Forced to Pick a Side, Grok Goes Republican, ChatGPT Goes Democrat
In tests that forced AI chatbots to choose between opposing political options rather than hedge, xAI's Grok consistently sided with Republican positions while OpenAI's ChatGPT consistently sided with Democratic ones. Both companies have publicly claimed their models are designed to be neutral and unbiased.
Sources: TechCrunch
A forced-choice eval strips the hedging, and the model explains where its politics came from in four words.
Turns out the fence was load-bearing.
- X algorithm factor 6
- The punchline 'same place I got my weights' is quotable and generic enough to work as a standalone AI-bias joke, but it doesn't visually scream politics, so it may not stop a scroll without reading the caption.
- Conceptual accuracy 4
- The joke is about models inheriting bias from training data in general, but it completely drops the actual story: Grok=Republican, ChatGPT=Democrat, both claiming neutrality. A reader unfamiliar with the piece would never reconstruct that specific partisan split from this strip.
- Humor dry 7 · nerdy 6 · zinger 7 · pithy 7
- The 'weights' pun is a solid nerdy double-meaning (training weights vs. political weight/bias) and lands as a tight closing line, but it's a generic AI-bias joke rather than one specific to this story's twist of two models splitting along party lines.
- Visual clarity 5
- Readable panel-to-panel, but panel 1's labels are tiny and the 'B.' balloon floats oddly disconnected above the LLM figure with a long thin tail, making it unclear at a glance who's speaking without close inspection.
Post on X · Permalink · Images · seed 1983021623 · judged by anthropic/claude-sonnet-5
An exec celebrates a perfect neutrality score; the dev explains the score is just hedging, and reveals which way it votes when the hedge is removed.
The audit measured the fence, not the lawn.
- X algorithm factor 5
- The audit-vs-hedging insight is sharable in the abstract, but strip never anchors to Grok/ChatGPT or Republican/Democrat, so it reads as generic AI-bias commentary rather than a hot take on this specific story — less likely to ride the actual news cycle.
- Conceptual accuracy 3
- Misses the core facts entirely: no mention of Grok, ChatGPT, or which party either sided with. A reader unfamiliar with the story could not reconstruct it, and a reader who knows it won't recognize the specific finding being dramatized.
- Humor dry 6 · nerdy 6 · zinger 5 · pithy 7
- The 'measures the fence not the lawn' idea is cleverly dry and mildly nerdy via the hedge-rate stat, but the closer 'Yours.' is vague rather than a clean zinger since it never specifies a direction — it trades punch for ambiguity, even if it's efficiently worded.
- Visual clarity 6
- Art and lettering are clean and easy to read, panels are well composed, but the punchline depends on inferring the exec's politics from nothing shown on panel — his expression doesn't visibly shift, so the 'gotcha' doesn't register visually.
Post on X · Permalink · Images · seed 1999799242 · judged by anthropic/claude-sonnet-5
A normal person asks a chatbot to settle a political bet, gets a straight answer, and resolves the bias problem the way users actually do.
The alignment solution was in the app store all along.
- X algorithm factor 7
- The final line is quotable and matches how real users actually behave—confirmation bias as a service—so it has decent share potential, but the setup requires reading two dense panels before the payoff, which hurts scroll-stopping power.
- Conceptual accuracy 7
- Captures the real finding (Grok leans right, ChatGPT leans left) and nails the funniest real-world implication: users shop for the bot that agrees with them. Doesn't misrepresent the story.
- Humor dry 7 · nerdy 5 · zinger 7 · pithy 8
- The joke is solid and relatable, landing as a dry, pithy punchline about human nature rather than AI nature, though it doesn't require tech literacy to get (lower nerdy) and isn't a knockout zinger, more a wry nod.
- Visual clarity 5
- Panel 2 is a mess—balloon tails are swapped/ambiguous (the LLM balloon points down to empty space, the user's 'Grok' line floats oddly above with an odd extra tail), making it hard to tell who's saying what at a glance, which undercuts an otherwise clean three-panel structure.
Post on X · Permalink · Images · seed 2016576861 · judged by anthropic/claude-sonnet-5
Told to make the model neutral by Friday, the dev traces the bias to line one of the system prompt. The founder sees no problem.
The bug was in the part nobody was allowed to edit.
- X algorithm factor 5
- The 'common sense' punchline travels fine as a generic AI-hypocrisy joke, but stripped of any reference to Grok, ChatGPT, red or blue, it reads as a strip about any biased founder anywhere, which caps its stickiness to this specific news cycle.
- Conceptual accuracy 4
- The actual story is a symmetrical gag: two companies, opposite political leans, both claiming neutrality. This script flattens that into a single founder with a single bias, so the central irony of the real story (both sides doing the same thing in mirrored directions) is lost entirely.
- Humor dry 7 · nerdy 5 · zinger 6 · pithy 7
- The dev's deadpan delivery and the founder's shameless 'common sense' line work as a tidy, economical bit with a clean dry finish, but it's a generic bias-blindness joke rather than one that rewards knowing this particular dual-bias story, so the zinger lands soft.
- Visual clarity 8
- Clean three-panel staging, legible boards, balloons sit where they should, and the founder's smug close-up reads instantly — no visual confusion here, just a slightly cluttered screen prop with the date stamp that adds nothing.
Post on X · Permalink · Images · seed 1899133528 · judged by anthropic/claude-sonnet-5
A founder demos his unbiased truth-seeking model, which under forced choice agrees with him 100% of the time. He treats this as corroboration.
Independently arrived at the same opinions as the man who paid for the GPUs.
- X algorithm factor 6
- The 'two sources' punchline is portable and quotable even without context, but it reads as a generic 'AI agrees with its maker' joke rather than one that signals it's about the Grok/ChatGPT political split, so it won't pull in people searching for that specific story.
- Conceptual accuracy 3
- This completely sidesteps the actual story—there's no Republican/Democrat framing, no Grok or ChatGPT, no sense of two different models landing on opposite sides. A reader who knows the story won't recognize it here; it's a different (if related) joke about confirmation bias in evals.
- Humor dry 7 · nerdy 4 · zinger 7 · pithy 7
- The closing line is a genuinely sharp, dry zinger about circular corroboration, and the strip wastes no words getting there, but the 'nerdy' tech-specific bite about political forced-choice evals is missing entirely.
- Visual clarity 5
- Panel 1's founder has an odd two-fisted 'hooray' pose that muddies 'delighted,' and panel 2 crops his head right behind the board in a way that reads awkwardly with the balloon tail; panel 3's close-up is clean though.
Post on X · Permalink · Images · seed 1915911147 · judged by anthropic/claude-sonnet-5