Everyone here is pointing at traffic quality, which is worth checking, but two other mechanisms fit your numbers better. If that campaign runs Smart Bidding, dropping 20 keywords in at once is a structural change that can push it back into learning, and Google's own docs say conversion rate dips while the system re-tests bids. The second is plain budget dilution the same daily budget now spreads across more keywords, so less of it lands on the ones that were already converting and total conversions fall even when nothing got worse. Before you start pausing things, separate conversion rate from conversion count. If your rate held and only the count dropped, that's budget maths, not junk traffic, and pausing keywords won't fix it.