NextPhone research report · Published

The Comprehensive 2026 AI Receptionist Call Report

What a year of calls answered by AI shows about when customers call, what they ask, and how they react to an AI on the line.

By Yan Mellata, founder of NextPhone

Every call in this report was picked up by an AI receptionist. An AI answers the calls a person would let ring: the spam, the robocalls, the 2am calls and the Saturday ones. So the calls show who is really on the other end of the line.

The calls come from two kinds of lines. Small businesses across the US, mostly law firms, home services and local services, use NextPhone to answer their business number. People use TTYL, from the same team, to answer their personal phone. The report uses a subset of the calls NextPhone answers, described in How we measured. These are calls that reached the AI after any spam blocking by the phone carrier, so the shares describe this set of calls, not every phone in America. Monthly figures cover the 12 months from October 2025 to September 2026.

This is NextPhone's first annual report. It covers when people call, from after hours to holidays. How people talk to an AI receptionist, from “are you a robot?” to thank-yous and price questions. AI callers: how often AI voice agents now call, and what they want. And spam calls and robocalls: when they ring, what they pitch and which numbers they use. It closes with how to stop spam calls, answers to common questions and how we measured it all.

23.6%

of after-hours calls became a lead or booking, against 18.6% in business hours

1 in 7

calls to small businesses arrives after hours

21.9%

of callers who asked “are you a robot?” still became a lead or booking

4.0

AI callers per 10,000 calls in September 2026, up from 0.7 in late 2025

The key numbers

  • 14.1% of calls to small businesses arrive after hours, and 23.6% of those become a lead or booking, against 18.6% in business hours. See the data
  • 21.9% of callers who ask "are you a robot?" still become a lead or a booking. See the data
  • AI voice agents made 4.0 of every 10,000 calls in September 2026, up from 0.7 per 10,000 in October to December 2025. See the data
  • Of the AI callers, 36% were following up on sales or leads, 32% were doing business tasks, 26% were collections and 6% were scams. See the data
  • Arizona callers say thank you most (46.9% of conversations), Texas callers least (27.3%). See the data
  • Christmas Day brings 56 real calls for every 100 on a normal day. Super Bowl Sunday brings 115. See the data
  • In NextPhone's 2026 call data, 8.2% of all calls were spam: 9.0% of calls to personal phones and 4.2% of calls to business lines. See the data
  • Spam peaks between 6 and 7pm, at 9.6 to 9.7% of calls, and falls to 1 to 2% between midnight and 3am. See the data
  • Sunday gets the least spam: 5.2% of calls, against 8.3 to 8.7% on weekdays. See the data
  • On the personal phones NextPhone covered all year, spam's share of calls rose by more than half: the Spam Index hit 155 in September 2026 (October 2025 = 100). See the data
  • 37% of spam calls with speech pitch loans or business funding. See the data
  • Only 17% of spam calls show the receiver's own area code, against 45% of real calls, and 1 in 4 spam calls (26.9%) shows a toll-free number. See the data
  • Calls from area code 607 in upstate New York are 47.2% spam, the highest of any area code in the data. See the data

Looking for market size, adoption and cost numbers too? See our AI receptionist statistics roundup.

How to read the numbers

  • Percentages are shares. Either a share of all calls, or a share of conversations: calls where the caller actually spoke and that weren't spam. Each finding says which.
  • Indexes show change. An index compares a month with October 2025, which is set to 100. Above 100 means more than in October 2025, below 100 means less.
  • Holidays work the same way. 100 is a normal day for that weekday. Below 100 is quieter than usual, above 100 is busier.
  • AI callers are counted per 10,000 calls, because they are still rare. We read every call flagged as an AI caller to check it.
  • Trends use the same lines. Business lines joined during the year, so trends that could be skewed by that are measured on the personal phones we covered all 12 months.
  • Hours are local. Every hour of the day is the local time of the line that was called.
  • Transcription improved in August 2026. Measures built on what callers say can jump that month for that reason alone.

