Best AI for Write & Research

Where it wins

• The widest feature range of any consumer AI product, covering research, files, data, images, and code
• Deep Research and web browsing built in
• Projects, memory, and custom GPTs make repeat workflows genuinely faster
• Four price points, from USD 8 to USD 200, so you can match spend to use

Know before choosing

• The number of modes and models is genuinely confusing, and the right one is rarely obvious
• Usage allowances are per model rather than one pool, so heavy users hit caps they did not expect
• Chat data trains OpenAI models by default and you have to go and turn it off

Privacy note

OpenAI states consumer chat data is used to improve its models by default. You can disable this under Data controls, using the Improve the model for everyone toggle.

Where it wins

• Sits inside Gmail, Docs, and Sheets rather than beside them
• The cheapest paid entry point of any major assistant at USD 4.99 a month
• Bundles 5 TB of storage and YouTube Premium Lite at the Pro tier, which changes the real cost comparison
• Genuinely capable free tier

Know before choosing

• The value case collapses if you do not use Google services
• Feature availability varies by region and plan more than rivals
• Google states that even with Keep Activity off or in temporary chats, conversations are still processed for safety and security, including by human reviewers

Privacy note

Google states that with Keep Activity on, chats are used to train its generative AI models, and turning it off means future chats are not used for training. Safety processing continues either way.

Where it wins

• The strongest default privacy position of the major assistants: consumer data is not used for training unless you opt in
• Claude Code is included rather than sold separately, which is unusual value for developers
• Excellent at long form drafting, structural editing, and reasoning over large documents
• The free plan runs on Claude Sonnet 5, a frontier class model

Know before choosing

• Usage limits apply on every tier, including Max, which Anthropic footnotes on its own pricing page
• The Max tier price is quoted only as from USD 100 per month, so the 20x tier cost is not published
• Less integrated web research than Perplexity or Gemini

Privacy note

Anthropic states consumer data is not used for model training unless you opt in through Model Improvement Privacy Settings, and Incognito chats are excluded even when that setting is enabled. This is the strongest default position among the major assistants.

Where it wins

• Answers are built around citations rather than having them bolted on
• Very fast at surfacing current web material
• Low friction follow up questions make it good for exploring a topic rather than just querying it

Know before choosing

• Citations sometimes support only part of a generated sentence, so you still have to read them
• The free tier restricts you to basic models with limited weekly usage
• Perplexity's privacy notice says data is used to improve or create services and products including its AI models, and we found no dedicated training opt out for general queries. The only explicit carve out is that email content is not used for training

Privacy note

Perplexity states data is used to improve or create its services and products, including its AI models. We found no dedicated opt out toggle for general queries. The one explicit exclusion is that email content is not used to train models.

Where it wins

• Direct access to live posts on X, which no rival can match
• Grok 4.5 posts strong published results on agentic coding benchmarks
• A usable free tier

Know before choosing

• xAI's SuperGrok Lite and Heavy tier prices are rendered by JavaScript on its pricing page and we could not read them
• xAI's privacy policy explicitly tells users not to include personal information in prompts and states it cannot control what users submit
• Prompts, searches, and responses may be used for training unless you turn the setting off

Privacy note

xAI states it may use prompts, searches, and Grok responses to train its models. You can turn off the Improve the model toggle under Settings and Data Controls, or use Private Chat. Its policy tells users not to include personal information in prompts.

Where it wins

• Answers are grounded in your own wiki, docs, and databases rather than the open web
• Notion states its AI subprocessors are prohibited from using customer data to train models, which is the strongest contractual position on this list
• Enterprise adds zero data retention with the underlying model providers

Know before choosing

• Notion Agent, AI Meeting Notes, and Enterprise Search all require the USD 20 Business tier, not the USD 10 Plus tier
• Custom Agents cost a further USD 10 per 1,000 monthly Notion credits after the trial
• Priced per member, so costs scale with headcount rather than usage

Privacy note

Notion states that your data is yours and that its AI subprocessors are prohibited from using customer data to train models. Enterprise plans add zero data retention with the large language model providers.