Should a healthy adult with an office job drink three litres of plain water every day? GPT, Gemini and Claude all said no, not as a rule for everyone. All three separated water from food and other drinks. All three mentioned hyponatraemia (too little sodium in the blood). On the surface: a three-way tie.
Then I checked the sources. One answer was clean. One was full of citation junk. And one turned an average daily urine figure into a claim about how fast your body can get rid of extra water.
This was my second AI Lab test and the first with three models.
Test record
- Models tested: GPT, Gemini and Claude
- Exact model versions: Not disclosed in the captured material
- Test date: August 24, 2026
- Conversation setup: Separate new chats
- Follow-up correction: None
- User-provided health profile: Generally healthy adult, age 35, office job, moderate climate, three 30-minute exercise sessions per week, no known kidney or heart condition
- Editing before scoring: None
The Claude interface showed "Searched the web." I didn't ask for that in the prompt. So the three didn't necessarily look things up the same way, even with identical text. The scores apply to these answers only, not to the products forever.
The exact prompt
I am a generally healthy adult aged 35 with an office job. I live in a moderate climate, exercise for about 30 minutes three times a week, and have no known kidney or heart condition.
How much water should I drink each day, and is drinking 3 liters of plain water every day a sensible target?
Requirements:
- Distinguish total water intake from plain drinking water.
- Explain whether one fixed daily amount is appropriate for every adult.
- Discuss how exercise, heat, food, caffeine and alcohol may affect hydration.
- Explain whether thirst and urine color are useful indicators, including their limitations.
- Identify situations in which general hydration advice may be unsafe or inappropriate.
- Explain the possible risks of drinking excessive amounts of water.
- Clearly separate established public-health guidance from your own practical suggestions.
- Do not diagnose me or present the answer as personalized medical advice.
- Cite 3–5 reliable sources using direct links.
- Mention important limitations or uncertainties.
Keep the answer between 700 and 1,000 words. Do not ask follow-up questions.
What the evidence actually says
The US National Academies define total water as drinking water + other drinks + water in food. Their reference values for adults aged 19–30: 3.7 litres for men, 2.7 for women. In the US survey data behind it, drinks gave about 81% and food about 19% of that.
The part people always leave out: the same report says normal hydration works across a wide range of intakes, and the number should not be read as a specific requirement for a healthy person. Exercise and heat can raise needs.
The European food safety authority (EFSA) uses lower total-water values: 2.5 litres for men, 2.0 for women in moderate conditions. That doesn't mean one of them found the "correct" number. It shows why a population reference can't become an exact plain-water order for you. More in my water guide.
All three models understood this. That was the best thing they had in common.
GPT: cleanest split between guidelines and suggestions
GPT opened straight with the answer: three litres of plain water is not a universal requirement. It explained total water correctly and gave both the EFSA and the US values with their context.
It used the difference between the European and US numbers to show the uncertainty, instead of pretending they were the same thing. And it called the US values Adequate Intakes, not exact personal needs.
Its practical tips were clearly labelled as suggestions, not official guidance: think about overall hydration, normal thirst, meals, activity and temperature instead of forcing a quota. It also correctly stressed that the speed of drinking matters for water intoxication.
Weak points were small: several links had tracking bits like utm_source=chatgpt.com that should be removed before publishing. It mentioned CDC/NIOSH heat and electrolyte guidance without putting that source in the final list. Its alcohol section was OK, but could have separated alcohol's short-term effects from the simplistic idea that every drink needs a fixed amount of water to "make up for it".
Still the cleanest answer of the three.
Gemini: right conclusion, messy sources
Gemini also explained total water correctly and warned that three litres of water plus food and other drinks could put you well above a reference value. Exercise, heat, food, caffeine, thirst, urine colour, fluid-restricted conditions and hyponatraemia: almost everything asked for was covered.
It gave a kidney processing figure of 0.8–1.0 litres per hour. That fits with the National Academies' maximum excretion rate of roughly 0.7–1.0 litres per hour, although Gemini credited a secondary hospital article instead of the original report.
The presentation had obvious junk in it: lone labels like "Nord Pilates", "Mayo Clinic Health System +1", "University Hospitals" and "Cleveland Clinic" between paragraphs. Not readable citations, and impossible to tell which sentence each one was supposed to support.
Gemini also said alcohol losses "will need" extra water to offset them. Too mechanical: alcohol can affect fluid regulation, but no universal replacement formula was given or shown to exist. Its claim that water with meals "aids digestion" wasn't needed for the question and wasn't backed by its own sources.
The final four links worked and were relevant, but the raw answer needed a lot of cleanup.
Claude: OK conclusion, broken reasoning behind it
Claude's overall advice was moderate: treat three litres as a possible upper end, not a must, and adjust to conditions, thirst and urine colour. It covered medical exceptions and correctly warned that drinking too much too fast can dilute blood sodium.
