free |
| Microsoft 365 Personal Copilot in desktop Office apps; Higher-than-free chat, visual and image usage. | $9.99 a month, or $8.33 a month billed yearly |
| Microsoft 365 Family Everything in Personal; Sharing for up to 6. | $12.99 a month, or $10.83 a month billed yearly |
| Microsoft 365 Premium Extensive chat, visual, audio and image usage; Limited agentic research and analytics. | $19.99 a month, or $16.67 a month billed yearly |
Grok
Best for X usersGrok scored 5 out of 5, though accuracy and ease of use were 4.5. It got 28 of 30 tasks right, made 0.6 errors per 1,000 words and scored 4.9 out of 5 with our blind judges. See our Grok review.
Grok works on the web, in its apps, inside X and in Tesla cars, with voice, image and video creation. Regulators in the EU, UK, California and Canada opened probes over sexualised deepfakes in January 2026, and it has 1.4 out of 5 on Trustpilot from 536 reviews (checked November 2026).
Who it's for
X users who want strong answers plus image and video tools.
Skip it if
Its safety record worries you. ChatGPT and Perplexity scored the same with fewer errors.
What we liked
- 28 of 30 tasks right
- Plus plan from $8 a month
- Image and video creation
What we didn't
- Regulatory probes over deepfakes
- Weekly pools pause paid features when used up
- Users report billing trouble
Key features
- Chat and reasoning on every plan
- Grok Imagine images and video
- Grok Build for coding
- Works inside X and Tesla cars
Plans and prices
| Plus Paid usage pool; Voice, Imagine images, Build, API access. | $8 a month, or $6.67 a month billed yearly |
| Premium 4x Plus pool; Video, Grok Bot, newest models. | $30 a month, or $25 a month billed yearly |
| Super 15x Plus pool; Priority access at peak times. | $100 a month, or $83.33 a month billed yearly |
| Ultra 50x Plus pool; Earliest access to new products. | $200 a month, or $166.67 a month billed yearly |
Claude
Best for long documents and codingClaude scored 4.5 out of 5. It got 27 of 30 tasks right, made 0.4 errors per 1,000 words and scored 4.5 with our blind judges. Ease of use was a full 5, and value for money its lowest part at 4. Our Claude review explains why.
Claude handles long documents and code, connects to Gmail, Slack and Microsoft 365, and includes Claude Code from Pro, with no ads. Limits reset every 5 hours and weekly, and payments are non-refundable outside the EEA and UK. It has 4.7 out of 5 on the App Store but 1.5 on Trustpilot from 2,102 reviews (both checked November 2026).
Who it's for
Writers, analysts and developers working with long documents and code.
Skip it if
You want a cheap paid plan or easy refunds. ChatGPT Go costs $8 a month.
What we liked
- Ease of use scored a full 5
- Claude Code included from Pro
- No ads on any plan
What we didn't
- 27 of 30 tasks right, behind the leaders
- Non-refundable payments outside the EEA and UK
- Usage windows can pause your work
Key features
- Connectors to Gmail, Slack and Microsoft 365
- Claude Code on Pro and up
- Research and Projects on Pro
- Scheduled and cloud tasks
Plans and prices
| Free Sonnet and Haiku; Limited usage. | Free |
| Pro Opus, Sonnet, Haiku; Claude Code, Projects, Research. | $20 a month, or $17 a month billed yearly |
| Max 5x Pro usage; Fable up to half weekly limit. | $100 a month |
| Ultra 10x Pro weekly usage; No 5-hour limit for verified accounts. | $200 a month |
DeepSeek
Best free chatDeepSeek scored 4.5 out of 5. It got 29 of 30 tasks right, made no factual errors we found and passed all 30 citation checks, but our blind judges rated its writing 4.1 out of 5. Read our DeepSeek review.
The chat is free and the API is very low-priced. The catch is data: its privacy policy says personal data is stored in China, which led to bans and probes in many countries. It has 2.4 out of 5 on Trustpilot from 162 reviews (checked November 2026).
Who it's for
Cost-conscious users whose prompts contain nothing personal or confidential.
Skip it if
You handle personal or client data. Duck.ai keeps chats out of training.
