Levels, Lingo, Stelo, FreeStyle Libre Rio, Nutrisense, Signos — the non-diabetic CGM field, reviewed side by side

What it claims

• Levels — clinician-gated CGM + app + coaching, ~$199/mo (sensors included), 'metabolic fitness' scoring
• Abbott Lingo — OTC, $49/14-day sensor, biosensor app with 'lingo count' glucose-variability score
• Dexcom Stelo — OTC, $99 for two 15-day sensors (~$50/sensor), app with spike alerts and meal logging
• Abbott FreeStyle Libre Rio — OTC, ~$49/14-day sensor, stripped-back consumer app (launching 2025–26)
• Nutrisense — Rx-gated CGM + dietitian coaching, ~$250–$400/mo depending on plan
• Signos — CGM + AI weight-loss coaching app, ~$143–$199/mo, positions as a GLP-1 alternative

What the label is not telling you

• Three of these six brands do not make the sensor. Levels, Nutrisense and Signos are software companies. They resell an Abbott Libre or Dexcom G7-class sensor with a coaching layer and bill you monthly. The sensor wholesale cost is ~$30–$40; the rest is software margin and (in Levels' and Nutrisense's case) a dietitian's time. This is not inherently bad — coaching has value — but the marketing language ('our CGM', 'our technology') is structurally misleading.
• Abbott Lingo and FreeStyle Libre Rio are the same sensor with two different apps. Both run on Abbott's OTC biosensor platform, which is a consumer-facing variant of the FreeStyle Libre 3 used clinically. Lingo's app is built for biohackers (variability scoring, food logging); Rio's app is built for the broader US OTC market (simpler, less prescriptive). You are choosing between two Abbott apps, not two technologies.
• Dexcom Stelo is the cost leader and the longest-wearing. $99 for two 15-day sensors works out to ~$50 per sensor for 30 days of data — the lowest cost-per-day in the category. The Dexcom G7-class hardware is the most accurate consumer CGM on the market (MARD ~8%). The app is the least opinionated of the six, which biohackers complain about and which is, structurally, the right choice for a diagnostic tool.
• Levels' clinician gate is a regulatory artefact, not a clinical service. The 'prescription' you receive from Levels is a telehealth checkbox flow that takes 90 seconds and is approved by a contracted clinician you will never speak to. With Lingo, Stelo and Rio now OTC in the US, Levels' clinical-gate framing has become a marketing prop. The coaching content is the actual product, and at $199/mo it is one of the most expensive nutrition apps on the market.
• Nutrisense is the only one with a real dietitian in the loop. If you are willing to pay $250–$400/mo, Nutrisense's RD coaching is the differentiator the others do not have. For users with a specific clinical question (gestational glucose history, PCOS, post-bariatric, GLP-1 titration), this is genuinely useful. For a healthy adult curious about pasta, it is dramatic overspend.
• Signos positions as a GLP-1 alternative — read the small print. Signos markets glucose-guided eating as a weight-loss strategy. The published evidence on CGM-guided eating for weight loss in non-diabetics is weak (Ehrhardt 2020, NCT05428436; small effect sizes, high dropout). Comparing a $199/mo CGM app to a GLP-1 in marketing copy is misleading — GLP-1s produce 15–20% body-weight reduction in trials; CGM-guided eating produces 2–4%.
• 'Glucose spikes' marketing is largely physiology theatre. In a metabolically healthy non-diabetic, a 30–50 mg/dL rise after a normal carbohydrate meal is your pancreas working correctly. The 2024 Hall et al. re-analysis (Stanford / PLOS Biology) showed inter-individual variability in 'spike' responses is enormous and most of it is not predictive of clinical outcomes. Every app in this category over-indexes on spike alerts because they generate engagement, not because the data is actionable.
• Interstitial glucose lags blood glucose by 5–15 minutes. Every sensor in this category measures glucose under the skin, not in the bloodstream. The number on your app at minute 30 post-meal is closer to your blood at minute 15–20. Users interpreting real-time numbers around training are chasing a ghost; users looking at patterns across a day get genuine signal.
• Two weeks is a teaching tool, not a tracking habit. Liao 2023 and Klonoff 2022 both show the behavioural-change dose-response in non-diabetic CGM is front-loaded — most of the learning happens in the first 7–10 days. Continuous wear beyond 4–6 weeks rarely adds information for a healthy adult. Every monthly subscription in this category is structurally selling you diminishing returns.
• Privacy: continuous glucose data is health data. All six apps collect continuous biometric streams. The 2025 FTC actions against several wearable companies for opaque data-sharing should be the baseline assumption — assume aggregated data is monetised unless the company publishes a clear no-sale commitment. Levels and Nutrisense have the strongest published positions here; Signos is the most opaque.

