The three-second pause that decides everything
Here is a moment that plays out thousands of times a day, in every corner of Canada. Someone phones a dental clinic in Calgary, a law office in Halifax, a flooring company in Mississauga. A warm, natural voice answers and asks how it can help. About three seconds in, the caller goes quiet. Then comes the question: "Sorry, am I talking to a real person?"
What your business does with that moment matters more than you would guess. Not because callers can always tell the difference — today's AI voice agents answering business phones are good enough that plenty of people cannot — but because the instant someone wonders, you have reached the real question. It was never whether you can quietly let an AI answer your phones. It is whether you should say so.
I will put my cards on the table. You should tell them. Upfront, in the first sentence, in plain words. The rest of this article is about why: the law, the money, and the part most vendors skate right past.
The short answer is yes — and the reasons aren't the ones you'd expect
Most owners assume the disclosure question is a legal one. It is, partly. But the stronger reasons are commercial and human, and they point the same way as the law is drifting.
So why does anyone hesitate? Two fears, mostly. The first is that saying "you're speaking with a virtual assistant" sounds cold, corporate, a little embarrassing — like admitting you were too cheap to hire a person. The second is sharper: that the caller will hang up the second they hear the word AI, and a lead worth hundreds of dollars evaporates over one sentence.
Both fears are understandable. Both turn out to be mostly wrong, and I will get to the evidence. First, the rules, because you cannot make a confident call about disclosure until you know what Canada actually requires — and what it doesn't.
What Canada's rules actually say (and what they don't)
Start with the part people get wrong. Canada already has robocall rules that can reach synthetic voices. The CRTC's Unsolicited Telecommunications Rules define an "automatic dialing-announcing device" as equipment that delivers a "pre-recorded or synthesized voice message." That word — synthesized — has sat in the text for years, long before the current wave of AI. You can read the framework yourself on the CRTC's telemarketing rules page.
For calls that go out to solicit business, the rules are strict. You generally need express consent before you dial, you have to identify who is calling, you have to honour the National Do Not Call List, and you have to stay inside permitted calling hours. Get it wrong and the penalties are real: the CRTC can levy up to CAD $1,500 per violation against an individual and up to CAD $15,000 per violation against a corporation. Multiply that by a dialer's call volume and you see why compliance teams lose sleep.
Now the distinction almost every article skips. Those rules were built for outbound solicitation — the business dialing the customer. An inbound AI receptionist is a different animal. When a customer phones you, the consent-to-call machinery does not apply the same way, because the customer initiated the contact. So let me be precise, because precision is the whole point here: having an AI answer your inbound line is not, on its own, illegal just because you didn't announce it.
That is exactly why this is an opinion piece and not a compliance memo. On inbound calls, the case for disclosure is not a blanket federal ban. It is built out of privacy law, recording law, provincial rules, reputation, and plain commercial sense. And once you stack those up, the case gets strong fast.
Quebec adds its own layer: Law 25 and French first
If your business takes calls from Quebec — and a lot of Canada-wide operations do — two extra things apply, whether your office is in Montreal or you are a Toronto company with Quebec customers.
The first is Law 25, Quebec's modernized private-sector privacy law. It leans hard on transparency. Among other duties, it requires an organization to inform a person when a decision about them is made exclusively by automated processing. An AI that screens, routes, qualifies or books based on what a caller says is doing automated processing of personal information, and the moment you record the call you have collected personal data that carries its own notice and consent obligations. We went deeper on this in our piece on what privacy law asks of AI phone calls, and it is worth your time if Quebec is in your footprint.
The second is the Charter of the French Language, Loi 96. Service to the public in Quebec has to be available in French. A voice agent that only greets callers in English is not just off-brand there — it is offside. Practically, that means a bilingual agent that opens in French and switches on cue. You can confirm the privacy side of this with the regulator directly through the federal Privacy Commissioner, and Quebec businesses should look to the province's own access-to-information commission.
Put the federal and Quebec pieces together and a pattern emerges. Nothing yet forces a scripted "I am an AI" line on every inbound call. But everything in the rulebook rewards honesty about who — or what — is on the line, and about what you are doing with the recording.
Where this is heading in 2026
Regulators are not standing still. In 2026 the CRTC opened a review of its Unsolicited Telecommunications Rules, with the rulebook out for public comment and AI voice squarely in the conversation. The direction of travel is toward clearer, firmer expectations, not looser ones.
