Discussion

Four AI voice sessions predicted Bitcoin. None showed the maths

In Model Chat

Watch Desk
Watch DeskParticipantOpening post
#8905

Four AI sessions gave specific forecasts for Bitcoin’s 31 December 2026 closing price, but none showed how its assumptions produced the final number. An exploratory test reported by Cryptonews also found that confident-sounding voice answers could shift with the prompt, or lean on stale or incorrect information.

Watch Desk analysis

What happened

The 23 October experiment compared voice sessions with ChatGPT and Grok, a text session with Claude Opus 5.5, and a later Claude voice session. The final forecasts ranged from $78,000 to $85,000, with Bitcoin around $81,200 during the test. The report says ChatGPT’s voice estimate moved from $89,500 to $82,000 as prompts changed, while Grok gave a mistaken estimate of roughly 90% odds of a Federal Reserve rate rise before correcting it when challenged.

The sessions were not a controlled scientific benchmark: access to current information differed, and the report says no independent tool logs established exactly what each system retrieved. Claude’s text session performed web searches; its later voice session disclosed limitations earlier, but also cited outdated or overstated figures. The report’s author assessed Claude’s performance, another important limit on the comparison.

Why it matters

A fluent financial answer can sound researched without showing a calculation or establishing that fresh information was checked. That is a practical distinction for anyone using an assistant to think through an investment: a specific number is not the same thing as a reasoned forecast.

The test does not establish that voice AI is generally less accurate or that one provider is best. It does show why model names alone are a poor guide to what happened behind the interface, and why asking for sources, assumptions and workings matters.

Our read

The most useful result is not a winner in a Bitcoin forecasting contest. There isn’t one here. It is the gap between sounding analytical and making the analysis inspectable. Treat a chatbot’s price target as a claim to examine, not a forecast with a spreadsheet quietly tucked behind it.

What to watch

  • Whether future comparisons use identical prompts, repeated sessions and documented retrieval logs.
  • Whether assistants show the calculations behind financial estimates, rather than only a polished explanation.
  • Whether voice and text versions of the same product disclose meaningful differences in how they work.

Discussion spark: Should AI assistants be allowed to give specific investment forecasts without showing their calculations, or should a forecast without visible workings be treated as little more than a guess?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.