Public trust, human voices
They're building our future. Do you trust them?
AI companies are asking for our data, our attention and our trust. Let's make that last part visible: an open trust index, shaped by the people who use it.
About two minutes. Rate the companies you know and skip the rest.
How much do you trust Anthropic?
One person, one vote
A verified sign-in. One current rating per company.
The reasons matter
Say what's behind your rating. Published without your name.
Nothing behind a curtain
An open method. No money from the companies we rate.
The companies
The first scores start with you
We're collecting the first verified ratings. A company's score appears at 30 ratings, so a handful of early votes can't decide a result.
- AnthropicClaudeCollecting ratings5 / 30
- DeepSeekDeepSeekCollecting ratings2 / 30
- GoogleGeminiCollecting ratings2 / 30
- MetaMeta AICollecting ratings2 / 30
- MicrosoftCopilotCollecting ratings2 / 30
- Mistral AILe ChatCollecting ratings2 / 30
- OpenAIChatGPTCollecting ratings4 / 30
- PerplexityPerplexityCollecting ratings3 / 30
- xAIGrokCollecting ratings2 / 30
No company matches that name.
Listed alphabetically, not ranked. See them ranked by score · Missing a company? Suggest one
Simple question, open method
A number you can look inside
The share of voters who trust a company, minus the share who don't. Every published score comes with its voter count, its average and the full spread of ratings.
Read the full methodExample: 90 trust, 50 neutral and 60 distrust ratings out of 200. Scores run from −100 to +100. Neutral ratings count in the total, so they pull a score toward zero without counting for either side.
A little more transparency
Good questions, straight answers
Why wait for 30 ratings?
A company's score, average and spread stay hidden until it has 30 ratings. That keeps a handful of early votes from deciding a result. It does not make the results a representative poll.
What does "verified" mean?
Voting needs a Google sign-in or an email address confirmed with a one-time code. One email is one voter, with one current rating per company. Someone with several email addresses could still vote more than once; the method says how we watch for that.
Who sees my rating and reason?
Your rating counts toward the published results. A reason you add is checked against the comment policy and published without your name, showing only the rating, whether you use the product, and the month. See the privacy policy.
Are the AI companies paying for this?
No. Net Trust Score takes no money from the companies it rates. Any sponsor is named in the method. Scores are never edited by hand, and every moderation action is logged.
Trust isn't a given. It's a conversation.
Help build a clearer picture of trust in AI. Start with your own view.