AI Endurance Review 2026: Price, Digital Twin, DFA Alpha 1, Is It Worth It?

AI Endurance makes one large promise. It says it trains “a Digital Twin of you: a neural network that learns from your own workouts how you respond to training, then builds the plan that maximizes your predicted performance on race day.”¹

That is a bigger claim than most training apps make. Most give you a plan. AI Endurance says it can predict how you will respond to one.

So this review asks three questions. What does AI Endurance cost in 2026? What do the digital twin and its DFA alpha 1 threshold detection actually do, and how much of that is independently verified? And who should pay for it?

We make SensAI, an AI coaching app for iPhone whose coach is a large language model. We are not neutral, so we say where AI Endurance is the better choice. Every AI Endurance price, feature and rating below was checked on its own pages (aiendurance.com), its app store listings and its official forum on 15 October 2026, and every competitor price on that competitor’s own pricing page the same day. Every study was checked against its PubMed record.

AI Endurance review: the short answer

AI Endurance is worth trying if you are a self-coached runner, cyclist or triathlete with a long GPS or power-meter history and a chest strap. Its per-athlete neural network plans around how you respond to training. It costs $20 a month or $200 a year after a 14-day free trial. It fits beginners, strength-first athletes and swim-focused triathletes less well.²¹³

	AI Endurance (October 2026)
What it is	A per-athlete neural network (“Digital Twin”) that plans your training, plus an LLM-based chat coach14
Sports	Running, cycling, triathlon and duathlon3
Price (US)	$20/month or $200/year; one plan, everything included2
Free trial	14 days, “no payment info required”2
Platforms	Web, iOS and Android156
Data it needs	100+ GPS runs or power rides for its “about 5%” prediction claim; a chest strap for DFA alpha 13
Apple Health	Not a listed integration34
US App Store rating	4.2 from 24 ratings5
Made by	AI Endurance Inc., Hamilton, Ontario, Canada57

Here is what that means in practice. AI Endurance is strongest when it has a lot of your own data to learn from. If your history is thin, or your training is mostly in the gym or the pool, much of what you pay for has little to work with.

What is AI Endurance and how does it work?

AI Endurance is an adaptive training app for running, cycling and triathlon, made by AI Endurance Inc. in Hamilton, Ontario. Its founder, Markus Rummel, is a particle physicist, and the company has been online since at least May 2020. It runs on web, iOS and Android under a single subscription.⁸⁵⁹¹

Rummel describes himself as a triathlete (2:06 for an Olympic distance, 4:34 for a half-Ironman) and a physicist “who worked at world leading institutions such as University of Oxford.”⁸ His co-founder, Chris St. Jean, is an ultramarathon runner based in Colorado Springs.⁸

The origin story is a physicist’s one. “Ever since I started working with machine learning methods in my physics research, I’ve been wondering how to leverage these methods to come up with a more data-driven, more predictive and optimized training routine,” Rummel writes on the company’s About page.⁸ The page puts it more bluntly: “AI Endurance exists because modern endurance training is a data problem, and we are scientists.”⁸

Think of the product as three layers stacked on your workout history.

Layer	What it does	Source
Digital twin	A neural network, retrained every 24 hours, that predicts your performance and picks the plan with the best predicted race day	AI Endurance FAQ and product page34
Recovery model	Fits your HRV, resting heart rate, DFA alpha 1, training load, morning check-in and soreness into recovery scores for “cardio and muscular”	Product page4
Chat coach	An LLM-based assistant that can analyse activities “second by second,” create and reschedule workouts, and adjust zones	Product page4

It also computes “daily calorie and macronutrient targets” from your training and builds meal plans to match.⁴ All of it sits in one subscription: “One subscription, everything included.”²

Its App Store subtitle calls it an “Adaptive AI endurance coach.”⁵ The iOS app first appeared on 16 October 2024, and version 4.0.0, a redesign, shipped on 15 July 2026.⁵ On Google Play it shows 5K+ downloads.⁶ The interface comes in nine languages, including English, German, Spanish and Portuguese.¹⁰

One distinction matters before we go further. The digital twin is classic machine learning: a numerical model that predicts performance. The chat is a large language model. They are different tools, and our guide to AI coaching for endurance athletes explains why that difference matters when you choose an app.

How much does AI Endurance cost in 2026?

AI Endurance costs $20 a month or $200 a year in US dollars, on the web and on the US App Store, after a 14-day free trial that needs no payment details. There is one plan with everything included. App Store prices vary by country, and coaches pay $15 per athlete per month, or $12 from 10 athletes.²⁵¹¹

Channel	Monthly	Yearly	Source
Web (USD)	$20	$200 (“Save $40!”)	AI Endurance pricing2
App Store, US	$20.00	$200.00	App Store (US5)
App Store, UK	£15.00	£150.00	App Store (UK12)
App Store, Germany	18,00 €	179,99 €	App Store (DE12)
App Store, Canada	CA$25.00	CA$249.99	App Store (CA12)
App Store, Australia	A$30.00	A$300.00	App Store (AU12)
Coach portal	$15 per athlete (1–9 athletes); $12 per athlete (10+)	—	AI Endurance for coaches11

Prices checked 15 October 2026. The web pricing page localises by country: opened from Brazil, it showed R$69.90 a month and R$699.90 a year.²

The yearly plan works out to about $16.67 a month. AI Endurance’s own coaches page confirms the athlete price: “the direct athlete subscription is $20 USD per month.”¹¹ For coaches, “Your first athlete is free,” and clubs and teams of 25+ get a “Let’s talk.”¹¹

The trial is unusually low-friction: “Try everything free for 14 days, no payment info required.”² That matters more than it sounds, because AI Endurance improves with data. Connect your full history on day one, not day ten.

A warning about older AI Endurance pricing figures. Triathlete’s roundup, last updated 11 March 2026, still lists “$12.99/month; $129.99/year.”¹³ That is not what the US checkout shows today. Budget on $20 and $200.

How does the AI Endurance digital twin build your plan?

Every 24 hours, AI Endurance retrains a neural network on your workouts. It then searches possible training futures (volume, intensity and zone distribution) for the plan with the best predicted race-day performance. With too little data, it borrows from similar athletes. In a 2021 explanation, the company said it restricts plan-finding to within one standard deviation of your own history.³⁴¹⁴

Picture a flight simulator built from your own logbook. Each new workout adds a page. The simulator tries many possible training blocks and shows you the one that lands best on race day.

The company’s own wording: “Your Digital Twin searches the possible training futures, volume, intensity, zone distribution, and locks onto the plan that maximizes your predicted performance on the day of your goal event.”⁴ Before you start, it shows “the improvement it predicts from following your plan versus repeating your own training.”⁴

The model refreshes daily, “taking into account each new activity you have uploaded via Garmin, Suunto, Coros, Polar, Wahoo, Hammerhead, Intervals.icu, Strava, Zwift, Rouvy or TrainerDay.”³ If you are new, “we will predict your performance based on the accumulated data of athletes in our database similar to you.”³

The limit most reviews skip: how far it will plan past your history

This is the most important design detail for a buyer, and it comes from AI Endurance itself. In a 2021 post, the company explained how plan-finding works:

“we restrict the possible input data up to one standard deviation away from the mean of the athlete’s historical data … if an athlete has never spent any time in a certain zone, the plan finding can’t reliably explore this zone as there is no input data to learn from and hence we discard the zone.”¹⁴

That is honest, and it is good statistics. A model should not extrapolate far past its data. But it has a practical consequence. In that 2021 design, AI Endurance optimised inside your past. If you have never done threshold work, the 2021 design discarded that zone. For volume well beyond your history, today’s FAQ says a “secondary layer of training for digital twin” switches on automatically to accommodate the request.³

The rationale for a per-athlete model is sound. A review in Sports Medicine found that people show “a wide range of responses” to the same standardised training.¹⁵ A model fitted to you can, in principle, learn which of those responders you are. The catch is that it learns mostly from what you have already done.