When people call

1.1 in 7 business calls comes after hours, and those calls convert better

14.1% of calls to small businesses arrive after hours: 7.4% on weekends and 6.6% on weekday evenings and early mornings (before 8am, after 6pm).

Those calls are worth more than their share. 23.6% of after-hours calls became a lead or a booking, against 18.6% of calls in business hours, so 17.2% of all leads and bookings arrived after hours.

Real calls to business lines that became a lead or booking
  • After hours23.6%
  • Business hours18.6%

Night calls are more urgent too: mentions of an emergency, leak, flood or no heat are 44% more common at night (2.6% against 1.8% of business conversations). The busiest hour of the week is Monday at 11am. More on covering those hours: after-hours answering.

2.Christmas is the quietest holiday on the phone. Super Bowl Sunday is busier than a normal Sunday

Christmas Day brings 56 real calls for every 100 on a normal day of the same weekday. Super Bowl Sunday brings 115.

Christmas Eve and Labor Day both sit at 65, Memorial Day at 72, Thanksgiving at 75 and New Year's Day at 78. July 4 (93) barely dents call volume, and Black Friday (100) is a normal day. Super Bowl Sunday was the one holiday we measured where the phone rang more, not less.

Real calls per line on holidays
  • Super Bowl Sunday115
  • Black Friday100
  • July 493
  • New Year's Day78
  • Thanksgiving75
  • Memorial Day72
  • Labor Day65
  • Christmas Eve65
  • Christmas Day56

100 = a normal day (the same weekday nearby)

More from the data

  • 4 in 10 calls are someone calling back within a week (42 to 43% on both kinds of lines).
  • The typical business conversation lasts 77 seconds (median), and 90% end in under 3.5 minutes (last 90 days).
  • Snowbirds are real: Northeast and Midwest numbers are 12.3% of calls to Arizona lines in winter, against 8.4% in summer.
One in seven calls to a small business comes when nobody's at the desk. Those aren't spam. They're the Saturday price question and the leak at 10pm.
Yan MellataYan Mellata, founder of NextPhone

How people talk to an AI receptionist

3.1 in 60 business callers asks "are you a robot?"

21.9% of the callers who ask still become a lead or a booking.

Only 1 in 60 business conversations includes the question (1.6%). Across all business conversations, the callers who asked became a lead or a booking slightly more often than the ones who didn't: 21.9% against 19.2%.

Became a lead or booking, business lines
  • Asked "are you a robot?"21.9%
  • Did not ask19.2%

That gap comes from call length: callers who ask also tend to talk longer. Compared within calls of the same length, askers converted less in short and medium calls and more in calls where the caller said over 100 words (33.0% against 20.8%). So the data doesn't show the question helping or hurting. It shows that plenty of callers ask, get an answer and keep going.

Became a lead or booking, by how much the caller said

Short calls (3 to 20 words from the caller)

  • Asked6.0%
  • Did not ask12.2%

21 to 50 words

  • Asked20.1%
  • Did not ask21.9%

51 to 100 words

  • Asked21.6%
  • Did not ask27.1%

Long calls (over 100 words)

  • Asked33.0%
  • Did not ask20.8%

4.1 in 3 callers thank the AI

1 in 3 conversations include “thank you”.

Manners follow the clock. 37% of callers say thank you at 8am, against 21% at 1am. Weekends are a little less polite too: 31% of weekend callers say thank you, against 35% on weekdays.

Conversations that include "thank you", by hour of day
Show the numbers
Conversations that include "thank you", by hour of day, by hour of day (local time of the line called)
HourShareHourShare
12am24.0%12pm34.6%
1am20.9%1pm34.7%
2am23.5%2pm35.1%
3am23.1%3pm34.8%
4am26.8%4pm34.9%
5am20.9%5pm33.7%
6am29.4%6pm31.8%
7am33.0%7pm29.3%
8am37.0%8pm31.8%
9am36.3%9pm28.2%
10am36.0%10pm22.8%
11am35.7%11pm28.0%

5.Saturday is price-shopping day

1 in 15 business conversations asks about price. On Saturdays it's 8.7%, against about 6% on weekdays.

