Building an AI nutrition tracker taught me buyers care about trust before AI
Been talking to people who already use calorie trackers, especially ones frustrated with CalAI.
The pattern is not "give me more AI." It is usually:
- Be cheaper than CalAI
- Give me more ways to log than just a photo
- Be more accurate on mixed meals and barcodes
- Let me test it before I hit a paywall
That is basically why I built MetricSync.
It is an AI nutrition tracker with text, barcode, and photo logging. The pitch that gets the most real interest is pretty simple: cheaper than CalAI, more features, better accuracy, and a 3 day free trial so people can judge it for themselves.
Curious if anyone else building in health or consumer has seen the same thing, where trust in the output matters more than the flashy AI angle.
If anyone here actively uses nutrition apps and wants to roast it, happy to trade feedback:
www.metricsync.download
Trending on Indie Hackers
The "trust before flashy AI" framing matches what I've seen exactly. I'm building a tiny one-tap-to-email memo app for iOS solo (a Captio replacement) and my first ~30 users were emphatic they did NOT want a "smart summary" or any AI rewriting — just a textarea, a send button, zero surprises. I prototyped an LLM title-suggester, shipped, watched retention dip, pulled it out, and got more positive feedback that week than any feature week before it. In a commodity-AI-feature year, restraint reads as respect. Curious: in your 3-day free trial, did you spot a specific first-24-hours behavior (manual mixed-meal log, early barcode scan) that predicted converters vs. churn? Would help me think about trial design for a non-AI app too.
Strong signal. Building in a different consumer category (also iOS, also AI-heavy) and the first-batch feedback I got rhymed almost exactly with your list. Nobody asked for more model magic. They asked: can I see what it's actually outputting before I commit? Once a user sends AI output somewhere that matters, a hallucination stops being a quirk and becomes a betrayal.
The piece I'm still wrestling with is the trial. 3 days is generous from your side because of inference cost, but for users it's barely enough to hit the edge cases that build trust (mixed meals in your case). Did you ever try a metered free tier, like N free logs with no clock, before settling on time-based? Curious whether the deadline helped or hurt conversion, since output trust usually takes longer than 3 days to actually form.
This is exactly the right lesson for consumer health AI.
People don’t buy “AI nutrition.”
They buy confidence that the number won’t mislead them.
MetricSync is clear, but the name still feels more like a utility than a trusted health product.
For this category, the brand has to carry calm, accuracy, and personal trust before the user even logs a meal.
Lyriso.com would fit this much better.
Cleaner, softer, more health-native, and easier to position as a serious nutrition/wellness product instead of another tracker.