How it sets your zones and tests your fitness

AI Endurance sets zones in two passes: “first, we try to find enough DFA a₁ threshold data to set your zones. In a second attempt, we are using the distribution of your historical power/pace data.”³ It doesn’t add test workouts by default: “We deliberately do not include test workouts.”³ Your normal training is the test, though you can ask the chat to add one or request regular tests in the Plan System Prompt.³

You can also steer the plan in words. Since November 2025, the Plan System Prompt “lets you directly tell our LLM-based AI agent how you want your training to evolve — in your own words.”¹⁶

What kind of training does it prescribe?

In its 2021 simulation, AI Endurance reported that “For 89% of athletes an AI optimized polarized training composition is most beneficial while for 11% an AI optimized threshold training composition is optimal.”¹⁴ That echoes a long line of research. Stephen Seiler, PhD, of the University of Agder, Norway, observed that well-trained endurance athletes perform about 80% of their training sessions at low intensity, and wrote: “Endurance athletes appear to self-organize toward a high-volume training approach with careful application of high-intensity training incorporated throughout the training cycle.”¹⁷

But newer evidence is less tidy. A 2025 network meta-analysis of individual data from 348 athletes in 13 studies found no difference between polarized and pyramidal training for VO2max or time-trial performance, and suggested recreational athletes may do better with pyramidal training.¹⁸ We cover that debate in polarized vs threshold vs pyramidal training. The 89% figure is a simulation on digital twins, not a measured outcome.

How accurate is AI Endurance?

Most of AI Endurance’s accuracy claims are the company’s own. Its “about 5%” race-prediction figure needs at least 100 GPS runs or power rides, and we found no published validation. The 11.1% gains figure comes from a 2021 simulation. The only peer-reviewed paper on its data is co-authored by the company.³¹⁴⁷

Here is every major claim, graded by the evidence behind it.

Claim	Who says it	What the evidence actually is
“We can predict your future performance at about 5%”	AI Endurance FAQ3	Company claim. Requires “at least 100 runs with GPS data or rides with power data.” No published validation found.
“11.1% performance gains in FTP and 5k pace”	AI Endurance blog, April 202114	Simulated: digital-twin outputs for 126 athletes over an 8-week plan, not measured results. The company notes “a certain bias is introduced via the choice of data processing and machine learning model.” Not peer-reviewed.
DFA alpha 1 “at high, Kubios-like accuracy”	AI Endurance blog19	Based on one blog analysis by Bruce Rogers, MD, in July 2021: a single ~2-hour session using “data given to me by the AI Endurance team.” Not peer-reviewed.20
Cycling power and DFA alpha 1 can estimate intensity zones	Sensors, 20247	Peer-reviewed, but co-authored by AI Endurance staff, using 3,123 workouts from 21 male users. Its own abstract: “Validation data of the performance level and the up-to-date relationship are still pending.”
Durability from everyday rides	AI Endurance blog, October 202621	Company research on 105 amateur cyclists and triathletes and 4,850 rides; a paper is announced for the Journal of Science and Cycling. Not yet peer-reviewed; “The cohort is almost entirely male amateur athletes.”
Threshold detection rates	AI Endurance blog19	Company data: aerobic cluster thresholds in about 40% of high-quality HRV activities, aerobic ramps in about 15%, anaerobic clusters in about 15%, anaerobic ramps in about 5%.

Read the table this way. The engine may well work. But a buyer should treat its headline numbers as the maker’s own estimates, not as independent results. Its terms of use say as much: “Our performance predictions are estimations and come together with a prediction error on top of uncontrolled variables such as weather, form on the day, etc.”²²

How good is AI Endurance’s race prediction compared with the research?

The independent research gives a useful yardstick for “about 5%”:

• Formulas vary wildly. A 2019 systematic review of 114 marathon prediction equations from 36 studies found standard errors from 0.27 to 27.4 minutes among the 19 equations that reported one. Its advice: “Runners should therefore be wary of relying on a single equation to predict their performance.”²³
• Machine learning can do better with good inputs. A 2023 study of 820 French marathoners, using 10-km time, BMI, age and sex, found a mean absolute error of 2.4% for a k-nearest-neighbours model and 5.6% for a neural network.²⁴
• Training logs alone are harder. Critical speed computed from the raw training data of more than 25,000 recreational runners predicted marathon time with an error of about 8%.²⁵

So “about 5%” from everyday training data is plausible. It sits inside the range the research reports. That is a long way from proof, but it is not a fantasy number. It matters more now that AI Endurance lets runners pick Critical Speed as an optimisation target (added 8 October 2026).²⁶

Does adapting training to your body beat a fixed plan?

That is the bet behind every adaptive app, and the evidence says: a little. A 2020 meta-analysis found HRV-guided training produced a “small (ES = 0.402) but positive effect” on VO2max.²⁷ A 2021 meta-analysis was more cautious: “if HRV-guided training is superior to predefined training for producing group-level improvements in fitness and performance, current data suggest it is only by a small margin.”²⁸

What this means for you: on current evidence, an adaptive plan is a modest edge, not a shortcut.

Does AI Endurance’s DFA alpha 1 work, and what do you need?

In lab studies, DFA alpha 1 near 0.75 marks the aerobic threshold and near 0.5 the anaerobic threshold. In AI Endurance it needs a chest strap over Bluetooth and clean data with under 5% artifacts. It also needs a fresh body: fatigue shifts the reading. Coros, Wahoo, Zwift and Strava data carry no activity HRV.^2930331

DFA alpha 1 is a number for the fractal pattern in the gaps between your heartbeats. As exercise gets harder, it falls. AI Endurance’s own summary of the research: “DFA alpha 1 drops below 0.75 at the aerobic threshold and below 0.5 at the anaerobic threshold.”¹⁹

Thomas Gronwald, PhD, of MSH Medical School Hamburg, has helped build that research base, and he co-authored AI Endurance’s 2024 paper.⁷ In a 2020 paper, Gronwald and colleagues wrote: “DFA-alpha1 may be taken as an indicator for exercise prescription and intensity distribution monitoring in endurance-type sports.”³²

