There already was a “no religion” option, in the last census it was the first checkbox under the question asking what religion you follow.

He does have a point in that the proposed question would have been a more neutral way of determining whether someone is religious, the combination of the implicit assumption in the old question that having a religion is normal and providing an single tick option for selecting common ones probably does make a small percentage say yes that would not in the proposed question. Claims of coercion and human rights abuse though seem a bit over the top and are probably coming more from a dislike of religions (and their political power) rather than a desire for accurate data.

What wasn’t mentioned in the article but is something I would consider likely that the main difference with the proposed question might not be from the question itself but from extra effort of writing out a religion name rather than ticking a box - it’s a small effort but there’d be a lot of people who just want to go through the questions as fast as possible.