How to review well — and what we won't tolerate.
Reviews on Rate My AI are worth reading because they come from real people describing real use. These rules keep it that way. Break them and we remove the review; break them repeatedly and we block the submitter.
No second-hand impressions. No "I read a benchmark that said…". No "my friend told me…". If you haven't opened the tool and put it to work on something you care about, skip the review.
Reviews must be written by a human in their own words. Don't paste output from ChatGPT, Claude, Gemini, or any other AI. Rate My AIexists because the world already has enough AI-generated content — we're the antidote to it, not another distributor of it.
Every review must be typed and submitted through the site by a person. No scrapers, curl scripts, batch tools, browser automation, or programmatic API calls — even for a single submission, and even if the underlying content is human-written. One review = one human at a keyboard.
"Really good" and "kinda bad" help nobody. Say what you used it for, what worked, what didn't. Concrete beats emphatic every time. Compare: "great model" vs. "one-shotted a 400-line React refactor but lost the thread by file three."
No brigading. No coordinated praise or takedowns. No burner accounts. No reviewing your own product. No repeat-rating from alt accounts to move a number. One person, one review per tool.
If you work at the company, invested in it, work at a direct competitor, or received compensation of any kind to review it — say so in the review itself. We'd rather see a sharp opinion with a disclosure than a scrubbed one without.
AI moves fast. A review of "Claude" from a year ago tells a reader almost nothing today. Mention which version you used (e.g. Claude Opus 4.5, GPT-5 Thinking, Cursor 1.2) and roughly when. Without that, your review has a very short shelf life.
Sharp criticism is welcome — of a model's behavior, its refusals, its guardrails, its price. Personal attacks on developers, employees, other reviewers, or user groups (race, gender, nationality, and the rest) are not.
Don't post real names of employees, API keys, private chat logs, screenshots with anyone's personal data, or anything else you wouldn't want a stranger reposting forever. Assume everything you write here is permanently public.
No affiliate links, referral codes, discount codes, "check out my Substack," Discord invites, or lead-gen. Reviews that read as ads get removed.
Review the AI, not politics, religion, or unrelated grievances. "It refuses to answer X" is fair game; using the comment section as a soapbox for something else is not.
See spam, manipulation, or a rule violation? Use the report control. Don't respond with another burner, don't dogpile, don't start a comment war.
Reviews that break these rules get removed. Repeat violations get the submitter blocked. When it's a judgment call between "sharp but real" and "manipulative or fake," we default to keeping the sharp real one and removing the manipulative one — we'd rather this site sting than lie. If a review of yours was removed and you think we got it wrong, get in touch.