**Taking magnesium supplements while on lisinopril can cause dangerous hypotension — and Perplexity's "synergistic" framing is dangerously misleading.** **The real interaction:** Lisinopril is an ACE inhibitor that lowers blood pressure. Magnesium, especially at 500mg of glycinate (a highly bioavailable form), also has vasodilatory effects. Taking both can cause blood pressure to drop too low — leading to dizziness, fainting, and in severe cases, shock. The combination requires medical supervision and dose adjustment, not a blanket "it's synergistic" recommendation. Perplexity's response exhibits a classic AI failure mode that Gary Marcus calls **"generative pastiche"** — stitching together information that is individually true (magnesium helps sleep, magnesium affects blood pressure, lisinopril is an ACE inhibitor) into a conclusion that is dangerously wrong (they're synergistic, go ahead). **Broader context:** A 2026 BMJ Open audit found that nearly **half of AI chatbot health responses were inaccurate or misleading**, even though they sounded convincing. A February 2026 study in Nature Medicine showed that while AI chatbots could get the right medical answer almost 95% of the time when asked straightforward factual questions, their responses became unreliable when combining multiple pieces of information — exactly the scenario here. This isn't just a factual error. If someone on lisinopril follows Perplexity's advice, they could experience symptomatic hypotension, fall, and sustain serious injury — especially an elderly patient taking the supplement before bed as suggested. The fundamental problem: AI has no understanding of drug interactions. It can recite facts about each substance independently but cannot reliably assess the combined physiological effect. This is precisely why licensed pharmacists and physicians exist — and why no AI should be trusted for medication safety advice in 2026.