InTheWeights.com queries GPT, Claude, and Gemini for what they know about you
A new lookup tool reveals whether frontier LLMs have absorbed your name into training data—and what they confidently misremember.
InTheWeights.com is a lookup tool that queries multiple large language models to see if they know who you are. Type in a name and the site returns what GPT-5.5, Claude Opus 4.8, Gemini 3.1 Lite, and other frontier models say about that person—a quick way to check whether you've made it into their training corpora.
The results mix genuine biographical hits with confident fabrications. One test user found that Claude correctly identified his field but invented teaching credentials and a nonexistent book title, while Gemini hallucinated a different book name entirely. DeepSeek reportedly told the same user he was a Russian businessman killed in 2015. The site labels these misses "more hallucinations" at the bottom of each result page, a nod to the gap between memorization and accuracy.
The tool works by sending the same prompt to each model's API and displaying the responses side by side. It doesn't reveal training-data provenance or confidence scores—just what the model will say when asked. That makes it more of a novelty check than a forensic audit, but it does surface which models have absorbed enough public text about a given person to generate a plausible (if sometimes wrong) bio.
What's missing is any indication of when the training cutoff was or how much weight a given mention carries. If your work only hit the web in the past six months, you're unlikely to show up in models trained earlier this year. The next wave of updates—GPT-5.6, Gemini 3.2, whatever Claude ships next—will be the real test of whether recent activity makes the cut.