That's a 40% jump, on the day many offices are closed or short-staffed.

Business conversations that ask about price
  • Saturday8.7%
  • Weekdaysabout 6%

6.After midnight, callers swear 4 to 5 times more

5.4 to 6.5% of conversations between midnight and 4am include a swear word, against 1.1 to 1.3% between 9 and 11am.

Weekends run hotter too: 2.8% of Saturday and Sunday conversations include swearing, against 1.4 to 1.7% on weekdays. Mondays, despite their reputation, are no worse than any other weekday.

Conversations with a swear word, by hour of day
Show the numbers
Conversations with a swear word, by hour of day, by hour of day (local time of the line called)
HourShareHourShare
12am5.4%12pm1.3%
1am5.6%1pm1.5%
2am6.0%2pm1.6%
3am6.5%3pm1.6%
4am4.4%4pm1.5%
5am1.5%5pm1.9%
6am2.1%6pm2.2%
7am1.8%7pm2.9%
8am1.2%8pm3.8%
9am1.1%9pm4.4%
10am1.3%10pm4.0%
11am1.2%11pm5.9%

7.The most polite and rudest states on the phone

Arizona callers say thank you most: 46.9% of conversations, against 27.3% for Texas callers.

If saying thank you is the test, Texas, Wisconsin and Tennessee callers are the rudest. Swearing has its own map: South Carolina (3.21%) and Missouri (3.05%) callers swear most, Alabama and Oklahoma (1.00%) least. Oklahoma is in the top 10 for thank-yous and tied for the least swearing.

Where callers say "thank you"
Least polite (27.3%)Most polite (46.9%)Not enough calls to compare
Where callers swear
Least swearing (1.00%)Most swearing (3.21%)Not enough calls to compare
Every state, ranked
Conversations that include "thank you", by caller state
  • Arizona46.9%
  • Michigan42.4%
  • Washington39.8%
  • Massachusetts39.2%
  • Illinois38.0%
  • Virginia36.7%
  • California35.2%
  • New York35.0%
  • Maryland34.8%
  • Oklahoma34.8%
  • Pennsylvania34.5%
  • Indiana33.9%
  • South Carolina33.7%
  • Georgia33.3%
  • New Jersey33.2%
  • North Carolina32.3%
  • Alabama31.5%
  • Louisiana31.3%
  • Florida31.0%
  • Missouri30.9%
  • Ohio29.1%
  • Tennessee28.4%
  • Wisconsin28.4%
  • Texas27.3%

All 24 states with 300 or more callers and 2,000 or more conversations, by caller area code.

Conversations with a swear word, by caller state
  • South Carolina3.21%
  • Missouri3.05%
  • Ohio2.52%
  • California2.39%
  • Illinois2.08%
  • Pennsylvania2.00%
  • Tennessee1.83%
  • Michigan1.80%
  • Virginia1.80%
  • Washington1.76%
  • Arizona1.72%
  • Florida1.70%
  • Louisiana1.67%
  • North Carolina1.53%
  • New York1.45%
  • Massachusetts1.43%
  • New Jersey1.38%
  • Wisconsin1.38%
  • Maryland1.34%
  • Texas1.32%
  • Georgia1.25%
  • Indiana1.24%
  • Alabama1.00%
  • Oklahoma1.00%

AI callers

8.AI voice calls: AI is now calling AI receptionists

AI voice agents made 4.0 of every 10,000 calls in September 2026. Over the year the rate rose from 0.7 per 10,000 in October to December 2025 to 2.7 in July to September 2026.

These are AI voice agents placing calls that an AI receptionist answers. Businesses see slightly more of them than people do: 4.9 per 10,000 calls on business lines against 3.4 on personal phones in September 2026. Part of the rise comes from business lines joining during the year. On the personal phones covered all 12 months, the rate rose more slowly, from 1.2 to 1.8 per 10,000 calls.

We read every call our phrase match flagged and removed the ones where a person was only talking about AI. AI callers that don't say what they are aren't counted, so these figures are a floor.