Look south and the picture sharpens. In February 2024 the U.S. FCC ruled that AI-generated voices count as "artificial or prerecorded voice" under the Telephone Consumer Protection Act, so they carry the same consent rules as robocalls. California, Utah and Colorado have all passed bot-disclosure laws. Canada and the United States rarely move in lockstep, but they rarely move in opposite directions either.
Here is the strategic read. Disclosure is shifting from a nice-to-have to a norm. The businesses that adopt the norm now, while it is still a choice, get to do it on their own terms and in their own voice. The ones that wait get to do it in a hurry, the week a rule lands, with legal breathing down their neck. One of those is a much better morning than the other.
The business case for saying it out loud
Set the law aside for a second and look at the numbers behind human behaviour. Transparent, upfront disclosure tends to improve call completion and customer satisfaction, not wreck it. That surprises people, so it is worth understanding why.
It comes down to when trust breaks. A caller who is told "you're speaking with our AI assistant" at hello knows the deal and adjusts. A caller who figures it out at minute four — after they have spelled their name twice and shared a health concern or a credit issue — does not feel clever. They feel tricked. And a customer who feels tricked mid-call is not just a lost call; they are a one-star review, a warned neighbour, a story told at dinner. Deception discovered late is one of the most expensive things a small business can buy.
There is an upside beyond avoiding harm, too. Saying it plainly signals confidence. It tells the caller the tool is good enough that you are not hiding it, and it sets a fair expectation: fast answers, available at 2 a.m., happy to pass you to a human. If you are still weighing vendors, honesty about disclosure belongs on your checklist — alongside the other questions worth asking before you buy an AI voice agent.
The market is going this way regardless. The virtual receptionist market is worth roughly USD $4.6 billion in 2026 and climbing. When a technology goes mainstream, the norms around using it honestly form fast — and customers start expecting them.
The best argument against disclosure — and why it loses
Let me make the other side's case as well as I can, because it deserves that. The argument goes: disclosure depresses conversion, it makes the interaction feel colder, and — the honest kicker — plenty of competitors are not doing it, so announcing your AI just hands them the callers who flinch. If a warm human-sounding voice books the appointment either way, why introduce friction?
It is a real argument, and on a single call it can even be right. Some callers do hesitate at the word AI. But the case falls apart over any timeframe longer than one phone call.
The conversion hit is smaller than feared and shrinks every month as AI answering becomes ordinary. The "colder" problem is a scripting problem, not a disclosure problem — a well-written line is warmer than a human reading from a bad one. And "nobody else is doing it" is not a strategy; it is a countdown. The competitor who skips disclosure is not winning, they are borrowing against the day a caller feels deceived or a rule changes. Building your phone experience on a practice you would rather customers not notice is a weak foundation, no matter how good this quarter's numbers look.
How to disclose without killing the call
Good news: doing this well takes about one sentence and a little care. Say it first, keep it warm, keep it short, and always leave an obvious door to a human.
Put the disclosure in the opening line, before you start collecting information — not buried after the caller has already answered three questions. Name the business, name that it is an AI assistant, say what it can do, and make the handoff to a person feel easy rather than hidden. If a caller might reach you from Quebec, greet in French and switch on request. And because you are almost certainly recording, fold recording consent into the same breath.
Here is a line you can lift for an English-speaking caller:
"Hi, thanks for calling [Business]. You're speaking with our AI assistant — I can book appointments, answer quick questions, or get you to someone on the team any time. This call may be recorded so we get the details right. How can I help?"
And the French equivalent for Quebec callers:
"Bonjour, merci d'appeler [Entreprise]. Vous parlez avec notre assistant virtuel — je peux prendre un rendez-vous, répondre à vos questions ou vous transférer à un membre de l'équipe en tout temps. Cet appel peut être enregistré. Comment puis-je vous aider?"
Notice what those lines do. They disclose, they reassure, they offer the human, and they get to business — all before the caller has to ask the awkward question. That is the entire trick.
The bottom line
Tell your callers. Not because a single Canadian law currently forces a disclosure line onto every inbound call — it doesn't — but because the CRTC's rules already reach synthetic voices, Quebec's Law 25 rewards transparency, the regulatory wind is at disclosure's back, and, most of all, because honesty at hello is simply better business than honesty forced out at minute four.
An AI that answers your phones should be something you are glad to mention, not something you hope nobody notices. If it is good, saying so costs you almost nothing and earns you trust you cannot buy back later. If you would rather your customers never found out, that tells you something more important than any rule ever could.
The best time to make disclosure your default was before the norm hardened. The second-best time is on your next call — ideally with a bilingual assistant that introduces itself, in the right language, by design.