Every chess player eventually asks some version of the same question: "I'm 900 on Chess.com but 1400 on Lichess, so what am I actually?" The unsatisfying, true answer is: both, and neither. A rating isn't a measurement of you, it's a prediction about your results against a specific pool of opponents, and different pools produce different numbers for the same player.
How a rating actually works
Nearly every system in use descends from the one Arpad Elo designed in the 1960s. The core idea is beautifully simple: your rating and your opponent's rating together predict your expected score. Equal ratings predict 50%. A 200-point edge predicts roughly 75%. After the game, your rating moves based on how your actual result compared to that prediction. Beat someone you were supposed to beat and you gain a little; beat someone 300 points above you and you gain a lot.
Modern sites use refinements of this. Lichess uses Glicko-2 and Chess.com uses Glicko, both of which add a "rating deviation": a measure of how confident the system is about your number. New accounts have high deviation, which is why your rating swings wildly in your first 20 games and settles down later. That's not the site being broken; that's the system admitting it doesn't know you yet and updating fast while it learns.
Why the same player gets different numbers
Ratings are only comparable inside their own pool. Lichess ratings run higher than Chess.com ratings for the same strength, typically by a few hundred points, mostly because Lichess starts new players at 1500 while Chess.com's defaults are lower, and because the average visitor to each site differs. Neither is "inflated" in any meaningful sense. Comparing your Lichess number to a friend's Chess.com number is comparing meters to feet without converting.
The same goes for time controls: bullet, blitz, rapid, and daily are separate ratings because they measure genuinely different skills. Plenty of players are 400 points stronger in rapid than bullet. And FIDE's over-the-board ratings are a different universe again, with a floor of 1400 and a pool of tournament players who are, on average, far more serious than an online pool.
What actually matters
The healthiest way to use a rating: track your OWN number, in ONE pool, over months. It's the trend that carries information, not today's value, and definitely not the comparison to someone else's number from a different site. Day-to-day swings of 50 points are statistical noise wearing a scoreboard costume.
And if you want a sense of how a single game measured up, that's a different question than rating entirely: it's about how much winning probability your moves kept or threw away against that specific opponent. That's exactly what a game review measures, and it's free to run on any game you've played.