Weight-loss plateau: what is happening and how an AI coach breaks it

Three reasons weight loss stalls

Cause	How big it is	How to tell	Source
Intake drift: portions creep, untracked bites, weekends, "I know this meal by now"	Large; the modelling study found intermittent lapses in adherence, not metabolic adaptation, explain the typical six-month plateau	Two weeks of weighing and logging everything usually reveals it	Thomas et al., 2014
A smaller body burns less: your maintenance falls as you lose weight, so the same intake becomes a smaller deficit	Moderate and predictable; roughly 20–25 kcal/day less maintenance for each kilogram lost, on top of which weight responds slowly, with a half-time of about a year	Recalculate maintenance at your current weight	Hall et al., 2011
Metabolic adaptation: expenditure drops more than the weight loss alone predicts	Real but smaller than its reputation except after extreme loss; in the Biggest Loser follow-up, after an average 58 kg loss, adaptation was about 500 kcal/day at six years and did not predict who regained	Cannot be measured at home; assume it is modest unless your loss has been extreme	Fothergill et al., 2016

Water and glycogen hide fat loss too. A high-salt or high-carbohydrate day, a hard training session, the second half of a menstrual cycle, or a new training program can each add one to two kilograms of water for days. That is why a plateau is defined over weeks, not days, and why a trend line matters more than any weigh-in. How to track fitness progress explains how to read a weight trend, and how often to weigh yourself covers the daily-versus-weekly question.

Diagnose it in three steps

Step 1: Is it a plateau? Weigh daily, same time, and look at the weekly average. If the weekly average has not moved for three weeks, it is a plateau. Two weeks of a flat line after a big drop is usually water and catching up.

Step 2: Recount the deficit at your current weight. The deficit you set at 95 kg is not the deficit you have at 85 kg. Re-estimate maintenance for your current weight and activity (the calories-to-lose-weight guide has the method) and compare it with what you are actually eating. If the gap is under about 300 calories, the arithmetic has caught up with you and a small cut or more activity is the fix.

Step 3: Audit the intake. For two weeks, log everything, weigh what you can, and include the bites, the drinks and the weekends. Most people find 200–400 calories a day they were not counting, which is most of a stalled deficit on its own. If two weeks of tight logging restarts the loss without any change to the target, the plateau was drift, which is the common case.

Only once those three are done does it make sense to talk about adaptation, and the answer there is the same as for a smaller body: a modest adjustment.

What to adjust, and by how much

Lever	Adjustment	Why this size
Calories	Reduce by 100–150 per day, hold for two to three weeks, reassess	Small enough to keep adherence; big enough to show on the trend within a month
Protein	Hold or raise toward 1.6–2.4 g/kg	Protects muscle as the deficit continues; see Helms et al., 2014 and Longland et al., 2016
Movement	Add 2,000–3,000 daily steps or one more training session	Non-exercise activity tends to fall during a diet; replacing it restores part of the deficit without eating less
Diet break	One to two weeks at estimated maintenance	A sanity reset for adherence; return to the deficit at the recounted number
Rate	Aim for 0.5–1 percent of bodyweight per week, not more	Faster rates cost more lean mass; in trained athletes, a slower rate preserved lean mass and strength better (Garthe et al., 2011)

What not to do: cut another 500 calories at once, add two hours of cardio, or stop eating carbohydrates. Each produces a quick water drop that looks like success, followed by hunger, worse adherence, and a bigger rebound in the oscillating pattern the plateau modelling describes.

What an AI coach adds

A plateau is a data problem: you need honest intake, a weight trend, and a rule for when to adjust. It is also a patience problem, because the data arrives slowly and the temptation is to react to one weigh-in. A coach that holds the numbers and applies the rule removes both.

Here is how Bay handles it, free right now:

• Honest intake, with low friction. Meals are logged as a sentence, a photo or a voice note, so the bites and the weekends get recorded rather than skipped. Verified label and chain-menu data is used where it exists and estimates are labelled as estimates. The two-week audit in step 3 is simply two weeks of texting what you eat.
• The trend, not the weigh-in. Daily weigh-ins are fitted to a trend so one salty dinner does not look like a plateau.
• The rule, applied weekly. Every Sunday, Bay compares your logged intake with the trend over the past 14 days and, if the observed rate of loss has fallen behind the goal rate, proposes a change of at most 150 calories, protein unchanged, never below 1,500 calories for men or 1,200 for women. It waits until it has four weigh-ins and ten logged days before proposing anything, so a noisy week cannot trigger a cut, and you approve each change with a tapback. If the trend clearly contradicts your goal mid-week, it sends a review early rather than letting a month go by.
• On demand. Text "am I on track?" any day and the same analysis comes back.

In other words, the plan above is what the weekly review already does; the difference is that it does it every week whether or not you remember to. The AI weight-loss coach guide covers the rest of what Bay does for a deficit.