The 2026 chatbot numbers that matter most
This page collects 55 chatbot statistics, and every one links to the report, filing or survey it came from. We rebuilt it in November 2026 because most statistics pages online, including the earlier version of this one, repeat figures that cannot be traced to a source. Where a number is a forecast, a vendor's own customer data or an older survey, the table says so.
| Statistic | What it says | Source |
|---|---|---|
| 87% | of customers say a company using GenAI for service must offer a way to reach a human | Gartner, Aug 2026 |
| 66% | of service organizations use agentic AI in 2026, up from 39% a year earlier | Salesforce, 2026 |
| 83.1% | average AI agent deflection rate in Freshworks' 2026 customer base | Freshworks, 2026 |
| 20% | of service leaders have cut agent headcount because of AI | Gartner, Dec 2025 |
| 1 billion+ | active threads with business accounts on Meta's apps every day | Meta, Oct 2025 |
| $9.6B | chatbot market in 2025, projected at $41.2B by 2033 | Grand View Research |
The short version: adoption is no longer the story. Most service teams already run some form of AI, and the gap now is between teams that deploy it well and teams that bolt it on. Customers are fine with bots that solve the problem and furious with bots that block the way to a person. If you are planning a bot this year, the numbers below point to the same three design rules: answer from real knowledge, hand off cleanly, and be honest that it is a bot. Our chatbot best practices guide turns those rules into a checklist.
Download the data (CSV) with every statistic, its source and the source date. You are welcome to reuse it with a link back to this page.
How many people use AI chat tools
Before you ask whether customers will use your chatbot, it helps to know how normal talking to software has become. The best public measures come from the Stanford AI Index and Pew Research Center, both of which publish their methods.
| Statistic | What it measures | Source |
|---|---|---|
| 53% | Population adoption of generative AI within three years of launch, faster than the PC or the internet | Stanford HAI 2026 AI Index, Apr 2026 |
| 88% | Organizations using AI, per the AI Index's organizational adoption measure | Stanford HAI 2026 AI Index, Apr 2026 |
| 4 in 5 | University students who now use generative AI | Stanford HAI 2026 AI Index, Apr 2026 |
| 34% | U.S. adults who have ever used ChatGPT (58% of adults under 30) | Pew Research Center, Jun 2025 |
| 28% | Employed U.S. adults who have used ChatGPT at work | Pew Research Center, Jun 2025 |
| 1 billion+ | Monthly actives using Meta AI across Meta's apps | Meta Q3 2025 earnings call, Oct 2025 |
| $172 billion | Estimated annual value of generative AI tools to U.S. consumers by early 2026 | Stanford HAI 2026 AI Index, Apr 2026 |
Two things stand out. First, generative AI reached majority adoption faster than the PC or the internet did, according to the AI Index, although the rate varies widely by country: the report puts the United States 24th at 28.3%, behind Singapore (61%) and the United Arab Emirates (64%). Second, the habit is forming young. A 58% majority of U.S. adults under 30 have used ChatGPT, which means the customers you will serve over the next decade expect to type a question in plain language and get an answer back.
That expectation is the real competitor for a business chatbot. A button-only menu that cannot understand "where is my order" now feels broken next to the assistants people use every day. It is why most new bots are built on a large language model that answers from the company's own knowledge base, with buttons kept for the moments where a structured choice is faster.
What customers actually think about chatbots
Customer sentiment looks contradictory until you read the questions carefully. People like AI that saves them time and dislike AI that stands between them and a person.
| Statistic | What it measures | Source |
|---|---|---|
| 87% | Customers who say a company using GenAI for service must offer a way to reach a human | Gartner survey of 3,566 customers, Aug 2026 |
| 50% | Customers who say interactions are easier when companies use GenAI | Gartner survey of 3,566 customers, Aug 2026 |
| 58% | GenAI users who have had it complete a task on their behalf (74% in B2B) | Gartner survey of 3,566 customers, Aug 2026 |
| 64% | Customers who would prefer companies did not use AI for service (survey fielded Dec 2023) | Gartner survey of 5,728 customers, Jul 2024 |
| 53% | Customers who would consider switching if a company used AI for service (same 2023 survey) | Gartner survey of 5,728 customers, Jul 2024 |
| 30% | Consumers who would work with an AI agent instead of a person for faster service (37% of Gen Z and millennials) | Salesforce State of the AI Connected Customer, Oct 2024 |
| Nearly 75% | Consumers who want to know when they are talking to an AI agent | Salesforce State of the AI Connected Customer, Oct 2024 |
| 45% | Consumers more likely to use an AI agent when there is a clear escalation path | Salesforce State of the AI Connected Customer, Oct 2024 |
| 74% | Consumers who now expect customer service to be available 24/7 | Zendesk CX Trends 2026, Nov 2025 |
| 74% | Consumers frustrated when they have to repeat information | Zendesk CX Trends 2026, Nov 2025 |
| 95% | Consumers who expect an explanation for AI-made decisions | Zendesk CX Trends 2026, Nov 2025 |
Gartner's two surveys, eighteen months apart, show the shift. In a survey fielded in December 2023, 64% of customers said they would prefer companies not use AI for service, and their top concern was that it would become harder to reach a person. By early 2026, half said GenAI made interactions easier, and 58% of GenAI users had let it complete a task for them. The condition attached to that acceptance is consistent: 87% say a human option is essential.
