they claim to be internally matched to 50 ohms without even specifying what frequency that would apply to
I looked at two datasheets for these parts.
Both have input and output return loss specs at 250, 950, and 2150 MHz.
One of them had typical Smith charts for S11 and S22 for the same span of frequencies.
How is it possible for this device to be automatically matched over its entire range?
A resistive termination with low capacitive parasitics will be matched across a wide band.
Would it be suitable for a basic, low bandwidth LNA?
That seems to be the application they sell them for.
If you're not using the full band, you likely want to filter the output to avoid capturing noise in the portion of the amplifiers gain band that you aren't using.
I'm between 1-10MHz which is too low for some MMICs
Given they don't give any specs below 250 MHz, you might want to get the eval board and test them out before you commit to designing them in.
Unfortunately they don't say much about the internal circuit from which you could make guesses about how to use it at low frequencies.
They seem to use external inductance in the VCC line to tweak the performance. You might want to try adding some additional inductance there to get good results at very low operating frequencies. Or conversely you might need higher bypass capacitance to give a clean VCC voltage at such a low frequency. Experimentation (or perhaps a call to your local NXP applications engineer) is in order.