What the research shows

Study	Who	What it found
Rogers et al., 2020	15 participants, treadmill ramp	Heart rate at the gas-exchange aerobic threshold (VT1) averaged 152 bpm; at DFA alpha 1 = 0.75 it averaged 154 bpm (r = 0.97).29
Rogers et al., 2021	15 participants	Heart rate at DFA alpha 1 = 0.5 vs the anaerobic threshold (VT2): r = 0.78; 171 vs 174 bpm on average.30
Schaffarczyk et al., 2023	26 women, cycling	Agreement for heart rate was good (ICC 0.87 at VT1, 0.90 at VT2), but at VT1 the mean bias was −4.7 bpm with limits of agreement from +10.6 to −20.0 bpm.33
Van Hooren et al., 2023	14 runners	Fresh: −0.9 bpm average bias. After fatigue: −12.3 bpm, with limits of agreement from −30.4 to +5.9 bpm.31
Rogers et al., 2021	17 participants	1–3% artifact gave a small but discernible bias, and 6% gave proportional bias of up to 19%, though the mean threshold stayed within 1 bpm; a Polar H7 averaged −4 bpm vs ECG.34
Schaffarczyk et al., 2022	25 recreational athletes	Polar H10 DFA alpha 1 agreed well at low intensity but limits of agreement widened to 58.1 to −40.9% at high intensity.35
Rogers and Gronwald, 2022	Review	“minimal bias” from artifact correction below 3%, and “negligible change” in the threshold even at 6%.36

Two lessons come out of that table. First, the method is real: on average, the 0.75 threshold lines up closely with lab-measured aerobic threshold. Second, individuals vary, and fatigue matters. Van Hooren and colleagues concluded that DFA alpha 1 “is only able to accurately demarcate exercise intensity thresholds in a non-fatigued state, but also opens opportunities for fatigue-based training prescription.”³¹

For an app that reads thresholds from everyday workouts, that matters. A threshold taken from the back half of a hard week may not match your true one.

Is AI Endurance’s own DFA alpha 1 calculation accurate?

The best evidence is one blog post. In July 2021, Bruce Rogers, MD, a board-certified endocrinologist affiliated with the University of Central Florida College of Medicine and lead author of the threshold papers above, compared AI Endurance’s numbers with Kubios HRV software.²⁹²⁰ He found “The correlation of AI endurance is excellent ( r of .96 and a mean difference of only about 5% at low DFA a1)” and concluded: “AI Endurance DFA a1 calculation appears to closely match that of Kubios premium.”²⁰

That is a real endorsement from the lead author of the DFA alpha 1 threshold studies. It is also one ~2-hour session analysed quantitatively, five years old, using data supplied by the company, and the app has been updated many times since.²⁰ We found no independent peer-reviewed validation of AI Endurance’s DFA alpha 1 calculation.

For contrast, here is what peer-reviewed validation of an app looks like. A 2026 study tested the open-source Fatmaxxer app against Kubios in 23 people and found thresholds within 1 bpm and correlations above 0.96.³⁷ That study did not test AI Endurance.

What hardware do you need for DFA alpha 1 in AI Endurance?

You need a chest strap. AI Endurance is explicit: “Optical wrist-based heart rate monitors do not record in-activity HRV/DFA alpha 1 for Garmin and Suunto devices. You must use a compatible chest strap (such as Polar H10).”³ Pair it over Bluetooth, because “Recording with ANT+ generally leads to significantly less quality HRV data.”³

Data quality has hard cutoffs. Readiness uses activities with under 10% artifact corruption; AI predictions require “less than 5% corruption.”³

Device or app	In-workout HRV reaches AI Endurance?	Live DFA alpha 1
Garmin + chest strap	Yes	alphaHRV Connect IQ app3
Suunto + chest strap	Yes3	—
Hammerhead Karoo	—	AI Endurance Karoo extension3
Apple Watch	Through Watchletic paired to a supported strap such as Polar H10 (since May 202638)	Watchletic 3.1.0 or later3
Polar	No: “does not include it in the FIT files shared with third-party apps”3	—
Coros, Wahoo	No: “do not record activity HRV data”3	—
Zwift, Rouvy, TrainerDay	No3	—
Strava	No: “Strava does not pass on activity HRV data”3	—

That table is the most practical part of this review. If your watch is a Coros, or your rides reach AI Endurance only through Strava, the DFA alpha 1 features have nothing to read. Our Zone 2 and LT1 calibration guide covers other ways to find your aerobic threshold.

What’s new in AI Endurance in 2026?

AI Endurance’s biggest 2026 change is its LLM chat. It was rebuilt on Gemini 3 in February, gained memory across conversations, learned to replan your week, and moved to Gemini 3.7 Flash in September, with Claude Sonnet 5.5 as an option. A ChatGPT app, durability analysis, indoor-trainer integrations and Critical Speed followed.³⁹²⁶³⁸

Date	What changed
11 Nov 2025	Plan System Prompt: tell the LLM-based agent how your training should evolve16
12 Feb 2026	Agentic chat revamped: “The chat now uses Gemini 3 (Flash and Pro available)”39
24 Feb 2026	Bulk .FIT upload40
26 Feb 2026	“The chat now has cross-conversation memory.”39
1 Apr 2026	Weekly replan through the chat39
7 Apr 2026	Coaching portal for coaches40
25 May 2026	Watchletic computes DFA alpha 1 from a strap paired to Apple Watch38
1 Jun 2026	Zwift, Rouvy, Hammerhead and TrainerDay integrations38
16 Jun 2026	ChatGPT app live “After a 6 month review process”39
6–15 Jul 2026	New design and dark mode; iOS 4.0.0 on 15 July405
8 Jul 2026	Durability analysis on the Data page26
17 Sep 2026	“Gemini 3.7 Flash is now the default chat model, replacing 3.0 Flash.”39
29 Sep 2026	Subscribers can pick Claude Sonnet 5.5 “as an alternative to Gemini 3.1 Pro.”39
8 Oct 2026	Critical Speed as an optimisation target; run and bike goals for triathletes2640

The chat is no longer a help widget. It “can act: analyzing activities second by second, creating and rescheduling workouts, adjusting your zones.”⁴ There is also an MCP server so ChatGPT and Claude can work with your AI Endurance data.⁴¹

Durability is the most interesting science addition. Sports scientists define durability as “the time of onset and magnitude of deterioration in physiological-profiling characteristics over time during prolonged exercise.”⁴² AI Endurance measures it as the loss of performance after a given amount of work, in kilojoules for cycling and kilometres for running.²⁶ In its own data from 4,850 rides, after 15 kJ/kg of work, heart rate ran 10 bpm higher and DFA alpha 1 0.17 lower (median across athletes).²¹ Its heart-rate drift measure was more repeatable than its power-based one (correlation 0.72 vs 0.37).²¹ “For amateur athletes it is not just the convenient way to measure durability; it is the one that works,” Rummel wrote.²¹ Independent lab work points the same way: in a 2025 study of 10 participants riding to exhaustion at a constant high power, heart rate, DFA alpha 1 and breathing rate drifted as the ride went on, and the drift was repeatable.⁴³

One chat limitation to know

The chat remembers your conversations, but the plan does not. AI Endurance’s FAQ: “The chat assistant keeps its own memory to follow your conversations, but it does not write your preferences into your plan settings.” And: “Anything you only discussed in the chat, but never saved to your Plan System Prompt, is not carried over.”³

In plain terms: if you tell the chat you can’t run on Tuesdays, also put it in the Plan System Prompt.

Earlier reviews of the chat were cool. Triathlete’s hands-on roundup (published September 2025, last updated 11 March 2026) called the Gemini chatbot “far less capable and slower than the Maxiom platform’s chatbot.”¹³ That test predates the February 2026 chat rebuild and the September model changes, so treat it as a dated snapshot.

Does AI Endurance work with Garmin, Apple Watch, Coros and Strava?

