An AI voice assistant calling or answering for a lead is not a grey area under Canadian telecom law — the CRTC has a specific rule set for exactly this, built around one question: is a synthesized voice being used to solicit, and does the person on the other end already have a relationship with you.
Key takeaways
Canada’s telemarketing rules do not have a separate category for AI. They have a category for any device that plays a recorded or generated voice, and an AI voice assistant fits it exactly. The CRTC’s Unsolicited Telecommunications Rules define an Automatic Dialing-Announcing Device, or ADAD, as equipment used, in the CRTC's own description, to convey a pre-recorded or synthesized voice message to a telecommunications number, (crtc.gc.ca) and that phrase — synthesized voice message — is the whole test. It does not matter whether the voice sounds robotic or indistinguishable from a person, and it does not matter whether the words are pre-scripted or generated live by a model. If the call plays a synthesized voice for the purpose of soliciting, the ADAD rules apply to it.
An AI assistant answering an inbound call from a lead who dialled you is a fundamentally different situation from an AI assistant placing an outbound call to someone who has not contacted you. The rules below are calling-out rules, aimed at telemarketing telecommunications initiated by or on behalf of a telemarketer. Answering your own phone with an AI assistant is not telemarketing in the sense these rules regulate — but the moment that assistant places an outbound call to solicit business, the full rule set below is live.
A telemarketing call to a number on the National Do Not Call List generally requires express consent. But not every call needs that consent in the first place — the DNCL’s own consumer guidance sets out when a relationship already exempts a call, and the windows are specific, not open-ended: “you are considered to have an existing business relationship… if you have purchased, leased, or rented a product or service in the last eighteen (18) months… a written contract… still in effect or expired within the last eighteen (18) months; or asked a telemarketer about a product or service within the last six (6) months.” A lead who inquired seven months ago and never became a client is outside the exemption window entirely, and an AI assistant calling that number needs the same express consent any human telemarketer would.
The rules also carry a plain exemption worth knowing on its own: they “do not apply to telemarketing calls made to businesses,” including a personal line also used for business — so an outbound call to a commercial listing contact is not governed by the same consumer-facing DNCL rules, though the ADAD rules below still are, since the ADAD Rules apply regardless of DNCL exemption status.
The CRTC’s ADAD-specific rules apply even to calls that are otherwise exempt from the DNCL, and they are explicit about it: the ADAD Rules apply whether or not the telemarketing telecommunication is exempt from the National DNCL Rules, per the CRTC's own rules. (crtc.gc.ca) An outbound AI voice call still needs its own express consent specific to the ADAD channel which, per the CRTC's rules, must include the specific telecommunications number to which the telemarketing telecommunication may be made, (crtc.gc.ca) and the rules add operational conditions on top: calls restricted, under the CRTC's rules, to 9:00 a.m. to 9:30 p.m. on weekdays and 10:00 a.m. to 6:00 p.m. on weekends, (crtc.gc.ca) in the recipient’s own local time; “sequential dialing… is prohibited”; the calling equipment must, under the CRTC's rules, disconnect within ten (10) seconds of the person receiving the telecommunication hanging up; (crtc.gc.ca) and the message itself has to open with an identification of who the call is on behalf of and how to reach them, repeated at the end if the message runs past sixty seconds. An AI assistant configured without these guardrails is not a grey-area risk — it is a rule violation with a specific, checkable definition.
The rules put the burden of proof squarely on the caller’s side: the CRTC's rules put the onus on the telemarketer to demonstrate that valid express consent was given, (crtc.gc.ca) and telemarketing registration records have to be kept “three (3) years” from creation, ready to produce to the Commission within thirty days of a request. An AI voice platform that does not log and retain consent records in a form you can retrieve on demand is not a compliance shortcut — it is a way to lose the one thing you would need if a complaint ever arrives.
Separate from CRTC’s rules, Canada’s own voluntary code for advanced generative AI systems asks managers of a public-facing AI system to “ensure that systems that could be mistaken for humans are clearly and prominently identified as AI systems.” It is voluntary and it does not bind every AI vendor, but it reflects the direction Canadian AI policy is moving in: a caller is entitled to know whether they are talking to a person or a system, on top of the consent and calling-hours rules above, not instead of them.
Related: the practical setup guide for an inbound AI voice assistant, the short answer on AI handling after-hours calls.
The CRTC rules covered here govern telemarketers initiating outbound calls, not a business answering a call someone chose to place to them. Answering with an AI assistant is a different question from the outbound rules above, though disclosure that the caller is speaking with an AI system is still good practice.
Not under the existing-business-relationship exemption — the DNCL guidance sets that window at eighteen months from a purchase or contract. Outside that window, the call needs fresh express consent like any other outbound telemarketing call.
No — the ADAD definition is specifically about a synthesized or pre-recorded voice message conveyed by telecommunication. A text or messaging-app outreach sits under different rules, including Canada’s anti-spam legislation, which is not what this article covers.
Answer truthfully. Beyond the general direction of Canada’s voluntary AI code on disclosure, a caller who is misled about whether they are speaking to a person is a trust problem for your business well before it is a compliance problem — and it undermines exactly the relationship a first call is meant to start.
A short call can confirm your consent records, calling hours and disclosure are actually built in, not assumed.