What we liked
- 29 of 30 tasks right
- Free chat with no paid tier needed
- Very low API prices
What we didn't
- Data stored in China
- Writing scored 4.1 with blind judges
- Users report suspensions
Key features
- Free chat on web and phone
- File uploads
- Pay-per-token API
- Open-weight models
Plans and prices
| API Flash Input cache miss $0.15/$0.30; Output $0.60/$1.20. | Custom contact sales |
| API V4-Pro Input cache miss $0.66/$1.32; Output $1.98/$3.96. | Custom contact sales |
Google Gemini
Best for Google usersGoogle Gemini scored 4 out of 5. It got 26 of 30 tasks right, made 1.4 errors per 1,000 words, the most among the top seven, and only 25 of 30 citations checked out. Our Gemini review has more.
The free tier includes Deep Research, image creation and 15 GB of storage, and paid plans add storage and Gemini in Gmail and Docs. Chats train Google's models while Keep Activity is on. It has 4.7 out of 5 on the App Store but 1.4 on Trustpilot from 1,422 reviews (both checked November 2026).
Who it's for
People who live in Gmail, Docs and Android.
Skip it if
You need the most accurate answers. Perplexity got all 30 tasks right.
What we liked
- Free tier includes Deep Research
- Storage bundled with paid plans
- Free year of AI Plus for students
What we didn't
- 1.4 errors per 1,000 words
- Only 25 of 30 citations checked out
- Training on with Keep Activity
Key features
- Deep Research and Gemini Live for free
- Gemini in Gmail, Docs and Vids
- Video creation from AI Plus
- Canvas and Gems
Plans and prices
| Free Flash-Lite; Deep Research. | Free |
| Google AI Plus 2x Free limits; Flash. | $4.99 a month |
| Google AI Pro 4x Free limits; Gemini 4 Argon, Pro and Flash. | $19.99 a month |
| Google AI Ultra 5x or 20x Pro limits; 10,000 or 25,000 Flow credits. | $199.99 a month |
Me.bot
Best for your own notesMe.bot scored 3 out of 5. It got 20 of 30 tasks right, made 2.5 errors per 1,000 words and scored 2.5 with our blind judges. More in our Me.bot review.
That's partly because it isn't a general assistant. Me.bot builds a personal AI from your notes, voice records and meetings, and turns them into shareable talks. It has 4.6 out of 5 on the App Store from 153 reviews (checked November 2026).
Who it's for
People who want an AI that remembers their own notes.
Skip it if
You want an everyday assistant. ChatGPT got 28 tasks right to Me.bot's 20.
What we liked
- Remembers your own notes
- Turns records into talks
- Free plan with no card
What we didn't
- Only 20 of 30 tasks right
- 2.5 errors per 1,000 words
- Free plan caps chat at 100 messages a month
Key features
- Personal memories
- Chat from your own records
- Shareable talks
- Web, iOS and Android apps
Plans and prices
| Free 100 memories and messages a month; 3 talks. | Free |
| Pro Unlimited memories and messages; 30 talks a month. | $10 a month, or $8.30 a month billed yearly |
| Max | $43 a month, or $36.58 a month billed yearly |
Poe
Best for trying many modelsPoe scored 3 out of 5. It got 22 of 30 tasks right, made 2.5 errors per 1,000 words and scored 2.6 with our blind judges.
Poe is Quora's app for using models from OpenAI, Anthropic, Google, Meta and others in one place, paid for with points. It has 1.9 out of 5 on Trustpilot from 63 reviews (checked November 2026).
Who it's for
Light users who want to try many AI models in one app.
Skip it if
You need reliable answers. Copilot's free tier was far more accurate.
What we liked
- Many providers in one app
- Starter plan from $4.99 a month
- Multi-bot chats
What we didn't
- Only 22 of 30 tasks right
- Points can run out fast
- No admin controls for teams
Key features
- Chat with many models
- Multi-bot chats
- Creator bot marketplace
- Image models paid with points
Plans and prices
| Starter | $4.99 a month |
| Standard or Premium | $19.99 a month, or $16.67 a month billed yearly |
| Higher tier | $49.99 a month |
| Higher tier | $99.99 a month |
Duck.ai
Best for privacyDuck.ai scored 2.5 out of 5. It got 17 of 30 tasks right and made 3.7 errors per 1,000 words, giving it 2 for accuracy, though our blind judges scored its writing 3.1. Read our Duck.ai review.
It's DuckDuckGo's free chat. It strips identifying details before prompts reach model providers, doesn't store chats by default and says no one trains on them. Better models need a Plus or Pro plan. DuckDuckGo as a whole has 2.5 out of 5 on Trustpilot from 656 reviews (checked November 2026).
Who it's for
Privacy-minded people with occasional questions.
Skip it if
You rely on AI for accurate work. Perplexity got all 30 tasks right.