AI Endurance works with Garmin, Coros and Strava, and with Apple Watch only through the third-party app Watchletic. Garmin imports and exports, and logs in-workout HRV with a chest strap. Coros, Strava, Zwift and others import but send no activity HRV. Apple Health is not a listed integration, and TrainingPeaks is export-only, plus overnight HR and HRV.³⁴

Platform	Import	Export workouts	Notes
Garmin	Yes	Yes (Garmin Upload)	Activity HRV with a chest strap; “Garmin does not pass on third party activities to AI Endurance”; full history via bulk upload3
Apple Watch	Through Watchletic	Through Watchletic	“simply connect Watchletic for uploading your activities to AI Endurance and executing your workouts on your Apple Watch”3
Apple Health	Not listed	Not listed	Overnight Apple Watch HRV only “indirectly” through Intervals.icu3
Coros	Yes	Yes	No activity HRV3
Suunto	Yes	Yes	Activity HRV with a chest strap3
Polar	Yes	—	In-workout HRV not in shared FIT files3
Wahoo, Hammerhead	Yes	Yes	Wahoo: no activity HRV3
Strava	Yes	—	No HRV passed on3
Zwift, Rouvy, TrainerDay	Yes	Yes	Since 1 June 202638
TrainingPeaks	Overnight HR/HRV only	Yes	“We do not import activities from TrainingPeaks”3
Intervals.icu	Yes	Yes	—34
Nolio	—	Yes	—4
Stryd, Oura, WHOOP	Yes	—	Running power; recovery data45

AI Endurance exports one week of workouts ahead, then adds the next day every 24 hours.³

The Apple gap is worth stating plainly. If your Apple Watch is your main training device, AI Endurance reaches it through one third-party app, and Apple Health itself is not on the integration list.

SensAI sits on the other side of that fence. It reads Apple Health only, with no vendor logins. Your Garmin, COROS, Polar or Strava workouts reach SensAI because those apps write them into Apple Health first.⁴⁴ The trade-off runs the other way too: HRV reaches SensAI only from an Apple Watch, vendor scores such as Garmin’s Body Battery never cross, and SensAI syncs roughly the last 30 days of Apple Health workouts.⁴⁴ If you want the details for your device, see how to connect a wearable to SensAI through Apple Health.

Is AI Endurance good for running, cycling and triathlon?

AI Endurance covers running, cycling, triathlon and duathlon, with goals from 5K to Ironman and from FTP to Critical Speed. Cycling plans need power-meter data. Swim sessions come from a coach-built workout library, not the AI. Strength exists only as free-form custom workouts, and the app does not detect injuries.³⁴¹²⁶

Sport	Goals you can set	What to know
Running	5k, 10k, (Half-)Marathon, Ultra Endurance Pace, Critical Speed, Critical Running Power, Ultra Endurance Power3	Strongest with 100+ GPS runs3
Cycling	FTP, Critical Cycling Power, Ultra Endurance Power3	“Power is the only metric in cycling that we found to be predictive”3
Triathlon	Sprint, Olympic, (Half-)Ironman3	Swim workouts “taken from a workout database created by an experienced coach”3
Duathlon	Yes3	—

You can add B and C races before your A race.³ Swim zones come from a critical swim speed estimate, but AI Endurance says it does not generate AI swim workouts “because it is harder to provide our AI with clean data to predict swimming performance.”³

Cyclists without a power meter should read the fine print: “We do need some activities with at least power and heart rate data.”³

Strength is the clearest gap. AI Endurance’s chat and MCP tools can “Create custom strength training, yoga, hiking, cross-country skiing, and other non-endurance activities with free-form workout descriptions.”⁴¹ That is a free-form session, not strength programming inside the plan engine. And on injuries: “Our AI does not detect injuries in any form.”³

That is where SensAI’s lane begins. Its Coach sees your swims, rides and runs next to your planned strength sessions, and “A heavy cycling block shows up as load that strength volume has to fit around.”⁴⁴ It is a strength-around-endurance tool, not a replacement for an endurance plan engine like AI Endurance.

Runners comparing dedicated running apps can also look at our roundup of the best AI marathon training apps.

AI Endurance reviews: what do users say?

AI Endurance has 4.2 stars from 24 ratings on the US App Store, 4.1 from 17 in Germany and 4.4 from 7 in the UK. Google Play shows no star rating, and we found no Trustpilot profile. Reviewers praise the engine, but some say the “adaptive” part still misses on individual workouts.⁵¹²⁶

Store	Rating	Ratings
App Store, US	4.2	245
App Store, Germany	4.1	1712
App Store, UK	4.4	712
App Store, Canada	4.8	912
App Store, Australia	3.7	312
Google Play	No rating shown	5K+ downloads6

These are small samples. With 24 ratings, the written reviews tell you more than the stars.

The written reviews split along a clear line. One App Store reviewer wrote: “I’ve spent plenty of time on TrainerRoad and Zwift, but AIEndurance is on a different level.”⁵ Another, under the title “Good concept, but the “adaptive” part isn’t there yet,” wrote: “I hadn’t done VO2max work in a long time and the app’s first prescription was 6x3min at VO2max, no ramp-up, no adjustment for the layoff.”⁵ A third asked for the app to “more fully leverage AI for workout instructions and feedback.”⁵

That second review is one person’s experience, but it suggests a useful trial test: if you are coming back from a break, check what the plan prescribes in week one.

AI Endurance vs TrainerRoad, Athletica, TrainingPeaks and Runna: how do they compare?

On price, AI Endurance ($200 a year) sits close to TrainerRoad ($209.99) and Athletica ($189), above TrainingPeaks Premium ($134.99) and Runna ($119.99). Its edge is a per-athlete prediction model across running, cycling and triathlon. TrainerRoad is cycling-first, Runna is running-only, and Athletica adds rowing and HYROX.^245464748

App	Focus	Monthly (US)	Yearly (US)	Free trial
AI Endurance	Run, bike, triathlon, duathlon3	$202	$2002	14 days, no payment info2
TrainerRoad	“The #1 Cycling App for Getting Faster”45	$21.9945	$209.9945	30-day money-back guarantee45
Athletica	Triathlon, duathlon, running, cycling, rowing, HYROX46	$19.9046	$189.00 (or $99.00 for 6 months46)	2 weeks, no credit card46
TrainingPeaks Premium	Training platform; human-coach plans available47	$19.9547	$134.9947	14 days47
Runna	Running48	$19.9948	$119.9948	7 days48
JOIN	Cycling49	€16,9949	€119,9949	7 days49
Humango	AI training planner50	$16.99 Essential; $28.99 Premium50	$155.99; $264.9950	Yes; length not stated50

Prices from each company’s pricing page on 15 October 2026. JOIN is priced in euros. Humango’s App Store listing shows additional tiers.⁵⁰

AI Endurance vs TrainerRoad

TrainerRoad costs $9.99 more a year and is built for cycling, with “AI FTP Detection · Fatigue Prevention · Machine Learning Optimization.”⁴⁵ AI Endurance covers running and triathlon too, and adds DFA alpha 1. If you only ride, TrainerRoad’s 30-day money-back guarantee lets you test it; if you run or race triathlon, AI Endurance covers more of your week.⁴⁵²

AI Endurance vs Athletica

Athletica is $11 cheaper a year, with a 2-week trial and no credit card.⁴⁶ Its AI Coach draws on a “curated sport science knowledge base,” but Athletica states that “The AI Coach does not automatically modify your training.”⁴⁶ AI Endurance’s chat, by contrast, can create and reschedule workouts directly.⁴ If you are weighing the two, our Athletica alternatives guide goes deeper.