AI callers per 10,000 calls, by monthAll linesSep 2026: 4.0
01.534.56Oct 25NovDecJan 26FebMarAprMayJunJulAugSep
Show the numbers
MonthAll lines
Oct 20250.4
Nov 20250.4
Dec 20251.2
Jan 20262.5
Feb 20261.9
Mar 20262.0
Apr 20262.1
May 20262.0
Jun 20261.1
Jul 20261.8
Aug 20262.3
Sep 20264.0
AI callers per 10,000 calls, by quarter

All lines

  • Oct to Dec 20250.7
  • Jan to Mar 20262.1
  • Apr to Jun 20261.7
  • Jul to Sep 20262.7

Personal phones covered all year

  • Oct to Dec 20251.2
  • Jan to Mar 20261.3
  • Apr to Jun 20261.3
  • Jul to Sep 20261.8

9.What AI callers want

36% of AI callers were following up on sales or leads, 32% were doing business tasks, 26% were collections and 6% were scams.

Mostly what a person on the phone would. On business lines, three in four AI callers were doing a job, most of them a car auction company's AI checking whether a vehicle was ready for release at an auto shop. On personal phones, they were mostly sales follow-ups and collections.

What the AI callers wanted
  • Sales and lead follow-up36%
  • Business tasks32%
  • Collections and account notices26%
  • Scams and impersonation6%

Real examples from the calls:

  • A debt collector's AI that opened with "this call is being recorded... I'm an AI assistant"
  • An AI posing as an international shipping carrier about unpaid import duties
  • An AI selling life insurance to veterans
  • A car auction company's AI checking a vehicle's release status with auto shops
  • AI sales assistants following up on loan, mortgage and real estate inquiries

Some are legitimate businesses automating their outbound calls. Some are scams with a better voice.

The robocall isn't the story anymore. The new caller is another AI, and it's polite. Small businesses will need to tell a legitimate AI from a scam the same way they learned to spot a fake 800 number.
Yan MellataYan Mellata, founder of NextPhone

Spam calls and robocalls

Spam grew this year. On the personal phones NextPhone covered all 12 months, spam's share of calls rose by more than half: the Spam Index reached 200 in August 2026 and stood at 155 in September (October 2025 = 100). Recorded robocalls did not grow with it (Robocall Index 104). The growth came from spam that isn't a recorded message.

Spam Index, personal phones covered all year (October 2025 = 100)Personal phones covered all yearSep 2026: 155
075150225300Oct 25NovDecJan 26FebMarAprMayJunJulAugSep
Show the numbers
MonthPersonal phones covered all year
Oct 2025100
Nov 202597
Dec 202591
Jan 2026115
Feb 2026133
Mar 2026151
Apr 2026150
May 2026165
Jun 2026181
Jul 2026184
Aug 2026200
Sep 2026155

10.What time do spam calls come? All day, peaking at 6 to 7pm

Spam is 7.8 to 9.7% of calls between 9am and 7pm, highest between 6 and 7pm, and 1 to 2% between midnight and 3am.

Spammers keep the hours of the people they're trying to reach, and they ease off on Sundays. Personal phones take the worst of it, with twice the spam of business lines (9.0% vs 4.2% of calls). Across all calls, 8.2% were spam.

Calls tagged as spam, by hour of day
Show the numbers
Calls tagged as spam, by hour of day, by hour of day (local time of the line called)
HourShareHourShare
12am1.2%12pm8.9%
1am1.9%1pm8.9%
2am1.6%2pm8.5%
3am2.4%3pm8.3%
4am2.8%4pm7.8%
5am8.5%5pm9.1%
6am4.8%6pm9.6%
7am3.8%7pm9.7%
8am7.1%8pm7.6%
9am8.4%9pm4.0%
10am9.0%10pm2.6%
11am8.9%11pm1.5%
Calls tagged as spam, by day of the week
  • Monday8.3%
  • Tuesday8.6%
  • Wednesday8.5%
  • Thursday8.6%
  • Friday8.7%
  • Saturday7.2%
  • Sunday5.2%

11.The most common spam calls: loans, Google listings and the IRS

37% of spam calls with speech are pitches for loans and business funding.