What we liked
- No chats stored by default
- No training on your chats
- Free, with no account needed
What we didn't
- Only 17 of 30 tasks right
- 3.7 errors per 1,000 words
- No memory, connectors or image tools
Key features
- Anonymous chat
- Models from several providers
- Optional sync and backup
- VPN included with Plus
Plans and prices
| Plus VPN; Advanced Duck.ai models. | $9.99 a month, or $8.25 a month billed yearly |
| Pro Everything in Plus; Claude Opus 4.8 and GPT-5.6 Sol. | $19.99 a month |
ChatHub
Best for side-by-side answersChatHub scored 2 out of 5. It got 17 of 30 tasks right, made 3.8 errors per 1,000 words, the most in our test, and scored 2.5 with our blind judges. Our ChatHub review has more.
It sends one question to up to 6 chatbots and shows the answers side by side. We couldn't find its free-plan limits, a privacy page or the company's name, and found no user reviews.
Who it's for
People who want to compare several models' answers on one screen.
Skip it if
You need accurate answers. Each of the top five got at least 27 tasks right.
What we liked
- Up to 6 chatbots in one view
- Apps for web, Chrome, phone and desktop
- One subscription instead of several
What we didn't
- Only 17 of 30 tasks right
- Most errors in our test
- Company and privacy details not published
Key features
- Side-by-side chat
- Image generation
- File analysis on paid plans
- Apps for every major platform
Plans and prices
| Pro 10,000 basic and 1,500 advanced queries; 300 images. | $19 a month, billed yearly |
| Unlimited Unlimited queries; 1,000 images. | $39 a month, billed yearly |
Meta Muse
Best inside Meta's appsMeta AI, powered by Meta's Muse models, came last with 2 out of 5. It got 16 of 30 tasks right, the fewest in our test, made 3.6 errors per 1,000 words and scored 2.6 with our blind judges. See our Meta Muse review.
Muse is the model family; the product is Meta AI, inside WhatsApp, Instagram, Facebook and Meta's glasses. It stays free, and Meta One plans add more image and video use. Meta has 1.6 out of 5 on Trustpilot from 93 reviews (checked November 2026).
Who it's for
Casual users who already spend their day in Meta's apps.
Skip it if
You need accurate answers for work. ChatGPT got 28 tasks right to Meta AI's 16.
What we liked
- Free inside apps you already use
- Image and video creation
- Instant and Thinking modes
What we didn't
- Fewest tasks right, 16 of 30
- Hard to switch off in messaging
- No document or work app connectors
Key features
- Meta AI in WhatsApp, Instagram and Facebook
- Voice, text and image input
- Works on Meta glasses
- Muse image and video models
Plans and prices
| Meta AI (free) Everyday Meta AI use. | Free |
| Meta One Core Instagram Plus, Facebook Plus and WhatsApp Plus; More Meta AI image creation and editing and video generation, powered by Muse models. | $7.99 a month |
| Meta One Premium Everything in Core; Highest Meta AI allowance across Meta's apps. | $19.99 a month |
What it costs to get started
Every tool here has a free tier, and DeepSeek's chat has no paid plan at all. The cheapest paid plans are Google AI Plus and Poe Starter at $4.99 a month, Meta One Core at $7.99, ChatGPT Go and Grok Plus at $8, Duck.ai Plus and Microsoft 365 Personal at $9.99 and Me.bot Pro at $10, all billed monthly. ChatHub Pro is $19 a month, billed yearly according to one listing.
The usual middle step is $20 a month for ChatGPT Plus, Claude Pro and Perplexity Pro, or $17 a month billed yearly for Claude and Perplexity. Top tiers reach $200 a month for ChatGPT, Claude, Perplexity and Grok. Some ChatGPT, Gemini and Poe prices come from a dated third-party listing, such as the App Store, because the official page couldn't be read.
| Tool | Cheapest way to start | Free plan | Free trial |
|---|---|---|---|
| ChatGPT | $8 a month (Go) | Yes | Not stated |
| Perplexity | $20 a month (Pro) | Yes | Not stated |
| Microsoft Copilot | $9.99 a month (Microsoft 365 Personal) | Yes | Not stated |
| Grok | $8 a month (Plus) | Yes | No |
| Claude | $20 a month (Pro) | Yes | Not stated |
| DeepSeek | Free | Yes | Not stated |
| Google Gemini | $4.99 a month (Google AI Plus) | Yes | Yes |
| Me.bot | $10 a month (Pro) | Yes | No |
| Poe | $4.99 a month (Starter) | Yes | Not stated |
| Duck.ai | $9.99 a month (Plus) | Yes | Not stated |
| ChatHub | $19 a month, billed yearly (Pro) | Yes | Not stated |
| Meta Muse | $7.99 a month (Meta One Core) | Yes | Not stated |
Prices from each tool's own pricing page (or a dated third-party listing where that page couldn't be read). Prices can change, so check before you buy.
