The reason that antipoverty policy is (or should be) harder than "just give people money" is that we *all* think there are some behavioral effects.

You can't argue that child allowances are better than TANF because they reduce the marginal tax rate on working or getting married unless you think behavioral effects are real.

But if you believe behavioral effects are real, then you have to acknowledge that while there is one margin where people who are single parents and non-working want to get married and work...

...there is another margin where people who are married (or might be married later) and work (or might later) want to (or will want to) not stay/get married or to work.

It's an empirical question which of those margins represents a bigger group and how responsive those groups are to specific policy reforms. But you can't wish away the margin you don't like. (Nor can conservatives.)

More traditional conservatives have dealt with this reality by favoring the combination of a less generous safety net that is (partly) time limited (for some) and work supports to make work more attractive.

That makes moving from marriage and work to non-marriage and non-work unappealing, and it makes moving from single parenthood and non-work to marriage and work more appealing.

(Incidentally, the best explication of this I can think of was @oren_cass 's "height of the net" essay: https://www.nationalreview.com/2014/01/height-net-oren-cass/ )

The more hardcore uber-conservative response (advocated by VERY few) would be to wipe out the entire safety net, in which case marriage and work become maximally attractive. That's too harsh. But there are also solutions that are too lenient.

The point is to do as much good while doing as little harm. But point-in-time poverty rates aren't the only (or even necessarily the most important) indicator of going good/harm.