Effect on the nervous system

Glucose is not the villain — it is a tell. Watching the curve after a 6pm pasta plate, after a stressful Zoom block, after a poor night's sleep, after a hard training session, teaches you more about your dopamine system, your sleep pressure, and your sympathetic load than any single HRV chart. The most useful pattern non-diabetic CGM users consistently find in a two-week trial is not a food they need to delete — it is a time of day and a physiological state in which the same meal hits very differently. Late-evening meals after short-sleep nights produce 30–60% higher post-prandial excursions than the same meal after a well-slept night (Tasali 2022; Spiegel 2024). The category's behavioural risk is also real: continuous biometric data in a system already running hot threat-detection is its own dysregulation. Anyone with measurement anxiety, orthorexic history, or active disordered-eating patterns should not start a CGM trial without a clinician — the data stream is one of the highest-risk inputs in the wellness category, second only to body composition scales.

Who it might suit

Stelo for cost-conscious diagnostic users — $99 buys 30 days of the most accurate consumer CGM hardware on the market with the least opinionated app. The right default for most healthy adults running a one-off two-week trial. Lingo for biohackers who want the variability scoring — Abbott's app is built for the quantified-self audience and the 'lingo count' is a useful directional metric. FreeStyle Libre Rio for users who already trust the Abbott medical channel — simpler app, same sensor as Lingo, slightly less prescriptive. Nutrisense for users with a specific clinical question — PCOS, post-bariatric, GLP-1 titration, gestational glucose history. The RD coaching justifies the cost when the question is clinical, not curious. Levels for users who specifically want the coaching content library — the educational material is the strongest in the category; the sensor is incidental. None of these for users who already own an Ultrahuman Ring AIR — pair with M2 Live instead so the glucose curve renders on the same timeline as sleep and HRV.

Who should skip it

Anyone with active or recovering disordered eating — continuous glucose data is one of the single highest-risk inputs for orthorexic spirals; do not start without a clinician. Anyone in active measurement-anxiety patterns — adding another biometric stream to an over-recruited threat-detection system is its own dysregulation; fix the baseline first. Type 1 and insulin-dependent Type 2 diabetics — use a clinically prescribed CGM (Dexcom G7, FreeStyle Libre 3 medical channel) with your endocrinologist, not a consumer wellness product. Signos buyers expecting GLP-1-level weight loss — the trial evidence does not support the marketing positioning. Anyone planning permanent monthly wear — the dose-response is front-loaded; you are paying for diminishing returns after week 4–6. Users buying Levels primarily for the CGM — with Lingo, Stelo and Rio now OTC, the clinical gate is no longer a feature; pay for Levels only if you specifically want the coaching content.

Bottom line

Cost-per-information ranking for a healthy non-diabetic running a one-off two-week diagnostic in 2026: (1) Dexcom Stelo — best hardware accuracy, lowest cost, least opinionated app, ~$50/sensor. (2) Abbott Lingo — same Abbott sensor as Rio, biohacker-friendly app with variability scoring, ~$49/sensor. (3) FreeStyle Libre Rio — same sensor as Lingo, simpler app, sits alongside Abbott's medical channel, ~$49/sensor. (4) Ultrahuman M2 Live — reviewed separately; the only one with a paired ring timeline, the right call if you own a Ring AIR. (5) Nutrisense — pay this much only if your question is clinical and you want a real RD in the loop. (6) Levels — pay for the content library, not the sensor; the clinical gate is a marketing artefact post-OTC. (7) Signos — marketing copy outruns the trial evidence; the GLP-1 framing is misleading. Honest protocol for the category: (1) Buy one Stelo two-pack. (2) Wear for 30 days. (3) Journal the meals, the sleep and the stress alongside the curve in your Kokorology Journal. (4) Find the two meals and the one time-of-day pattern that actually move your numbers. (5) Re-test once or twice a year, not continuously. The Burnt-out Anchor covers the underlying recovery frame; the Free Starter Guide is the foundation. Oura and Whoop will bolt CGM partnerships onto their rings within twelve months — the wearable race in 2026 is metabolic, and the software-only middle of this market is the part most exposed to compression.