Testing the Current LLM Knowledge Gap

So here's a question I needed an answer to:

Please give me a list of all members of the U.S. House of Representatives and U.S. Senate ranked in order of age from oldest to youngest. Include state and party affiliation

And here is the response from ChatGPT's standard model:

I cannot provide a fully complete list of all 535 voting members of the U.S. House of Representatives and Senate ranked by age (with birthdates, party, and state) in a single response — that would require up-to-the-minute biographical data for every Member of the 119th Congress sourced from authoritative congressional databases or similar, which isn’t straightforwardly available in summary form on the web. However, I can provide a partial ranked list of the oldest members, with verified ages, party affiliation, and state, and explain how you can obtain the full dataset.

Here is the response from Gemini 3:

As of January 2026, the 119th Congress features a wide generational span, from its oldest member at 92 to its youngest at 29. Because a full list of all 535 members (100 Senators and 435 Representatives) is extremely long, I have organized the most significant members by age into the categories below. [NOTE: Gemini then provided a complete Google Sheet listing of all members and ages]

Couple of things stand out. One encouraging thing for people worried about AI hallucinations is that both ChatGPT and Claude Sonnet (which gave me a similar response to ChatGPT) models acknowledged that they couldn't accurately assemble the relevant data, rather than making up an answer. This is real progress and encouraging for everyone working with these models.

Second is that this reinforces our belief at Rover that Gemini currently provides the best model for our use case. Rover users need accurate, up-to-date information, even when the query requires the model to assemble an answer from a diverse set of sometimes hard-to-find sources.

Rovers' mission is to provide custom podcast programming that satisfies every user's curiosity, and it's so exciting to see this coming together in a high-quality way. Join us and hear for yourself!