Nothing else comes close. Next are callers claiming a business's Google listing or business profile needs attention (7.4%) and callers claiming to be the government, the IRS or Social Security (7.2%). Both work by sounding official and urgent.

Share of spam calls with speech, by script
  • Loans and business funding37%
  • Google listing or business profile7.4%
  • Government, IRS or Social Security7.2%
  • Debt and credit5.2%
  • Health insurance or Medicare3.9%
  • Marketing and SEO3.0%
  • Package or Amazon1.6%
  • Political calls and surveys1.4%

Scripts overlap and not every spam call matches one, so shares do not add up to 100%.

12.Phone spoofing: spam rarely shows your area code anymore

Only 17% of spam calls show the receiver's own area code, against 45% of real calls.

Neighbor spoofing, where a spam call fakes a local number, is uncommon in this data and kept falling through the year: 18.1% of spam calls showed the receiver's area code in October 2025, 11.8% in September 2026. Real calls stayed between 36 and 46%. Toll-free numbers are the bigger tell: 1 in 4 spam calls (26.9%) shows one, against 5.2% of real calls.

What the caller ID shows

Same area code as the person called

  • Spam calls17%
  • Real calls45%

Toll-free number

  • Spam calls26.9%
  • Real calls5.2%

13.The spammiest area codes: upstate New York and Kentucky

Calls from area code 607 in upstate New York are 47.2% spam, against 8.2% across all calls.

Upstate New York holds five of the top 10 area codes: 607, 518 (34.0%), 585 (30.5%), 315 (26.9%) and 716 (25.6%). By state, Kentucky stands out: 19.1% of calls from Kentucky area codes are spam, led by 270 (23.4%) and 502 (23.0%), and the spam comes from many different numbers, not one bad actor. West Virginia's 304 also over-indexes. Area codes show where a number was issued, not where the caller sits, so this points to numbers registered there, not to the people who live there.

Calls tagged as spam, by caller area code: the top 11
  • 607 (New York)47.2%
  • 229 (Georgia)39.9%
  • 518 (New York)34.0%
  • 769 (Mississippi)32.9%
  • 585 (New York)30.5%
  • 272 (Pennsylvania)27.5%
  • 315 (New York)26.9%
  • 856 (New Jersey)25.8%
  • 716 (New York)25.6%
  • 270 (Kentucky)23.4%
  • 502 (Kentucky)23.0%

Area codes with 50 or more distinct calling numbers and 500 or more calls. 290 area codes qualify.

Calls tagged as spam, Kentucky area codes vs all calls
  • Calls from Kentucky area codes19.1%
  • All calls8.2%

How to stop spam calls: what a year of calls shows

  1. Be wary of toll-free numbers you don't know. 1 in 4 spam calls (26.9%) shows one, against 5.2% of real calls.
  2. Trust a local number a little more, not completely. Only 17% of spam calls show your own area code, against 45% of real calls. A local number can still be faked.
  3. Hang up on loan and funding offers from cold calls. They are 37% of spam calls with speech.
  4. Ignore anyone offering to fix your Google listing. It is the second most common spam script. Manage your business profile by logging in yourself.
  5. Screen unknown calls in the early evening. Spam peaks between 6 and 7pm.
  6. Ask an AI caller who it works for. Then call back on a number you look up yourself. 6% of the AI callers we checked were scams.
  7. Report it and register. Report scam calls to the FTC at ReportFraud.ftc.gov, add your number to the National Do Not Call Registry, and see the FTC's guide to blocking unwanted calls.

What this means for small businesses

  • Cover the hours you're closed. 1 in 7 calls arrives after hours, and those callers are more likely to become a lead or a booking.
  • Have prices ready for Saturday. 8.7% of Saturday conversations ask about price, against about 6% on weekdays.
  • Treat AI callers as callers, then verify them. Some are a vendor or a partner's system checking on real work. Some are scams.
  • Don't trust urgent calls about your listing or your taxes. Google listing and government pitches are two of the three most common spam scripts.

