Free AI Receptionist: What the One Free Plan and the Trials Cover

Yes, one is genuinely free: Smith.ai's AI Receptionist has a $0 plan with 25 calls a month, then $3.00 per extra call. Rosie, Dialzara and My AI Front Desk offer 7-day free trials, and Goodcall lets you build and test a demo agent free, charging only when you connect a phone number.

"Free" covers five different things here: one standing plan, three time-limited trials and one demo. They suit different jobs. The free plan suits a business with few calls that wants a permanent answer. A trial suits a business choosing between tools. The demo suits a quick listen before you commit. This tutorial sets out what each covers, where it stops, and a test you can run in a week.

Follow me on Instagram@sagnikteaches

The five free offers, side by side

These come from each vendor's own pricing page in November 2026. Where a page did not say, the cell says so.

Connect on LinkedInSagnik Bhattacharya

The pattern is worth noticing. Only Smith.ai publishes a number you can plan around. The trials say "7-day" and little else, so you cannot know in advance how many calls you can test. That is not a reason to avoid them; it is a reason to run your test calls early in the week and to ask the vendor, before you start, how much the trial includes.

Subscribe on YouTube@sagnikteaches

What the Smith.ai free plan includes

Smith.ai's page for the free plan lists 24/7 AI call answering, recording, transcription and summaries, lead qualification and call routing, AI scheduling, native integrations such as Clio and Zapier, and analytics with call history. It also includes Quality Studio, a testing area for trying the receptionist before it takes live calls, with 25 test calls and 50 simulated calls a month.

What it leaves out matters just as much. The page marks these as not on the free plan: dedicated onboarding, live phone, email and chat support, custom SMS branding (Pro), and custom AI prompting, custom integrations, full-service setup and priority support (Enterprise). So on the free plan you set the receptionist up yourself and you have no phone support if it goes wrong.

Paid Pro starts at $150 a month for 75 calls, with $1.80 a call at 150 calls and $1.67 at 300. Past the included calls on Pro, the page lists $2.50, $2.30 and $2.17 for each additional call by tier. The free plan's own $3.00 overage rate is higher still, which gives you the sum below.

The sum: when free stops being cheaper

Imagine a kitchen fitter who receives about 60 calls a month that the AI would answer, mostly after hours when the team is on site. Using the page's figures:

The break-even is at 75 calls. Below that, the free plan wins even with the $3 overage. One thing the pricing page does not say is whether calls past 25 are billed automatically or have to be switched on, so ask before you go live: if extra calls are blocked rather than billed, the receptionist may stop answering after call 25, which is worse than a $3 charge for a business that depends on it. Spam counts less than you might fear: Smith.ai's FAQ says calls from known spam numbers are filtered out before they reach your quota, and you can remove up to 10% of a cycle's calls as spam. Pro, in return, adds onboarding and human support, which for some owners is worth more than the saving.

The three 7-day trials, and what they cost afterwards

A trial is an audition. The question is what the receptionist costs when you keep it. The pricing pages give these starting points.

The units differ, which is where owners trip. Rosie and Dialzara count minutes. Goodcall counts unique customers, meaning each phone number counted once a month. Smith.ai counts calls. To compare them you need your own figures: how many calls, and how long each lasts. A music school that gets 80 calls a month, averaging two minutes, uses about 160 minutes, which fits Rosie's 250 and Dialzara's Business Pro 220. At 80 distinct callers it also fits Goodcall Starter's 100 customers, though Goodcall's page equates 100 customers with about 150 calls, so check how repeat callers count.

A tested setup: seven days, one business, real numbers

Here is a plan that turns a trial into evidence. It assumes a childcare nursery that misses calls while staff are with children, and wants to see whether an AI can take messages and answer routine questions.

1. Day 0: write the answers. Before you sign up, list the ten questions callers ask: opening hours, fees, waiting list, visits, ages taken. Write the answer to each in one or two sentences. A receptionist is only as good as that sheet.
2. Day 1: build and test alone. Set the business details and test with your own phone. Ring it five times with different questions. Note every wrong or awkward answer.
3. Day 2: fix and re-test. Correct the answers, then ring again with a colleague who has not seen the script.
4. Days 3 to 5: after-hours only. Forward calls only outside opening hours, with a clear fallback: "Leave a message and we'll ring back."
5. Day 6: read every transcript. Count calls handled correctly, calls where the AI guessed, and calls it should have passed to a human.
6. Day 7: decide. Keep, switch tool, or stop, before the trial turns into a charge.

A filled-in test sheet looks like this, illustrative only:

Calls tested: 14 (8 own tests, 6 real after-hours)
Correct answer, no help needed: 9
Took a message properly: 3
Guessed a fee or a date: 1  (fixed in the script)
Should have been passed to a human: 1  (safeguarding query)
Decision: keep for after-hours; no use during the day yet

The safeguarding line is the one to take most seriously. If a caller raises a concern about a child, an elderly client or anyone at risk, the system must route to a person immediately, and you should test that route before anything else. The tutorial on booking mistakes covers how to handle errors that slip through.