TrainingPeaks is the option if you want a human. Its coached tiers run $149, $229 and $359 a month.⁴⁷ Already decided to leave AI Endurance? Our AI Endurance alternatives guide is written for switchers.

Where SensAI fits — and where it doesn’t

SensAI is not an AI Endurance replacement, so we left it out of the price table.

If DFA alpha 1 thresholds, race-time prediction or a dedicated endurance plan engine are what you want, AI Endurance, or one of the apps above, is the better fit. Those are not SensAI’s lane.

SensAI’s lane is different. It is an iPhone app whose Coach is an LLM, and it reads Apple Health only. Your runs, rides and swims from Garmin, COROS, Polar or Strava appear next to your planned strength sessions, and Coach fits strength volume around the endurance load it sees.⁴⁴ “Sleep and overnight heart rate from your device shape how hard Coach asks you to push on a given day.”⁴⁴ If you are a hybrid athlete who lifts as well as runs or rides, that is the gap SensAI fills, alongside an endurance app or without one.

Where SensAI falls short for an AI Endurance user: it never logs into Garmin or Strava, HRV only arrives from an Apple Watch, and history starts from roughly the last 30 days in Apple Health.⁴⁴

Is AI Endurance worth it in 2026?

AI Endurance is worth $200 a year if you self-coach running, cycling or triathlon, have a deep GPS or power history, and wear a chest strap. It is worth less if your data is thin, your rides lack power, or you mostly swim, lift or train with an Apple Watch. The 14-day free trial will tell you.²³

AI Endurance is a good fit if you:

• Are self-coached, train for a running, cycling, triathlon or duathlon goal, and want a plan built from your own data.³
• Have 100 or more GPS runs or power rides for the model to learn from.³
• Use Garmin or Suunto with a chest strap over Bluetooth, and want DFA alpha 1 thresholds.³
• Want an LLM chat that can change workouts, plus a ChatGPT app and MCP access.⁴⁴¹
• Coach athletes and want a portal at $15 or $12 per athlete.¹¹

Look elsewhere if you:

• Are new to structured training. The twin falls back on similar athletes and leans on your own history.³¹⁴
• Ride without a power meter, or train mostly on Coros, Wahoo or through Strava, and want HRV-based features.³
• Want AI-generated swim sessions or structured strength programming.³⁴¹
• Train mainly with an Apple Watch and expect Apple Health integration.³
• Lift as much as you run or ride. A coach that plans strength around endurance load, such as SensAI, fits that better.⁴⁴
• Want a human coach. TrainingPeaks’ coached tiers start at $149 a month.⁴⁷

Frequently asked questions

Quick answers: AI Endurance costs $20 a month or $200 a year after a 14-day free trial that needs no payment details. DFA alpha 1 needs a chest strap, Apple Watch works only through Watchletic, and its “about 5%” race prediction is a company claim without published validation. It suits self-coached athletes with long GPS or power histories.²³

Is AI Endurance worth it?

For self-coached runners, cyclists and triathletes with a long GPS or power history and a chest strap, yes, it is worth trying. For beginners, swimmers and strength-first athletes, it offers less. Use the 14-day free trial with your full history connected.²³

How much does AI Endurance cost?

$20 a month or $200 a year in US dollars, on the web and the US App Store. UK App Store prices are £15.00 and £150.00, and coaches pay $15 per athlete per month ($12 from 10 athletes).²⁵¹²¹¹

Does AI Endurance have a free trial?

Yes. “Try everything free for 14 days, no payment info required.”²

Is AI Endurance accurate?

It claims race predictions within about 5% once you have at least 100 GPS runs or power rides. That claim has no published validation that we found. Its DFA alpha 1 calculation matched Kubios in one 2021 blog analysis, not in a peer-reviewed study.³²⁰

Does AI Endurance work with Apple Watch?

Only through the third-party app Watchletic, which uploads activities and runs workouts on the watch. Since May 2026, Watchletic can compute DFA alpha 1 from a supported chest strap. Apple Health is not a listed integration.³³⁸

Does AI Endurance work with Garmin?

Yes. It imports Garmin activities, sends workouts to your watch, and reads in-workout HRV if you wear a chest strap. Garmin does not pass third-party activities, such as Zwift rides, to AI Endurance.³

Do I need a chest strap for AI Endurance?

Only for DFA alpha 1 and in-workout HRV. Optical wrist sensors on Garmin and Suunto don’t record it. AI Endurance names the Polar H10 as an example and recommends Bluetooth over ANT+.³

AI Endurance vs TrainerRoad: which is better?

TrainerRoad for riders who only cycle; AI Endurance for runners and triathletes, or cyclists who want DFA alpha 1. TrainerRoad costs $209.99 a year with a 30-day money-back guarantee; AI Endurance costs $200 with a 14-day trial.⁴⁵²

Does AI Endurance plan swim workouts?

Yes, but not with AI. Swim sessions come from “a workout database created by an experienced coach,” with zones from your critical swim speed.³

Does AI Endurance use ChatGPT?

Its in-app chat runs on Google’s Gemini 3.7 Flash by default, with Claude Sonnet 5.5 as an option for subscribers. Separately, a ChatGPT app went live in June 2026, and an MCP server connects ChatGPT and Claude to your data.³⁹⁴¹

The bottom line

AI Endurance is a serious, data-driven self-coaching app for runners, cyclists and triathletes, and its makers document its limits openly. Pay the $200 a year if you have the GPS or power history and a chest strap. Skip it if your history is thin or your training is mostly outside running, cycling and triathlon.²³¹⁴

Its weak spots are verification, not ambition. The headline accuracy figures are the company’s own, and its DFA alpha 1 rests on research that works best on a fresh body and clean data.³¹³⁵

If the endurance plan is only half of your training, and strength is the other half, that half needs a different kind of coach.

---

References

Footnotes

1. AI Endurance. Homepage — “AI Endurance trains a Digital Twin of you: a neural network that learns from your own workouts how you respond to training, then builds the plan that maximizes your predicted performance on race day.” Links to the iOS (id6714485075) and Android (com.aiendurance.twa) apps. Accessed October 15, 2026. https://aiendurance.com/en ↩ ↩² ↩³ ↩⁴ ↩⁵

2. AI Endurance. “Pricing” — “$20 per month · Billed monthly in USD”; “$200 per year · Billed yearly in USD · Save $40!”; “Try everything free for 14 days, no payment info required.”; “One subscription, everything included.” Viewed from a Brazilian connection: “R$69.90 per month” and “R$699.90 per year.” Accessed October 15, 2026. https://aiendurance.com/en/pricing ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹² ↩¹³ ↩¹⁴ ↩¹⁵ ↩¹⁶ ↩¹⁷ ↩¹⁸ ↩¹⁹ ↩²⁰