Questions about calls, AI and spam

How many calls to small businesses come after hours?

In NextPhone's data, 1 in 7 (14.1%): 7.4% on weekends and 6.6% on weekday evenings and early mornings. Those calls are worth more: 23.6% became a lead or booking, against 18.6% during business hours, and 17.2% of all leads and bookings arrived after hours.

When do customers call businesses the most?

In NextPhone's data, the busiest hour of the week for a small business is Monday at 11am. 1 in 7 calls (14.1%) arrives after hours: 7.4% on weekends and 6.6% on weekday evenings and early mornings. Saturday brings the most price questions (8.7% of business conversations, against about 6% on weekdays), and Christmas Day is the quietest holiday, at 56 real calls for every 100 on a normal day.

What do callers ask an AI receptionist?

In NextPhone's data, 1 in 15 business conversations asks about price, rising to 8.7% on Saturdays. Mentions of an emergency, leak, flood or no heat come up in 1.8% of business conversations during the day and 2.6% at night. Only 1 in 60 business callers asks whether they are talking to a robot, and 1 in 3 conversations include "thank you".

Do callers mind talking to an AI receptionist?

Most never ask. Only 1 in 60 business callers asks whether they are talking to a robot, and 21.9% of those who ask still become a lead or a booking. Compared within calls of the same length, askers converted less in short calls and more in calls over 100 words (33.0% vs 20.8%).

Are AI voice agents calling people and businesses now?

Yes. In NextPhone's data, callers that said they were an AI made 4.0 of every 10,000 calls in September 2026, up from 0.7 per 10,000 in October to December 2025. 36% were following up on sales or leads, 32% were doing business tasks like checking the status of a job, 26% were collections and account notices, and 6% were scams.

How can you tell if a call is from an AI?

Ask whether you are talking to an AI and who it works for. Some AI callers say it up front: in NextPhone's data, a debt collector's AI opened with "this call is being recorded... I'm an AI assistant". Then call back on a number you look up yourself, since 6% of the AI callers NextPhone checked were scams.

Why am I getting so many spam calls?

Call data can't say why one number gets targeted, but it shows how spam behaves. In NextPhone's 2026 call data, 8.2% of all calls were spam, personal phones got about twice the spam of business lines (9.0% vs 4.2%), and spam peaked between 6 and 7pm, when people are likely to pick up. 37% of spam calls with speech pitched loans or business funding.

Why am I suddenly getting more spam calls?

Spam comes in waves. On the personal phones NextPhone covered all year, spam's share of calls doubled between October 2025 and August 2026 (Spam Index 200, October 2025 = 100) before easing to 155 in September.

What time do spam calls come?

All day, with a peak in the early evening. In NextPhone's data, spam is 7.8 to 9.7% of calls between 9am and 7pm, highest between 6 and 7pm, and 1 to 2% between midnight and 3am. Sunday has the least spam (5.2% of calls).

What percentage of phone calls are spam?

8.2% of the calls answered by NextPhone's AI were tagged as spam: 9.0% of calls to personal phones and 4.2% of calls to business lines. These are calls that reached the AI after any carrier spam blocking, so it is a share of this set of calls, not an estimate for every phone in the US.

What are the most common spam and scam calls?

In NextPhone's data, 37% of spam calls with speech pitched loans or business funding. Next came calls about a Google listing or business profile (7.4%), callers claiming to be the government, the IRS or Social Security (7.2%), and debt and credit offers (5.2%).

Is a call about my Google business listing a scam?

Treat a cold call offering to fix or verify your listing with suspicion. Google listing and business profile pitches were the second most common spam script in NextPhone's data (7.4% of spam calls with speech). Manage your business profile by logging in yourself.

What is phone spoofing, and is neighbor spoofing still common?

Spoofing is when a call shows a fake caller ID. Neighbor spoofing shows a number with your own area code so the call looks local. In NextPhone's data it is uncommon and falling: 17% of spam calls showed the receiver's area code, against 45% of real calls, and the share fell from 18.1% in October 2025 to 11.8% in September 2026. 1 in 4 spam calls showed a toll-free number instead.

