Why isn't it recommended that you use Simple Linear Regression for Time Series data?

Why is it recommended that instead we use autoregressive models (AR Models)?

Regression is designed to learn a pattern in the spatial spread of the data.

The pattern in a time series data lies in the preceding values, it is sequential (as opposed to spatial).

For time series data, Auto Regressive and Moving Average models (or a combination of the two) are used.

Auto Regressive learn tries to predict the current value using the preceding values. Moving Average computes the changing average over time.

Both are powerful basic methods.