Salesforce's consumer research adds the transparency angle. Nearly three in four consumers want to know when they are talking to an AI agent, and 45% are more likely to use one when there is a clear escalation path. If you are deciding how your bot should introduce itself, our guide to AI chatbot disclosure laws covers where saying so is now a legal requirement, not just good manners.
The practical takeaway: design the human handoff first, then automate around it. A bot that answers 60% of questions and passes the rest to a person with full context will score better than one that tries to answer everything. Conferbot's live chat handover hands the full conversation to an agent in the same inbox, so the customer never has to repeat themselves, which Zendesk found frustrates 74% of consumers.
How many businesses use chatbots and AI agents
The adoption question has largely been answered. The more useful numbers now are how deep the deployment goes and how much work it actually takes on.
| Statistic | What it measures | Source |
|---|---|---|
| 85% | Customer service leaders who planned to explore or pilot customer-facing conversational GenAI in 2025 | Gartner survey of 187 service leaders, Dec 2024 |
| 66% | Customer service organizations using agentic AI in 2026, up from 39% in 2025 | Salesforce State of Service: AI Agents Edition (3,075 pros), 2026 |
| 85% | Service organizations using at least one form of AI | Salesforce State of Service: AI Agents Edition (3,075 pros), 2026 |
| 30% to 50% | Share of service cases handled by AI today, and the share teams project for 2027 | Salesforce State of Service, 7th ed. (6,500 pros), 2025 |
| 82% / 87% | Senior support leaders who invested in AI for service in 2025 / plan to in 2026 | Intercom 2026 Customer Service Transformation Report (2,400+ pros), 2026 |
| 10% | Support teams that have reached mature, at-scale AI deployment | Intercom 2026 Customer Service Transformation Report (2,400+ pros), 2026 |
| 1 million+ | Businesses using Meta Business Agents to talk to customers or complete sales every week | Meta Q2 2026 earnings call, Jul 2026 |
| 1 billion+ | Active threads people have with business accounts on Meta's messaging apps every day | Meta Q3 2025 earnings call, Oct 2025 |
Salesforce's two State of Service editions show the speed of change: 39% of service organizations used agentic AI in 2025 and 66% did in 2026. Intercom's 2026 survey of more than 2,400 support professionals shows the depth problem. Most senior leaders invested in AI, but only 10% of teams describe their deployment as mature, meaning fully integrated and working at scale.
Messaging apps are where much of this volume lands. Meta reported more than a billion active threads between people and business accounts every day on its messaging apps, and more than a million businesses using its Business Agents every week by mid-2026. For a small business that means the chatbot question is often a channel question: which of WhatsApp, Instagram, Messenger, Telegram or the website widget do your customers already use? Our chatbot channel strategy guide walks through that choice.
Chatbot results: resolution, speed and satisfaction
Outcome data is the hardest to trust, because most of it comes from vendors measuring their own customers. We include it because it is the only large-scale data available, but read each row with its source in mind.
| Statistic | What it measures | Source |
|---|---|---|
| 83.1% | Average AI agent deflection rate across industries in 2026 (Freshworks customer base) | Freshworks 2026 Customer Service Benchmark, 2026 |
| 9m 21s to 4m 36s | Average resolution time when AI agents handle 0 to 25% versus 50 to 100% of conversations | Freshworks 2026 Customer Service Benchmark, 2026 |
| 70% | Service organizations with AI agents that see measurable value within 60 days | Salesforce State of Service: AI Agents Edition (3,075 pros), 2026 |
| #1 | Customer satisfaction ranked as the most improved KPI after deploying AI agents | Salesforce State of Service: AI Agents Edition (3,075 pros), 2026 |
| 62% / 87% | Teams reporting improved service metrics after AI, all teams / mature deployments | Intercom 2026 Customer Service Transformation Report (2,400+ pros), 2026 |
| 2.3 million | Conversations Klarna's AI assistant handled in its first month, two-thirds of its service chats | Klarna press release, Feb 2024 |
| Under 2 min | Klarna's time to resolve an errand with the assistant, down from 11 minutes | Klarna press release, Feb 2024 |
| 25% | Drop in repeat inquiries Klarna reported after launch | Klarna press release, Feb 2024 |
| 64.7% vs 64.2% | CSAT on chatbot-handled chats versus all rated chats on LiveChat | LiveChat Customer Service Report (2024 data), 2024 |
| 85% | Conversations Movida's WhatsApp business agent resolved without a human | Meta Q2 2026 earnings call, Jul 2026 |
Freshworks' 2026 benchmark, drawn from more than 33,000 customer accounts, is the clearest picture of the link between automation share and speed: resolution time roughly halves as AI agents handle more of the conversation. The deflection rates are high because they measure conversations that never reached a human, which includes customers who gave up. Pair any deflection target with a satisfaction measure.