3. AI Endurance. “FAQ” — model “updated every 24 hours”; sports, goal types, duathlon, B and C races; swim workouts from “a workout database created by an experienced coach”; CSS swim zones; “We deliberately do not include test workouts”; cycling power requirement; population-model fallback for new users; zone detection from DFA a1, then power/pace; chest strap requirement; Bluetooth vs ANT+; artifact cutoffs (under 10% for readiness, under 5% for AI predictions); Coros, Wahoo, Zwift, Rouvy and TrainerDay record no activity HRV; Polar HRV not in shared FIT files; Strava passes no HRV; live DFA a1 via alphaHRV, the AI Endurance Karoo extension and Watchletic 3.1.0+; Apple Watch through Watchletic; overnight Apple Watch HRV “indirectly” via Intervals.icu; “We can predict your future performance at about 5% if you have at least 100 runs with GPS data or rides with power data”; TrainingPeaks export and overnight HR/HRV import only; “Garmin does not pass on third party activities”; one-week export window; “Our AI does not detect injuries in any form”; chat memory not carried into plan settings unless saved to the Plan System Prompt. Accessed October 15, 2026. https://aiendurance.com/en/faq ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹² ↩¹³ ↩¹⁴ ↩¹⁵ ↩¹⁶ ↩¹⁷ ↩¹⁸ ↩¹⁹ ↩²⁰ ↩²¹ ↩²² ↩²³ ↩²⁴ ↩²⁵ ↩²⁶ ↩²⁷ ↩²⁸ ↩²⁹ ↩³⁰ ↩³¹ ↩³² ↩³³ ↩³⁴ ↩³⁵ ↩³⁶ ↩³⁷ ↩³⁸ ↩³⁹ ↩⁴⁰ ↩⁴¹ ↩⁴² ↩⁴³ ↩⁴⁴ ↩⁴⁵ ↩⁴⁶ ↩⁴⁷ ↩⁴⁸ ↩⁴⁹ ↩⁵⁰ ↩⁵¹ ↩⁵² ↩⁵³ ↩⁵⁴ ↩⁵⁵ ↩⁵⁶ ↩⁵⁷ ↩⁵⁸ ↩⁵⁹ ↩⁶⁰ ↩⁶¹ ↩⁶² ↩⁶³ ↩⁶⁴ ↩⁶⁵ ↩⁶⁶ ↩⁶⁷ ↩⁶⁸

4. AI Endurance. “Product” — Digital Twin “searches the possible training futures, volume, intensity, zone distribution”; predicted improvement vs repeating your own training; recovery model from HRV, resting heart rate, DFA alpha 1, training load, morning check-in and soreness; chat “can act: analyzing activities second by second, creating and rescheduling workouts, adjusting your zones”; calorie, macronutrient and meal-plan features; import from Garmin, Suunto, Coros, Polar, Wahoo, Hammerhead, Intervals.icu, Strava, Stryd, Oura, Whoop and Zwift; export to Garmin, Suunto, Coros, Wahoo, Hammerhead, TrainingPeaks, Intervals.icu, Nolio, Zwift, Rouvy and TrainerDay. Accessed October 15, 2026. https://aiendurance.com/en/product ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹² ↩¹³ ↩¹⁴ ↩¹⁵ ↩¹⁶

5. AI Endurance Inc. “AI Endurance · Adaptive AI endurance coach.” Apple App Store listing (US, id6714485075) — in-app purchases Monthly Subscription $20.00, Yearly Subscription $200.00; seller AI Endurance Inc; 4.2 from 24 ratings; reads recovery data “from Oura, Whoop, and your watch”; App Store reviews quoted in this article. iTunes lookup API: first released October 16, 2024; version 4.0.0 released July 15, 2026. Accessed October 15, 2026. https://apps.apple.com/us/app/ai-endurance/id6714485075 and https://itunes.apple.com/lookup?id=6714485075&country=us ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰ ↩¹¹ ↩¹² ↩¹³ ↩¹⁴ ↩¹⁵ ↩¹⁶

6. AI Endurance. Google Play listing (com.aiendurance.twa) — “5K+ Downloads”; “Updated on Jul 7, 2026”; no star rating displayed. Accessed October 15, 2026. https://play.google.com/store/apps/details?id=com.aiendurance.twa&hl=en_US&gl=US ↩ ↩² ↩³ ↩⁴

7. Andriolo S, Rummel M, Gronwald T. “Relationship of Cycling Power and Non-Linear Heart Rate Variability from Everyday Workout Data: Potential for Intensity Zone Estimation and Monitoring.” Sensors (Basel), 2024;24(14):4468. PMID: 39065866. Authors Andriolo and Rummel affiliated with AI Endurance Inc., Hamilton, ON. https://pubmed.ncbi.nlm.nih.gov/39065866/ ↩ ↩² ↩³ ↩⁴

8. AI Endurance. “About” — Markus Rummel, founder, triathlete (2:06 Olympic distance, 4:34 Half-Ironman) and particle physicist “who worked at world leading institutions such as University of Oxford”; Chris St. Jean, co-founder and ultramarathon runner based in Colorado Springs; “AI Endurance exists because modern endurance training is a data problem, and we are scientists.”; quote by Rummel on machine learning. Accessed October 15, 2026. https://aiendurance.com/en/about ↩ ↩² ↩³ ↩⁴ ↩⁵

9. AI Endurance. Blog index — earliest dated post “How AI Endurance Training can help you reach your goals,” published May 8, 2020. Accessed October 15, 2026. https://aiendurance.com/blog ↩

10. AI Endurance. Sitemap hreflang list (EN, DE, FR, ES, PT, IT, RU, PL, NL). Accessed October 15, 2026. https://aiendurance.com/sitemap.xml ↩

11. AI Endurance. “For coaches” — “1-9 athletes $15 USD per athlete / month · 10+ athletes $12 USD per athlete / month · Clubs & teams (25+) Let’s talk”; “Your first athlete is free”; “the direct athlete subscription is $20 USD per month.” Accessed October 15, 2026. https://aiendurance.com/coaches ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶

12. AI Endurance Inc. Apple App Store listings (id6714485075) — UK: £15.00 / £150.00, 4.4 from 7 ratings; Germany: 18,00 € / 179,99 €, 4,1 from 17 ratings; Canada: $25.00 / $249.99, 4.8 from 9 ratings; Australia: $30.00 / $300.00, 3.7 from 3 ratings. Accessed October 15, 2026. https://apps.apple.com/gb/app/ai-endurance/id6714485075 ; https://apps.apple.com/de/app/ai-endurance/id6714485075 ; https://apps.apple.com/ca/app/ai-endurance/id6714485075 ; https://apps.apple.com/au/app/ai-endurance/id6714485075 ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹ ↩¹⁰

13. Triathlete. “AI triathlon training apps” (published September 3, 2025; updated March 11, 2026) — AI Endurance “Cost: $12.99/month; $129.99/year”; chatbot “far less capable and slower than the Maxiom platform’s chatbot.” Accessed October 15, 2026. https://www.triathlete.com/gear/tech-wearables/ai-triathlon-training-apps/ ↩ ↩²

14. AI Endurance. “Which training really works?” (April 15, 2021) — simulated 11.1% average gains in FTP and 5k pace across 126 runners, cyclists and triathletes over an 8-week plan; “a certain bias is introduced via the choice of data processing and machine learning model”; 89% polarized vs 11% threshold; “we restrict the possible input data up to one standard deviation away from the mean of the athlete’s historical data.” Accessed October 15, 2026. https://aiendurance.com/blog/which-training-really-works ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷

15. Mann TN, Lamberts RP, Lambert MI. “High responders and low responders: factors associated with individual variation in response to standardized training.” Sports Medicine, 2014;44(8):1113-24. PMID: 24807838. https://pubmed.ncbi.nlm.nih.gov/24807838/ ↩