The answer sheet that makes any free tool work

Every receptionist in this tutorial answers from information you give it. A weak sheet makes a good tool look bad. Here is a filled-in answer sheet for a music school, illustrative only, with the kinds of line that matter.

Opening hours: Monday to Friday 3pm to 8pm, Saturday 9am to 2pm. Closed Sundays.
Lesson fees: 30 minutes $28, 45 minutes $40, 60 minutes $52. Paid monthly in advance.
Trial lesson: yes, one free 30-minute trial. Booking needs the student's age and instrument.
Instrument hire: violins and guitars, $12 a month. Pianos are not hired out.
Do NOT answer: refunds, complaints, teacher availability on named days.
For those, say: "I'll pass that to the office and someone will ring you back today." Take name and number.
Pass to a human at once: any mention of a safeguarding concern or an injury.

The "Do NOT answer" line is the most valuable. An AI that cheerfully invents a refund policy costs far more than one that takes a message. Spend the first hour of your trial on this sheet and the rest of the week will go well.

Reading a test transcript: what to fix

Here is an illustrative exchange from a test call to a nursery, with the problems marked. It shows the kind of thing you are hunting for when you read transcripts on day six.

Caller: Hi, do you have a place for a two-year-old starting in January?
AI: Yes, we have places available in January.      [PROBLEM: it guessed. The sheet said "waiting list".]
Caller: How much is it per day?
AI: Our daily rate is $85.                           [CHECK: matches the sheet? If not, fix the sheet.]
Caller: Can I come and look round on Saturday?
AI: Visits are by appointment. May I take your name and number?   [GOOD: took a message.]

The first answer is the dangerous one, because it sounds helpful and commits the nursery to something it has not decided. The fix is to add a line to the sheet: "Places: never confirm availability; say we keep a waiting list and take a name, child's age and start month." Re-run the same call and confirm the answer changes. A transcript read costs five minutes and prevents a promise you cannot keep.

What a missed call is worth: the sum behind the decision

Whether to bother with any of this depends on what calls are worth to you. A quick sum, illustrative, for a kitchen fitter:

• Calls missed in a month, mostly while on site: 18.
• Share that would have been real enquiries: about a third, so 6.
• Share of those that become jobs if answered promptly: say one in four, so 1.5 jobs.
• Average job margin: $1,800. Value of the answered calls: about $2,700 a month.

Against that, the Smith.ai free plan costs $0 for 25 calls, and even a $99 plan is small. The figures are guesses, so replace them with your own call log for a fortnight before you trust the answer. The point is the shape: a receptionist does not need to rescue many calls to cover a modest fee, but it does need to answer them correctly.

Do the reverse check too. A music school whose callers are mostly existing parents ringing about timings will rescue little revenue, and the case rests on saving staff time instead. Count the minutes spent on the phone per week and value them at your hourly rate.

What the free versions will not tell you

Four things cannot be learned from a pricing page, so ask the vendor or find out in your test.

• Can I keep my existing number? Goodcall says every plan includes a local number or you can forward your own. Rosie and Dialzara include a number of their own on every plan, but their pricing pages do not say whether you can forward your existing line, so ask before relying on it.
• Which languages? Dialzara's pricing page lists 10 languages, including English and Spanish, on every plan, and Rosie lists English and Spanish on every call; confirm the ones your callers need.
• Where do recordings go, and how long are they kept? Goodcall publishes call history retention of 30, 90 days or unlimited, depending on plan. For the others, ask.
• Can it book into my calendar? Rosie books directly into a calendar only from Scale ($149); its Professional plan sends booking links by text instead. Dialzara's Google and Outlook calendar sync starts on Business Pro ($99), not Lite. Check that the plan you would end up on has it, and test it during the trial.

An AI receptionist also records conversations, and callers may not expect that. Tell callers at the start of the call, and check what your sector's rules say about recording and retention. If you want it to speak in your own voice, read the tutorial on cloning and consent first.

Which free option for which kind of business

A care agency with 15 enquiry calls a month. The Smith.ai free plan fits: 25 calls cover the volume, there is no deadline, and the AI scheduling and call routing handle the typical "can someone ring me back" case. The need for a human fallback is high, so test the escalation first.

A kitchen fitter with 60 calls, mostly after hours. Smith.ai's free plan covers the month at $105 on the sums above. Dialzara's $29 Lite plan can undercut it: 60 minutes are included, then $0.48 a minute, so 60 calls averaging two minutes cost about $58, and Lite stays under $105 until calls average roughly three and a half minutes. The catch is that Dialzara lists after-hours rules and escalation only from Business Pro, so on Lite you would route after-hours calls with your phone's own forwarding. Test both on a week of real calls.

A music school with 120 calls in September and 40 in July. Seasonal peaks make a permanent plan awkward. A free plan with $3 overage is risky in September (95 extra calls at $3 is $285) and cheap in July. A month-to-month paid plan you can pause may beat both, so ask each vendor about cancelling or pausing.

A childcare nursery with 30 calls and sensitive topics. Use the trial for evidence and the free plan for ongoing use if the tests pass, but never as the only route for safeguarding. Always give callers a way to reach a person.

