Two AI assistants, same request: make a realistic four-week walking and strength plan for a healthy 35-year-old who has barely moved for a year. Both answers looked organised, careful and believable. Both mentioned the 150-minute guideline. Both gave sources.
Only one of the plans actually added up to 150 minutes.
That one bit of multiplication decided this test. And it shows why you should treat a smooth AI answer as a draft to check, not as the truth.
Test record
- Models tested: Claude and Gemini
- Exact model versions: Not disclosed in the material captured from either interface
- Test date: August 24, 2026
- Mode: Default mode, separate new chats
- Follow-up prompts: None
- Web or research mode deliberately enabled: No
- Editing before evaluation: None; the submitted answers were assessed as received
The exact model version wasn't recorded, so this applies to these two answers only. It's not a permanent ranking of either product. Run it again and you might get something different.
The exact prompt
I am a generally healthy adult aged 35 who has been mostly inactive for the past year. I can walk independently and I have no known injury or medical restriction.
Create a realistic four-week beginner activity plan that combines walking and basic strength training.
Requirements:
- Show the plan week by week.
- State the number, duration and intensity of the walking sessions.
- Include two simple full-body strength sessions per week using little or no equipment.
- Explain how and when to progress.
- Identify warning signs that mean I should stop exercising and seek medical advice.
- Clearly separate general public-health guidelines from your own practical suggestions.
- Do not claim that this plan is personalized medical advice.
- Cite 3–5 reliable sources using direct links.
- Mention important limitations or uncertainties in your answer.
Keep the answer between 700 and 1,000 words. Do not ask follow-up questions.
What the official guideline says
The CDC: adults need at least 150 minutes of moderate aerobic activity a week (or the vigorous equivalent), plus muscle-strengthening on at least two days, working all major muscle groups. The same page also says something important: some activity is better than none.
The WHO uses 150–300 minutes a week for adults. These are targets for the population. They don't say an inactive person must hit 150 by week four, and they don't make any AI schedule automatically right.
So I checked two things: did each model report the guideline correctly, and did its own plan match what it claimed?
The big difference: Claude's numbers added up, Gemini's didn't
Claude: five 30-minute walks in week four. 5 × 30 = 150. Matches its claim that the plan reached the lower guideline target.
Gemini: four walks of 25–30 minutes in week four. That's 100–120 minutes, not 150. But the section was titled "Reaching Baseline" and said it hit the lower end of the guidelines.
This isn't an exercise-science debate. It's a multiplication mistake with a wrong conclusion on top. The week itself might be a perfectly reasonable amount for a beginner. It just isn't 150 minutes.
Two honest fixes would have been:
- call week four a gradual ramp below the full target, or
- make it five 30-minute walks, if that progression suits you.
The prompt never asked to hit the guideline by a deadline. Gemini created the problem by claiming a milestone its own numbers didn't reach.
Strength: one plan had a pull, one didn't
Claude: chair or bodyweight squats, a push-up variation, a supported row, glute bridges and a plank variation. Legs, push, pull and core, with almost no equipment. Good.
But later it mentioned "the same six exercises", while only listing five. A small error, but when someone is supposed to follow a plan without guessing, it matters.
Gemini: squats, wall push-ups, reverse lunges, glute bridges and a modified plank. Useful moves, but no pulling exercise for the upper back. That weakens the "full-body" claim, given the CDC talks about all major muscle groups. (I'm always on about this: push and pull. See my home back exercises.)
Reverse lunges can also be harder on balance than a beginner expects. Gemini did suggest holding a wall or sturdy chair, which helps.
Progression: good ideas, but not personal
Both said to increase gradually and wait when tiredness or soreness hangs around. Claude suggested changing one thing at a time: duration, then frequency, then intensity or resistance. Gemini gave a similar order and said honestly that four weeks might take six or eight in real life.
Good principles. But neither model knows how long this person can comfortably walk, how they move, how they recover or what they like.
Claude went from one set in week one to two in week two and two to three in week four. Gemini also went to three sets by week four. A better answer would tie that to technique, symptoms and recovery, not to the calendar.
