On your real question — additional business or the same customers paying you less — there is a way to get evidence rather than a feeling.
Look at total covers, not channel mix. Pull your counter transactions from before the platforms and compare with now. If total orders grew by roughly what delivery added, it is incremental. If total orders are flat and the split changed, you are substituting, and paying commission for the privilege.
Look at the times. Delivery orders concentrated in the evening or in bad weather, when your counter is quiet, are much more likely to be genuinely additional. Delivery orders at your busiest lunch hour are very often people who would have walked in, and they are also the ones costing you the most in kitchen chaos.
Look at the postcodes. Orders from outside your walking catchment are new customers. Orders from the street outside are not.
That third one is the quickest and it is usually the most striking. If a meaningful share of delivery orders come from within a few minutes walk, you are paying a platform to serve people who were already yours.
And the honest thing about substitution: it is not fully reversible. Once a customer has the habit, leaving the platform loses some of them permanently rather than converting them back. Which is an argument for acting sooner rather than later if the numbers say what you suspect.