Also worth knowing about
We didn't test these, but they come up often.
Le Chat (Mistral)
A chatbot aimed at privacy-minded users in the EU, with Pro listed at $14.99 a month. We haven't tested it.
You.com
An AI assistant listed as a Copilot alternative from $15 a month. We haven't tested it.
How we tested
Every tool got the same 30 everyday tasks between October 16 and October 22, 2026, copied in exactly, on the plan most buyers would pick and with no custom instructions. You can read the full set-up in how we test.
We checked each answer against a written answer key, counted factual errors per 1,000 words and checked every citation by hand. Three judges rated the writing from 1 to 5 without knowing which tool wrote it.
Each tool got a score out of 5, in half points, for output quality, accuracy, ease of use, features and limits, and value for money. These make the overall score, and the ranking follows it.
How to choose
Copilot fits Microsoft 365 and Windows, Gemini fits Gmail, Docs and Android, and Meta AI lives in WhatsApp and Instagram. ChatGPT and Claude connect to other apps through connectors.
Perplexity links every answer to its sources and got all 30 tasks right. Copilot and DeepSeek passed all 30 of our citation checks, while Gemini passed 25.
ChatGPT, Copilot and Gemini use consumer chats for training unless you turn it off, and DeepSeek stores data in China. Duck.ai doesn't store chats or train on them, and business plans from ChatGPT and Claude don't train by default.
Claude payments are non-refundable outside the EEA and UK, and users of Perplexity, Grok and Gemini report trouble cancelling. Start on the free tier before paying.
Terms you'll see
- Citation
- A link to the source a chatbot used, so you can check the claim yourself.
- Deep Research
- A mode that searches many web pages and writes a longer report.
- Connector
- A link that lets a chatbot read or act in another app, such as your email or files.
- Usage window
- A set period, such as five hours or a week, after which your usage limit resets.
Frequently asked questions
What is the best AI chatbot?
In our test, ChatGPT, Perplexity, Microsoft Copilot and Grok all scored 5 out of 5. ChatGPT is the best all-rounder, Perplexity got all 30 tasks right and suits research, and Copilot fits Microsoft 365 users. Claude is a strong pick for long documents and coding.
How much do AI chatbots cost?
Every tool we tested has a free tier. Most standard plans cost about $20 a month, such as ChatGPT Plus, Claude Pro and Perplexity Pro. ChatGPT Go and Grok Plus are $8 a month, Google AI Plus is $4.99, and top tiers reach $200 a month.
Which AI chatbot is best for privacy?
Duck.ai doesn't store chats by default and says neither it nor its model providers train on them. ChatGPT, Copilot, Claude and Gemini let you turn training off in settings, and ChatGPT and Claude business plans don't train by default. DeepSeek stores personal data in China.
How did we test these AI chatbots?
We gave every tool the same 30 everyday tasks in October 2026, from research with sources to emails, a PDF summary, a spreadsheet question and a coding fix. We checked answers against a written key, counted factual errors, checked citations by hand and had three blind judges rate the writing.
Do AI chatbots make mistakes?
Yes. Even the leaders slipped on a task or two, and the bottom tools made up to 3.8 errors per 1,000 words in our test. Users of every major chatbot report wrong or invented answers, so check anything important before you use it.
Do AI chatbots show ads?
ChatGPT shows ads on its Free and Go plans in the US, while Plus and higher plans are ad-free. Perplexity dropped its ad plan in February 2026, and Claude has no ad-supported tier. If ads bother you, check the plan first.
The bottom line
For most people, ChatGPT is the best place to start; pick Perplexity for cited research and Copilot if you live in Microsoft 365.
Make your case
Think a tool is missing or wrong?
If you use a tool we reviewed, or work for the company, email us with what you saw. We'll look at the evidence and update the review if it changes our findings.
Think your company belongs on this list?
Send us your product, website and the category it fits. We'll test it the same way as every other tool. It's free, and we can't promise a date or a score.
Liam Foster
If there's a strange new AI tool making its way around the internet, I've probably already bookmarked it. My favorite part is figuring out what these platforms can actually do.
Update history
- November 4, 2026: published with the results of our test, run between October 16, 2026 to October 22, 2026.