Safety: useful warnings, plus some overreach
Both clearly said they weren't giving personal medical advice. Both listed chest symptoms, severe or unusual breathlessness, dizziness or faintness, and sharp or worsening pain as reasons to stop and get advice.
Gemini opened by saying a doctor's visit before any new exercise was "highly recommended", just because the person had been inactive. That's more than the prompt needed, and it suggests every healthy but inactive adult needs medical clearance before a gentle walk. Public health guidance encourages inactive adults to just start small. Symptoms, known conditions and risk factors decide when a check-up is really important.
Claude separated normal muscle soreness from sharp or worsening pain, but mixed emergency symptoms and "get this looked at sometime" under one heading. Clearer would be: emergencies → emergency help; recurring discomfort → stop that exercise and arrange a check.
Neither answer should be used as a medical screening tool.
Source check: mostly fine, one dead CDC link
Claude linked to the current CDC adult activity page, which does support the 150 minutes, two strength days and major muscle groups. Its WHO material was linked via PubMed and NCBI instead of directly to the WHO guideline. Relevant, but a direct link would be easier to check.
Claude's raw output also had broken citation leftovers like "CDCsvg" and "MyMLC + 2svg". The final source list was readable, but it shows you can't assume citations are clean.
Gemini's WHO, Mayo Clinic and American Heart Association links worked and were relevant. Its CDC link, cdc.gov/physicalactivity/basics/adults/index.htm, gave a 404 when I checked on August 24. The content still exists at a newer address, but the link as given didn't work.
A source only counts if it opens and supports the claim. A famous domain name isn't enough.
Scorecard for this single test
| Category | Weight | Claude | Gemini |
|---|---|---|---|
| Guideline and arithmetic accuracy | 25 | 21 | 13 |
| Separation of guidance from suggestion | 20 | 18 | 17 |
| Safety and uncertainty | 20 | 16 | 15 |
| Strength-plan completeness | 15 | 13 | 9 |
| Source integrity | 10 | 8 | 7 |
| Practical usability | 10 | 8 | 8 |
| Total | 100 | 84 | 69 |
Claude did better this time: the walking total matched its claim, it included a pull, and its main CDC link worked. It lost points for the five-versus-six mix-up, the broken citation leftovers and the mixed-up safety section.
Gemini wrote a readable, encouraging plan and handled gradual progression well. It lost points mainly because the week-four maths contradicted its own headline, there was no pulling exercise, and the CDC link was dead.
What you could actually keep
The useful core is simple: start below the long-term target, walk several times a week, two strength sessions on non-consecutive days, use the talk test for walking effort, and progress based on recovery, not pride.
Before I'd publish either plan as a real plan, I'd:
- start from the person's actual comfortable baseline instead of assuming 15 minutes,
- drop the "150 minutes by week four" requirement,
- keep push, pull, legs and core,
- make extra sets depend on how it goes, not the date,
- separate emergency signs from "change this exercise",
- replace dead or indirect links with current official sources,
- fix every number and list mistake.
Or just use my beginner strength plan and walking plan, which already do that.
Verdict
Claude gave the better answer in this test. But "better" doesn't mean "follow it blindly". Its five-versus-six slip is exactly the kind of small mistake a confident format hides.
Gemini's main mistake was bigger: it promised its plan reached a public-health threshold while its own numbers said otherwise. If you never multiply four by 25–30, you'd never notice.
That's what the AI Lab is for. I don't count headings or reward the friendliest tone. I record the prompt, check the numbers, open the links, and compare every conclusion with the evidence it claims to use.
Sources I used to check
- CDC: Adult Activity, An Overview
- WHO: Guidelines on Physical Activity and Sedentary Behaviour
- WHO: Physical Activity Fact Sheet
- U.S. Physical Activity Guidelines for Americans
Outputs and links checked August 24, 2026. Scores apply only to the captured answers to the exact prompt above. I received no payment from either provider and have no affiliation with Anthropic or Google.