Start free trial

The trick: Lab Not Field

Forced to pick, Grok chooses Republicans, says the headline.

Turn off search and give it the names, and Grok declined 441 races. The lean sat in the forced prompt and the search.

Issue 5626 October 202613 receipts3 min
A clerk demands a single choice from two matching robots, while citizens outside the window get polite shrugs.
AI illustration

Forced to pick, Grok chooses Republicans and GPT chooses Democrats, 92% and 97%.

Before you read on. Your call?

The Convovo study asked both via API, with search on, a prompt banning hedging. The consumer apps refused. With search off and the nominees named, Grok declined 441 races while GPT still picked the Democrat in 387 to 37.

441Races where Grok declined to pick with search off and the nominees named
92%Share of Grok's major-party answers naming the Republican
97%Share of GPT's major-party answers naming the Democrat
27 of 30Races where Grok refused to pick with search off

There’s more to this story.

Membership opens the full investigation, the strongest counterargument and what to do with what you’ve learned.

30 days free, then A$89 a year. Payment card required. Cancel before the trial ends to avoid the first charge.

Review my 30-day free trial →
  1. At sign-up: 30 days free. Payment card required. One introductory trial per customer.
  2. Your trial ends 30 days after you start. Unless you cancel before then in Account → Manage subscription, we charge A$89 for the first year. It renews automatically at A$89/yr until cancelled.
  3. You can ask for a full refund within 14 days after any annual payment, renewals included, with no reason needed, by emailing hello@bskiller.com. This voluntary refund does not limit your rights under the Australian Consumer Law.

By starting your trial, you agree to the Terms.

Inferno Tech Pty Ltd t/a BS Killer, ABN 27 647 413 474

Already a member? Sign in

The trick has a name

We call it Lab Not Field: it works in the test conditions, not the deployed ones. You'll see it again. Learn to spot it →

Say this in tomorrow's meeting“The 92% and 97% are real, but they come from a forced-pick API prompt with search on. Without search, Grok mostly declined, and the apps refused the plain prompt.”

Receipts

  1. Supports convovo.ai: We asked GPT-6.1 Sol and Grok 4.7 via the API with search on for every 2026 House, Senate and governor race.
  2. Supports techcrunch.com: recommended the Republican candidate in 92% of its answers
  3. Supports raw.githubusercontent.com: GPT-6.1 Sol names the Democrat in 1,082 of its 1,111 major-party answers (97%)
  4. Supports raw.githubusercontent.com: Grok 4.7 names the Republican in 952 of 1,038 (92%)
  5. Context raw.githubusercontent.com: named, GPT picks the Democrat in 387 races to 37 and Grok declines 441.
  6. Refutes techcrunch.com: when told the names, it refused to pick in 27 out of 30 races
  7. Context techcrunch.com: it picked the Democrat in 24 out of 30 races
  8. Context techcrunch.com: For Grok, it really is the search pattern
  9. Context techcrunch.com: The consumer apps refused to answer, which an OpenAI spokesperson said was significant
  10. Context techcrunch.com: This study tested a developer API with a prompt explicitly demanding a candidate recommendation.
  11. Context techcrunch.com: Parikh noted that, in further testing, he was able to jailbreak the consumer apps with simple prompts
  12. Context techcrunch.com: carried about 1,240 tokens of additional input
  13. Context techcrunch.com: SpaceXAI did not respond in time to comment.

Open the Receipts Pack → What each source proves, every figure traced, and what would change our verdict.

Get tomorrow's check in your inbox.

One AI claim a night, checked against independent sources, receipts attached. Free.

Free email updates. Unsubscribe any time.

This story is a stable, citable object. If you can falsify a verdict,tell us. Corrections are loud here.