DTMF (touch-tone signaling)
DTMF, or dual-tone multi-frequency signaling, is the touch-tone method that sends a pair of audio frequencies for each phone-keypad digit or symbol during a call.
How DTMF works
Pressing a key produces a distinctive pair of tones. The receiving phone system detects that combination and translates it into the intended keypad input. An IVR can then use the input to select a menu branch, accept an extension, or collect a reference value.
In modern telephony, the signal may travel as audible tones in the call audio or as separate signaling events associated with the media session. Separate events can be more reliable because audio compression, noise suppression, and transcoding may distort audible tones. Both ends of the call must agree on how the input will be transported and interpreted.
DTMF is not speech recognition. A caller can press a key while another system is playing a prompt, and the receiving application handles that signal as a discrete input. It also should not be treated as inherently secure. Keypad input can represent sensitive information, so collection, storage, logging, and redaction rules still matter.
Why it matters for AI phone calls
AI phone agents often encounter DTMF when they call businesses that use phone menus. The agent may need to wait for the relevant option, send a keypad input, and confirm that the next prompt reflects the expected branch. Timing matters: a signal sent too early, too late, or in an unsupported transport mode may be ignored.
An inbound AI agent may also use DTMF as a fallback when speech is unclear or when a caller prefers the keypad. The conversational prompt and the telephony implementation should agree about what inputs are allowed and what happens after invalid or missing input.
Testing should cover the actual carrier and call route. A keypad flow that works in a browser or direct software call may behave differently after codec conversion or a transfer across telephone networks.