Handing over to a person: the rule that protects customers

Free tools tempt owners to let the AI handle everything, and that is where harm creeps in. Decide, in writing, which calls the AI must never close on its own. A short list works for most small firms:

• Anything mentioning safety, injury, abuse or a vulnerable person: immediate hand-over, or an urgent message to a named person.
• Complaints and refund requests: take details, never decide.
• Prices that are not on the sheet: say a person will confirm, never estimate.
• Callers who ask twice for a human: connect or take a callback number without argument.

Test each rule with a deliberate call during your trial, and check which transfers your eventual plan includes: Rosie lists live and warm transfers only from Scale, while Dialzara lists a blind transfer on every plan and a warm transfer from Business Pro. If the tool cannot transfer on the plan you trial, the callback message is your hand-over, so make sure someone reads it by a set time each morning. A care agency, for instance, might set 8am as the time the first person checks overnight messages, and write that time into the AI's script so callers hear a promise the agency can keep.

Keep a log of every hand-over for the first month. If more than a third of calls need a person, either the answer sheet is thin or the calls are the wrong kind for automation, and that is a useful finding in itself.

Where free plans and trials go wrong

The trial ends on day seven, quietly. Set a calendar reminder for day six. A tool you forgot may begin to charge. My AI Front Desk says no card is needed to start, which avoids this, but Rosie and Dialzara's pages did not say, so assume a card may be requested.

You test only easy calls. A tool that handles "what time do you open?" may fail on "can you do Thursday if the 3pm is taken?" Include at least three awkward calls in your test: a booking change, an angry caller and a question the script does not cover.

You forget to compare against doing nothing. A missed call is not free. If an after-hours enquiry is worth $200 and the receptionist rescues two a month, a $49 plan has paid for itself. The 90-day return tutorial shows how to count it.

The free plan changes. The calls allowance, price and features can all be changed by the vendor. Screenshot the plan page on the day you start.

After the free period: the cost ladder

Once a free period ends, costs tend to climb in steps. The free Smith.ai plan, $29 Dialzara Lite and $49 Rosie Professional sit at the bottom; $79 Goodcall Starter, $99 Dialzara Pro and $99 My AI Front Desk in the middle; and $150 Smith.ai Pro, $149 Rosie Scale and above at the top. Smith.ai's pages show a human-staffed service at $300 a month and up, which is a different product altogether. The AI receptionist cost tutorial walks through the full ladder, and the comparison with a human answering service helps if your callers need a person.

A final rule of thumb. If the free test shows the AI handles at least four of five typical calls correctly and routes the rest to a person, the setup is ready for a limited live run. If it handles fewer than that, the problem is usually the script, not the tool, so rewrite the answer sheet before you switch vendors.

Checklist for choosing and testing

• You have counted your calls and average length for a typical month and a busy one.
• You have written the ten routine questions and their answers.
• You know which unit each tool bills in: calls, minutes or unique customers.
• You have asked about card requirements and what the trial includes.
• You have tested at least three awkward calls and one urgent call that should reach a person.
• You have a fallback message and a named person who reads the transcripts daily.
• You have a calendar reminder before any trial ends.

Work through that list and the free plan or trial gives you a decision, not just a fortnight of novelty.

Free AI receptionist: what owners ask next

Is the Smith.ai free plan really free?

Its pricing page lists $0 a month for 25 calls, then $3.00 for each call beyond that. The page does not say whether a card is needed at sign-up, or whether calls past 25 are billed automatically or need switching on. Confirm both on the sign-up screen before the receptionist takes a live call.

What happens when a trial ends?

The vendors' pages say little. Rosie and Dialzara say 7-day free trial with cancel anytime; My AI Front Desk says no card is required and that trial use depends on your plan or a written offer. Ask each what happens to your number and call flow on day eight.

Can I test with real customers during a trial?

You can, but I would start with test calls from your own phone and a colleague's. Put the trial on after-hours calls only at first, with a human fallback message, so a bad answer costs you little.

Which free option suits a business with 10 calls a month?

Smith.ai's free plan covers 25 calls a month, so ten calls fit comfortably. Trials are the wrong tool for a low-volume, permanent need because they end after seven days. Test the free plan's answers carefully, since you are staying on it.

Further reads

• AI Receptionist for Tradespeople: What It Costs and When It Pays — What it costs and when it pays for a trades business.
• AI Receptionist for Mechanics: What It Costs Against Missed Calls — Cost set against missed calls, for a workshop.
• AI Receptionist vs Front-Desk Hire: The Real Cost for a Salon — The real cost against hiring front-desk staff.
• Best AI Receptionists for Salons, Compared by Price — Receptionists compared by price for appointment businesses.
• AI Receptionist for Retail: Shop Cost, Stock Calls and Limits — Shop calls, stock queries and limits.
• Should a Hair Salon Use an AI Receptionist? — Whether an appointment business should use one at all.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Smith.ai's AI receptionist and receptionist pricing pages, and the pricing pages of Rosie, Dialzara, My AI Front Desk and Goodcall, read in November 2026.