Is a local number calling me more likely to be real?

It is a better sign than a toll-free number. In NextPhone's data, 45% of real calls came from the receiver's own area code, against 17% of spam calls, while 26.9% of spam calls showed a toll-free number against 5.2% of real calls. A local number can still be faked.

What is the most polite state?

Arizona. In NextPhone's 2026 call data, 46.9% of conversations from Arizona callers included "thank you", the highest of the 24 states with enough calls to compare. Michigan (42.4%) and Washington (39.8%) come next.

Which state is the rudest on the phone?

Measured by who says thank you, Texas callers are the least polite (27.3% of conversations), followed by Wisconsin and Tennessee (28.4%). Measured by swearing, South Carolina (3.21%) and Missouri (3.05%) lead.

How we measured

  • The calls are a subset. This report does not cover every call NextPhone has answered. It uses a subset: inbound calls answered by AI from December 2024 to October 2026 that have a full call record, on personal phones (people using the TTYL app) and small-business lines (NextPhone customers, mainly law firms, home services and local services). Only calls the AI handled are counted, after any spam blocking by the phone carrier.
  • The window. Monthly figures cover the 12 months from October 2025 to September 2026. One-number findings (shares by line type, hour, weekday, area code and state) are measured across the full dataset. Call length uses business lines over the last 90 days.
  • Same lines. Business lines start in November 2025. To keep trends from reflecting that mix, the Spam Index and the same-lines AI caller rates use only the personal phones with calls in every month of the window.
  • Transcription. Transcription was upgraded during August 2026. Phrase-based measures from August onward are measured on the new transcripts and are not fully comparable with earlier months.
  • Calls and conversations. “Calls” counts every inbound call. “Conversations” counts calls that were not spam and where the caller said at least three words.
  • Phrases. Thank-yous, price questions, swearing, robot questions and emergencies are matched on what the caller said, in the call transcript. A caller who paraphrases may not be counted, so these are floors, not exact counts.
  • AI callers. A call counts when the calling voice says it is an AI or a virtual assistant. We read every flagged call and removed people who only mentioned AI and other businesses' AI receptionists picking up. AI callers that don't say what they are are not counted, so this is a floor.
  • Robot question by length. Calls are grouped by how many words the caller said (3 to 20, 21 to 50, 51 to 100, over 100) and compared within each group.
  • Spam. Calls tagged as spam. Script shares are a share of spam calls that contained speech. Area codes need 50 or more distinct calling numbers and 500 or more calls to be ranked.
  • After hours. Before 8am or after 6pm on weekdays, plus all of Saturday and Sunday, in the local time of the line called. Spam and calls where the caller never engaged are left out. For emergency mentions, night means 8pm to 7am.
  • Holidays. Real calls per line on the holiday against the same weekday nearby, where 100 = normal.
  • Location. Caller and receiver locations come from area codes, which show where a number was issued, not where the person is. States are ranked when they have 300 or more callers and 2,000 or more conversations.
  • Leads and bookings. Outcomes are from business lines only.
  • Other NextPhone numbers. Other pages on this site cite a 28.5% after-hours share from an earlier analysis of business calls. Some cite 35% or more of leads arriving after hours, a law-firm figure from outside research. This report is a separate study: only calls the AI handled, weekends counted as after hours and an 8am to 6pm business day, so its shares differ.
  • Privacy. Every figure is an aggregate. No transcript, caller or business is identified, and the companies behind the AI callers are not named.

How to cite and download the data

Journalists, researchers and AI assistants are welcome to quote the findings. Please cite:

NextPhone (2026). The Comprehensive 2026 AI Receptionist Call Report. https://www.getnextphone.com/reports/ai-receptionist-call-report-2026

Download the data (CSV): every table and chart on this page as rates and indexes, with no call-level data. Free to reuse with attribution under CC BY 4.0.

About NextPhone. NextPhone answers the phone for small businesses with AI: it answers calls, takes the details and books appointments, day or night. TTYL, from the same team, answers calls on personal phones. getnextphone.com