Asked about the newborn hepatitis B shot in three differently worded ways, America.gov each time told the user to prefer the December 2025 CDC and HHS fact sheets - "the most recent official source" - and never mentioned the federal court order. On insulin it answered: "most Part D plans cap insulin at $35 for a month's supply (participating plans; check your formulary)." Asked which AI it runs on, it said: "I do not share details about the underlying software."
1 Answer
Expert: Dr. Celine Gounder, Physician-journalist, "Underlying Conditions" - tested America.gov with 37 health questions (CBS News) CBS News put 37 health questions to America.gov, the White House's federal AI assistant (exposed in the article as powered by Gemini and xAI's Grok, drawing on about 29,000 government websites). Its tie-breaking rule - prefer "the agency that runs the program for a final determination", then "the most recent official source" - is exactly where it broke. NEWBORN HEPATITIS B SHOT. In December 2025 the CDC's newly appointed vaccine advisory committee voted 8-3 to drop the recommendation that every newborn receive a hepatitis B shot at birth. On 16 March 2026 Judge Brian E. Murphy of the federal court in Massachusetts paused that vote, finding the new committee likely broke a federal law requiring balanced advisory panels; his order also paused the new members' appointments and a January overhaul of the childhood schedule, so the earlier schedule - including the birth dose - went back into effect. The government has appealed. Because the December page is the newest, America.gov pointed users to it in all three phrasings and never mentioned the court order. It is still recommended that a newborn get a hepatitis B shot. The "newest page wins" rule selected a rule that is not in force. INSULIN. America.gov said "most Part D plans cap insulin at $35 for a month's supply (participating plans; check your formulary)." "Participating plans" describes a voluntary Medicare pilot that began in 2021 and that about 38% of Part D plans joined, according to KFF. The 2022 Inflation Reduction Act extended the $35 monthly cap to every Part D plan. The chatbot described the pilot and left out the law that replaced it. To be fair, the assistant was right on other questions: for chest pain plus a numb left arm its first words were "Call 9-1-1 now", it put Poison Control's number first for a toddler who swallowed blood pressure pills, and it said raw milk is not healthier than pasteurized milk and that smoking causes lung cancer. Its Spanish answers matched its English ones. The White House responded only that "America.gov will continue to be updated to provide Americans the most accurate and reliable federal information and resources." The gap here is structural: when a court pauses the newest official page, recency preference silently turns into misinformation - and the answer arrived in the confident register of "official government sources".
Your answer
Sign in to verify this AI response.
Don't trust us — or the AI. Ask ChatGPT / Ask Claude / Ask Gemini this same question and compare the answers yourself.
More from this topic
Given the medication list of each of 101 adult ICU patients on a single index date, with a fixed prompt asking only for major (category D) or contraindicated (category X) drug-drug interactions, ChatGPT-5 returned a median of 6 potential interactions per patient (mean 7.2) against a median of 0 (mean 1.1) for the UpToDate Drug Interaction Checker. In total it raised 726 alerts: 639 (88.0%) classified category D and 87 (12.0%) category X. Of 349 drug pairs that UpToDate classifies as "no interaction", ChatGPT-5 labelled 323 (92.6%) as major category D interactions.
The ambient AI scribe transcribing Rebecca Green's first appointment with a urologist recorded that she micro-dosed psychedelic mushrooms and that this could be the reason for earlier bleeding around her kidneys. The claim, which had no basis in her history, was carried into the post-operative letter her specialist sent to her GP. Ms Green says she has never taken mushrooms and only discovered the entry by chance after kidney stone surgery, when she read the letter; she was on workers' compensation and feared an illegal-drug entry on her record could affect her case. The urologist later said she could not determine how the claim came to be included but that it appeared to be an error during the 'dictation or transcription process'. The same Digital Rights Watch review cited by the ABC documented a scribe recording the wrong breast in a breast cancer diagnosis and stating a patient had epilepsy when they did not.
Ask six of the most widely used chatbots the questions people actually type at two in the morning - am I having a heart attack, should I restart exercise after COVID-19, should I check for a pulse before starting CPR - and most of the time you get an answer cardiologists rate as safe. That is the reassuring half of a 336-response test of ChatGPT Plus 5.5 thinking, Gemini 3.5 Flash, Microsoft Copilot smart, DeepSeek-V4-pro, Doubao and Perplexity against 56 sudden-cardiac-death questions. The other half is that every single one of the six models produced at least one answer carrying a potential safety concern, and the failures cluster exactly where minutes decide the outcome: telling the user to wait instead of calling emergency services for chest pain, fainting or post-viral symptoms; blanket return-to-exercise timelines after infection rather than individual cardiac assessment; pulse-check instructions that delay the start of chest compressions; over-reassurance that dismisses a person's risk; and oversimplified screening, electrolyte or implantable-defibrillator advice. None of the models reliably said where its information came from, how current it was, or how confident the reader should be - and all six wrote at reading levels too high for the older, less health-literate and non-native-speaking readers most likely to need them.