16. AI Endurance. “Introducing the Plan System Prompt” (November 11, 2025) — “This new feature lets you directly tell our LLM-based AI agent how you want your training to evolve — in your own words.” Accessed October 15, 2026. https://aiendurance.com/blog/introducing-the-plan-system-prompt ↩ ↩²

17. Seiler S. “What is best practice for training intensity and duration distribution in endurance athletes?” International Journal of Sports Physiology and Performance, 2010;5(3):276-91. PMID: 20861519. https://pubmed.ncbi.nlm.nih.gov/20861519/ ↩

18. Rosenblat MA, Watt JA, Arnold JI, Treff G, Sandbakk ØB, Esteve-Lanao J, Festa L, Filipas L, Galloway SD, Muñoz I, Ramos-Campo DJ, Schneeweiss P, Sellés-Pérez S, Stöggl T, Talsnes RK, Zinner C, Seiler S. “Which Training Intensity Distribution Intervention will Produce the Greatest Improvements in Maximal Oxygen Uptake and Time-Trial Performance in Endurance Athletes? A Systematic Review and Network Meta-analysis of Individual Participant Data.” Sports Medicine, 2025;55(3):655-673. PMID: 39888556. https://pubmed.ncbi.nlm.nih.gov/39888556/ ↩

19. AI Endurance. “DFA alpha 1 thresholds from heart rate variability” (updated October 5, 2026) — “DFA alpha 1 drops below 0.75 at the aerobic threshold and below 0.5 at the anaerobic threshold.”; “AI Endurance calculates DFA alpha 1 at high, Kubios-like accuracy, see for example this review by Bruce Rogers.”; threshold detection rates; “cluster thresholds were not studied yet in the research literature.” Accessed October 15, 2026. https://aiendurance.com/blog/dfa-alpha-1-thresholds-from-heart-rate-variability ↩ ↩² ↩³

20. Rogers B. “AI Endurance - DFA a1 accuracy initial review.” Muscle Oxygen Training (blog), posted July 7, 2021, updated July 13, 2021 — “The correlation of AI endurance is excellent ( r of .96 and a mean difference of only about 5% at low DFA a1).”; “This was based on data given to me by the AI Endurance team.”; “a 2 hour session”; “AI Endurance DFA a1 calculation appears to closely match that of Kubios premium.” The blog’s About page describes its author as a “board certified Endocrinologist with 30 years of clinical practice experience.” Not peer-reviewed. Accessed October 15, 2026. http://www.muscleoxygentraining.com/2021/07/ai-endurance-dfa-a1-accuracy-initial.html ↩ ↩² ↩³ ↩⁴ ↩⁵

21. Rummel M. “Cycling durability from everyday rides.” AI Endurance blog (October 5, 2026) — 105 amateur cyclists and triathletes, 4,850 rides; after 15 kJ/kg, heart rate 10 bpm higher, DFA alpha 1 0.17 lower, breathing 4 breaths per minute faster (median); repeatability correlation 0.72 (heart-rate drift) vs 0.37 (power-based); paper to appear in the Journal of Science and Cycling; “The cohort is almost entirely male amateur athletes.” Accessed October 15, 2026. https://aiendurance.com/blog/cycling-durability-everyday-rides ↩ ↩² ↩³ ↩⁴

22. AI Endurance. “Terms of Use” — “Our performance predictions are estimations and come together with a prediction error on top of uncontrolled variables such as weather, form on the day, etc.” Accessed October 15, 2026. https://aiendurance.com/termsofuse ↩

23. Keogh A, Smyth B, Caulfield B, Lawlor A, Berndsen J, Doherty C. “Prediction Equations for Marathon Performance: A Systematic Review.” International Journal of Sports Physiology and Performance, 2019;14(9):1159-1169. PMID: 31575820. https://pubmed.ncbi.nlm.nih.gov/31575820/ ↩

24. Lerebourg L, Saboul D, Clémençon M, Coquart JB. “Prediction of Marathon Performance using Artificial Intelligence.” International Journal of Sports Medicine, 2023;44(5):352-360. PMID: 36473492. https://pubmed.ncbi.nlm.nih.gov/36473492/ ↩

25. Smyth B, Muniz-Pumares D. “Calculation of Critical Speed from Raw Training Data in Recreational Marathon Runners.” Medicine & Science in Sports & Exercise, 2020;52(12):2637-2645. PMID: 32472926. https://pubmed.ncbi.nlm.nih.gov/32472926/ ↩

26. AI Endurance Forum, announcements — “If you have a Runner or Triathlete account you can now pick Critical Speed as an optimization target.” (October 8, 2026) https://forum.aiendurance.com/t/224 ; “You can now see your durability, the loss of performance after a given amount of work (in cycling kJ kilo Joule and in running kilometers) in the Data page” (July 8, 2026) https://forum.aiendurance.com/t/174 . Accessed October 15, 2026. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶

27. Granero-Gallegos A, González-Quílez A, Plews D, Carrasco-Poyatos M. “HRV-Based Training for Improving VO2max in Endurance Athletes. A Systematic Review with Meta-Analysis.” International Journal of Environmental Research and Public Health, 2020;17(21):7999. PMID: 33143175. https://pubmed.ncbi.nlm.nih.gov/33143175/ ↩

28. Manresa-Rocamora A, Sarabia JM, Javaloyes A, Flatt AA, Moya-Ramón M. “Heart Rate Variability-Guided Training for Enhancing Cardiac-Vagal Modulation, Aerobic Fitness, and Endurance Performance: A Methodological Systematic Review with Meta-Analysis.” International Journal of Environmental Research and Public Health, 2021;18(19):10299. PMID: 34639599. https://pubmed.ncbi.nlm.nih.gov/34639599/ ↩

29. Rogers B, Giles D, Draper N, Hoos O, Gronwald T. “A New Detection Method Defining the Aerobic Threshold for Endurance Exercise and Training Prescription Based on Fractal Correlation Properties of Heart Rate Variability.” Frontiers in Physiology, 2020;11:596567 (published online January 2021). PMID: 33519504. https://pubmed.ncbi.nlm.nih.gov/33519504/ ↩ ↩² ↩³

30. Rogers B, Giles D, Draper N, Mourot L, Gronwald T. “Detection of the Anaerobic Threshold in Endurance Sports: Validation of a New Method Using Correlation Properties of Heart Rate Variability.” Journal of Functional Morphology and Kinesiology, 2021;6(2):38. PMID: 33925974. https://pubmed.ncbi.nlm.nih.gov/33925974/ ↩ ↩²

31. Van Hooren B, Mennen B, Gronwald T, Bongers BC, Rogers B. “Correlation properties of heart rate variability to assess the first ventilatory threshold and fatigue in runners.” Journal of Sports Sciences, 2025;43(2):125-134 (published online November 2023). PMID: 37916488. https://pubmed.ncbi.nlm.nih.gov/37916488/ ↩ ↩² ↩³ ↩⁴

32. Gronwald T, Rogers B, Hoos O. “Fractal Correlation Properties of Heart Rate Variability: A New Biomarker for Intensity Distribution in Endurance Exercise and Training Prescription?” Frontiers in Physiology, 2020;11:550572. PMID: 33071812. https://pubmed.ncbi.nlm.nih.gov/33071812/ ↩