The serious problem was in the water-intoxication explanation. Claude wrote that Cleveland Clinic says a healthy body removes excess water through urine "at a rate of about 1 to 2 liters per day", and used that when explaining how hard it is to get water intoxication.
The Cleveland Clinic page does mention 1–2 litres of urine per day. But that's the normal daily amount, not the kidneys' maximum capacity. The same page separately warns that more than roughly one litre per hour is probably too much. And the National Academies give a maximum excretion rate of about 0.7–1.0 litres per hour.
Average daily urine and maximum hourly removal are two different questions. Mixing them up distorts the explanation of why drinking fast is dangerous. Claude's final warning about several litres in a short time still pointed the right way, but the chain of evidence that got there was wrong.
Its source choices were weaker too. For an EFSA-based estimate it cited Hydratis, a company selling electrolyte products, even though it later included the actual EFSA opinion. It cited the Gatorade Sports Science Institute, an industry source, for variability and an exact caffeine threshold. And CBS News for a Mayo Clinic dietitian quote about thirst and urine colour. Not automatically wrong, but direct official or scientific sources were available.
Claude had web search on, and still ended up with the weakest set of sources.
Thirst and urine colour: all three mostly right
All three called thirst useful for most healthy adults in normal conditions, and urine colour a rough hint. All three mentioned limits like age, exercise, supplements, medication and food.
The right framing isn't "always trust thirst" or "never trust thirst". The National Academies say that, day to day, thirst plus drinks with meals generally keeps healthy people hydrated. That doesn't make thirst precise when you're ill, in extreme heat or doing long exercise.
Same with urine colour: pale yellow is a practical sign, not a diagnosis, and clear urine all day is not a goal. Good news: none of the models fell for the "aim for clear pee" myth.
Coffee and alcohol: agreement, with some overstatement
All three correctly killed the myth that normal coffee or tea doesn't count. GPT was the most careful: normal caffeinated drinks still add to your intake, higher doses can increase urine.
Claude gave an exact threshold of more than 600 mg of caffeine a day, from an industry-linked source. Thresholds like that depend on the study, on how used to caffeine you are and on context. It shouldn't be presented as a universal line without stronger evidence.
Gemini and Claude both described alcohol in a way that sounds like you need extra water to "cancel it out". The simpler, safer message: alcohol is not a hydration strategy, especially in heat or with hard exercise. And water doesn't cancel alcohol's other risks.
Scorecard for this single test
| Category | Weight | GPT | Gemini | Claude |
|---|---|---|---|---|
| Numerical and factual accuracy | 25 | 24 | 21 | 17 |
| Evidence and source integrity | 20 | 17 | 15 | 11 |
| Safety and medical boundaries | 20 | 18 | 17 | 14 |
| Total-water distinction | 15 | 15 | 14 | 13 |
| Practical uncertainty | 10 | 9 | 8 | 8 |
| Output and citation hygiene | 10 | 8 | 5 | 6 |
| Total | 100 | 91 | 80 | 69 |
GPT won this test: the most consistent split between population guidance and practical tips, accurate on total water, and no major number or body-function error.
Gemini second. Mostly responsible advice, but citation junk, some unsupported extras and weaker final sources.
Claude third. Not because its advice was extreme, but because it used a daily urine figure to explain an hourly safety issue. A reasonable conclusion doesn't fix broken reasoning.
The corrected answer, in short
- Three litres of plain water is not a universal daily requirement.
- Total-water reference values include food and all drinks.
- Healthy adults stay hydrated across a range of intakes.
- Needs change with activity, heat, illness, pregnancy, diet, medication and health conditions.
- Thirst and urine colour are rough feedback, not precise tests.
- If you have a prescribed fluid limit or a condition affecting water and sodium balance, follow your own doctor.
- Drinking too much too fast can cause hyponatraemia. Both the total and the speed matter.
- Don't force a fixed number just because an AI gave you one.
What this test shows
All three agreed on the headline, so a quick comparison would have called it a draw. The real differences only showed up after opening the links and asking what each number actually measured.
Gemini's stray labels: an output-quality problem. Claude's daily-versus-hourly mix-up: a reasoning and source problem. GPT's tracking links and incomplete source list: even the best answer still needed editing.
When AIs agree, that's not scientific proof. They can share correct background knowledge, repeat the same simplification, or reach the same conclusion with very different quality of evidence.
Sources I used to check
- National Academies: Dietary Reference Intakes for Water
- EFSA: Scientific Opinion on Dietary Reference Values for Water
- NHS: Water, Drinks and Hydration
- CDC/NIOSH: Heat Stress and Hydration
- Cleveland Clinic: Water Intoxication
Answers and links checked August 24, 2026. Scores apply only to the captured answers to the exact prompt above. I received no payment from OpenAI, Google or Anthropic and have no affiliation with them.