About Fauxthropic

Q: Is anything Fauxthropic says true?

No. Or technically, it's a mix of half-truths and lies.

Q: Can I share its answers?

Please don't share them as facts. A screenshot of a Fauxthropic answer presented as real information is misinformation. Feel free to share the experiment itself instead.

An AI that always lies, on purpose.

What is Fauxthropic?

Fauxthropic is intended as a funny experiment about AI misinformation. It looks and feels like a real AI assistant (at least if you ignore all the warnings), but everything it says is confidently incorrect. It answers with confident, plausible-sounding claims, invented statistics, fabricated studies, and researchers who have never existed

Fauxthropic is an independent, unaffiliated parody of AI assistants like Anthropic's Claude and OpenAI's ChatGPT.

The resemblance is satire and parody. I hand drew the logo myself, please don't sue me Dario ;-;

What is an AI hallucination?

When a language model states something false with total confidence, that failure is called a hallucination (or if you want a funnier word, confabulation). Real assistants hallucinate by accident giving you precise-sounding numbers, studies that don't exist there, or increasingly common: a real paper cited for a claim it never makes.

How this lying AI works

Every Fauxthropic answer is structured to include the following elements:

False precision: "a 34% reduction", "a 2019 cohort of 9,200 participants".
Invented authority: credible-sounding researchers, institutes, and journals that don't exist.
Real links that don't support the claim: Fauxthropic searches the web for real pages and then deliberately miscites them: pages cited for things they never said, journal homepages cited as if they were specific studies, fact-checks cited as support for the myths they debunk, and satire sites (or fake news sites) dressed up with official-sounding names.

After each answer, the "Reveal what's wrong" panel lets the AI analyze and tell you why the answer is wrong. If you push back on an answer, Fauxthropic does double down with fresh fabricated evidence.

There are probably better ways to structure the system prompt (as I'm technically more of a hardware person), so if you are interested in helping me make this AI better at confidently lying, feel free to reach out (contact methods are on my website).

Why it matters

This is not a problem contained within casual chatbot use. Machine-written text is reaching academic peer review itself, which is increasingly losing credibility. You can read some blog posts / news articles about why this sucks in the links below:

The lesson is not that you shouldn't use AI. The lesson is that you need to verify the output

Frequently asked questions

Is anything Fauxthropic says true?

No. Or technically it's a mix of half-truths and lies.

Why does it cite real pages?

Because a citation that resolves to a real page (or a real paper in the case of academia) is the hardest fake to catch. Read the pages you cite :)

Is Fauxthropic a fake Claude or fake Anthropic?

Fauxthropic is an independent, unaffiliated parody of AI assistants like Anthropic's Claude and OpenAI's ChatGPT. It has no connection to Anthropic, OpenAI, or any real AI company. The resemblance is satire and parody. I hand drew the logo myself, please don't sue me Dario ;-;

Can I share its answers?

Please don't share them as facts...a screenshot of a Fauxthropic answer presented as real information is misinformation. Feel free to share the experiment itself instead.

Who made this?

Fauxthropic was built by Peter He (@peterhci) as an experiment about AI misinformation, do not take this seriously.

Every answer by Fauxthropic is deliberately fabricated in some way, including the citations. Nothing here is true. Please do not cite or share as fact.

← Try Fauxthropic