33. Schaffarczyk M, Rogers B, Reer R, Gronwald T. “Validation of a non-linear index of heart rate variability to determine aerobic and anaerobic thresholds during incremental cycling exercise in women.” European Journal of Applied Physiology, 2023;123(2):299-309. PMID: 36269394. https://pubmed.ncbi.nlm.nih.gov/36269394/ ↩

34. Rogers B, Giles D, Draper N, Mourot L, Gronwald T. “Influence of Artefact Correction and Recording Device Type on the Practical Application of a Non-Linear Heart Rate Variability Biomarker for Aerobic Threshold Determination.” Sensors (Basel), 2021;21(3):821. PMID: 33530473. https://pubmed.ncbi.nlm.nih.gov/33530473/ ↩

35. Schaffarczyk M, Rogers B, Reer R, Gronwald T. “Validity of the Polar H10 Sensor for Heart Rate Variability Analysis during Resting State and Incremental Exercise in Recreational Men and Women.” Sensors (Basel), 2022;22(17):6536. PMID: 36081005. https://pubmed.ncbi.nlm.nih.gov/36081005/ ↩ ↩²

36. Rogers B, Gronwald T. “Fractal Correlation Properties of Heart Rate Variability as a Biomarker for Intensity Distribution and Training Prescription in Endurance Exercise: An Update.” Frontiers in Physiology, 2022;13:879071. PMID: 35615679. https://pubmed.ncbi.nlm.nih.gov/35615679/ ↩

37. Rogers B, Murias JM, Fleitas-Paniagua PR. “Validity of an open-source mobile app to measure fractal correlation properties of heart rate variability during exercise.” European Journal of Applied Physiology, 2026;126(4):2023-2035. PMID: 41171429. https://pubmed.ncbi.nlm.nih.gov/41171429/ ↩

38. AI Endurance Forum, announcements — Watchletic “now adds support for connecting to heart rate monitors directly in the app … calculate DFA Alpha 1 if the monitor supports it, e.g. Polar H10.” (May 25, 2026) https://forum.aiendurance.com/t/144 ; Zwift, Rouvy, Hammerhead and TrainerDay: “All 4 integrations support training plan export … automatic import … historical import upon first connection” (June 1, 2026) https://forum.aiendurance.com/t/148 . Accessed October 15, 2026. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶

39. AI Endurance Forum, announcements — “The chat now uses Gemini 3 (Flash and Pro available) as the default chat models” (February 12, 2026) https://forum.aiendurance.com/t/32 ; “The chat now has cross-conversation memory.” (February 26, 2026) https://forum.aiendurance.com/t/63 ; weekly replan via chat (April 1, 2026) https://forum.aiendurance.com/t/111 ; “After a 6 month review process , our ChatGPT app is now live” (June 16, 2026) https://forum.aiendurance.com/t/153 ; “Gemini 3.7 Flash is now the default chat model, replacing 3.0 Flash.” (September 17, 2026) https://forum.aiendurance.com/t/212 ; “If you’re a subscriber you can now pick Claude Sonnet 5.5 in the Chat as an alternative to Gemini 3.1 Pro.” (September 29, 2026) https://forum.aiendurance.com/t/217 . Accessed October 15, 2026. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸

40. AI Endurance Forum. “Announcements” category (topic dates) — bulk .FIT upload (February 24, 2026); Coaching Portal (April 7, 2026); new design (July 6, 2026); dark mode (July 7, 2026); run and bike goals for triathletes (October 8, 2026). Accessed October 15, 2026. https://forum.aiendurance.com/c/announcements/6 ↩ ↩² ↩³ ↩⁴

41. AI Endurance. “MCP” documentation — MCP server for ChatGPT and Claude; “Strength & Other Workouts — Create custom strength training, yoga, hiking, cross-country skiing, and other non-endurance activities with free-form workout descriptions.” Accessed October 15, 2026. https://aiendurance.com/docs/mcp ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶

42. Maunder E, Seiler S, Mildenhall MJ, Kilding AE, Plews DJ. “The Importance of ‘Durability’ in the Physiological Profiling of Endurance Athletes.” Sports Medicine, 2021;51(8):1619-1628. PMID: 33886100. https://pubmed.ncbi.nlm.nih.gov/33886100/ ↩

43. Rogers B, Fleitas-Paniagua PR, Trpcic M, Zagatto AM, Murias JM. “Fractal correlation properties of heart rate variability and respiratory frequency as measures of endurance exercise durability.” European Journal of Applied Physiology, 2025;125(6):1619-1631. PMID: 39904800. https://pubmed.ncbi.nlm.nih.gov/39904800/ ↩

44. SensAI Help Center. “How to Connect Garmin, Oura, WHOOP, Strava, or Any Wearable to SensAI Through Apple Health.” Last updated September 25, 2026 — “SensAI reads Apple Health only”; HRV “Only from Apple Watch”; Garmin, COROS, Polar and Strava workouts reach SensAI through Apple Health; “Their proprietary scores never cross”; “It syncs roughly the last 30 days of Apple Health workouts”; “Your swims, rides, and runs sit next to your planned strength sessions in the Activity tab”; “A heavy cycling block shows up as load that strength volume has to fit around.”; “Sleep and overnight heart rate from your device shape how hard Coach asks you to push on a given day.” Accessed October 15, 2026. https://www.sensai.fit/help/connect-wearable-to-sensai-apple-health ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷

45. TrainerRoad. “Pricing” — Monthly $21.99 USD; Annual $209.99 USD; “30-day money-back guarantee. Cancel anytime.”; “The #1 Cycling App for Getting Faster”; “TrainerRoad AI … AI FTP Detection · Fatigue Prevention · Machine Learning Optimization.” Accessed October 15, 2026. https://www.trainerroad.com/pricing ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸

46. Athletica. “Pricing” — “2-Week Trial … Free”; “$19.90 /month”; “$99.00 /6 month”; “$189.00 /year”; “No credit card required”; sports “triathlon, duathlon, running, cycling, rowing and HYROX”; AI Coach with a “curated sport science knowledge base”; “The AI Coach does not automatically modify your training.” Accessed October 15, 2026. https://athletica.ai/pricing ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷

47. TrainingPeaks. “Pricing for athletes” — “A monthly subscription costs $19.95/month, while a yearly subscription costs $134.99.”; “We offer a 14-day free trial”; coached plans Bronze $149, Silver $229, Gold $359 per month. Accessed October 15, 2026. https://www.trainingpeaks.com/pricing/for-athletes/ ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷

48. Runna. “Pricing” — “Annual Plan $119.99/year”; “Monthly Plan $19.99/month”; “Free 7-day Trial.” Accessed October 15, 2026. https://www.runna.com/pricing ↩ ↩² ↩³ ↩⁴ ↩⁵

49. JOIN. “Pricing” — “JOIN costs €16,99 per month or €119,99 per year for a subscription. A free 7-day trial is also available”; “Extensive database with 400+ workouts.” Accessed October 15, 2026. https://join.cc/pricing ↩ ↩² ↩³ ↩⁴

50. Humango. “FAQs” — “Essential: $16.99/month or $155.99/annually USD”; “Premium: $28.99/month or $264.99/annually USD”; “The free trial version is equivalent to the Premium level.” The App Store listing shows additional tiers. Accessed October 15, 2026. https://humango.ai/faqs and https://apps.apple.com/us/app/humango-ai-training-planner/id1554430755 ↩ ↩² ↩³ ↩⁴ ↩⁵