The LiveChat figure is a useful counterweight. Across 1.67 billion chats, bot-handled conversations scored 64.7% satisfaction against 64.2% overall: about the same, not dramatically better. Well-built bots match people on routine questions. They do not magically delight customers. Our containment rate benchmarks and chatbot KPI guide explain which numbers to track so you see both sides, and the deflection rate and CSAT glossary entries define them precisely.
To estimate what these rates would mean for your own volume, the free chatbot cost calculator runs the arithmetic with your numbers rather than someone else's case study, and our guide to calculating chatbot ROI shows the formula.
Chatbots and customer service jobs
The fear that chatbots replace support teams is the most repeated claim and the least supported by data so far.
| Statistic | What it measures | Source |
|---|---|---|
| 20% | Customer service leaders who have reduced agent staffing because of AI | Gartner survey of 321 service leaders, Dec 2025 |
| 55% | Leaders reporting stable staffing while handling higher customer volumes | Gartner survey of 321 service leaders, Dec 2025 |
| 42% | Organizations hiring AI-focused roles such as conversational AI designers | Gartner survey of 321 service leaders, Dec 2025 |
| 85% | Service leaders expanding human agent responsibilities as AI absorbs volume | Gartner survey of 321 service leaders, Apr 2026 |
| 50% by 2027 | Companies that cut service staff due to AI that will rehire for similar work | Gartner prediction, Feb 2026 |
| 95% | Service leaders who plan to keep human agents to define AI's role | Gartner poll of 163 service leaders, Jun 2025 |
Gartner's October 2025 survey of 321 leaders found that only 20% had cut agent headcount because of AI, while 55% held staffing steady and handled more volume. A follow-up found 85% are expanding what human agents do, and Gartner predicts half of the companies that cut staff because of AI will rehire for similar work under new titles by 2027. The pattern is redeployment: bots take the repetitive questions, people take the complex ones, and new roles appear around designing and tuning the bot.
For a small team, that is the realistic goal. A bot that covers nights and weekends and answers the twenty most common questions lets two people do the work of four without anyone being replaced. See after-hours support with a chatbot for how teams set that up.
Chatbot forecasts for 2027 to 2029
Forecasts are opinions with a method, not facts. These come from analysts who publish their predictions and revisit them, which is the best you can do.
| Statistic | What it measures | Source |
|---|---|---|
| 80% by 2029 | Common customer service issues agentic AI will resolve without a human, with 30% lower operating costs | Gartner prediction, Mar 2025 |
| 60% by 2028 | Brands that will use agentic AI for one-to-one interactions | Gartner prediction, Jan 2026 |
| 50%+ by 2028 | Customer service organizations that will double technology spend | Gartner prediction, Mar 2026 |
The common thread is agentic AI: bots that do things, not just answer. Booking an appointment, changing an order and taking a payment inside the conversation are what separate the next generation from FAQ bots. Our explainer on what an AI agent is and the comparison of AI agents and chatbots cover where the line sits today, and AI chatbot trends for 2026 goes deeper on what is changing this year.
Chatbot market size
Market sizing is the least comparable data in this field, because every firm defines the market differently. The figures below come from two research firms' public summaries and Gartner's spending forecast. Do not add them together.
| Statistic | What it measures | Source |
|---|---|---|
| $9.6 billion | Global chatbot market size in 2025 | Grand View Research, 2026 |
| $41.2 billion by 2033 | Projected chatbot market, 19.6% CAGR from 2026 | Grand View Research, 2026 |
| 31.3% | North America's share of the chatbot market in 2025 | Grand View Research, 2026 |
| $17.05 billion | Conversational AI market in 2025, projected to reach $49.80 billion by 2031 | MarketsandMarkets, 2025 |
| $29.2 billion | Worldwide spending on AI agents and assistants forecast for 2026, up from $16.5 billion | Gartner forecast, Sep 2026 |
Grand View Research sizes the chatbot market at $9.6 billion in 2025 with 19.6% annual growth to 2033. MarketsandMarkets sizes the broader conversational AI market, which includes voice assistants and platforms, at $17.05 billion. Gartner's forecast counts spending on AI agents and assistants at $29.2 billion in 2026. Three definitions, three numbers, one direction.
For the investment and spending side in detail, including total AI spending and where it goes, see our companion page of conversational AI market statistics.
Where chatbots run: channel reach
A chatbot is only as useful as the channel it lives on. These are the audience sizes the platforms themselves report.
| Statistic | What it measures | Source |
|---|---|---|
| 3 billion+ | Monthly WhatsApp users | Meta Q1 2025 earnings call, reported by Nairametrics, Apr 2025 |
| 3 billion | Monthly Instagram users | Mark Zuckerberg, reported by The Star / Reuters, Sep 2025 |
| 1 billion+ | Monthly Telegram users | Pavel Durov, reported by TASS, Mar 2025 |
| 10 million+ | Bots on Telegram's bot platform | Telegram Bot Platform page, 2026 |
| 90 million+ | Daily active Discord users | Discord company page, Q4 2025 data |
Each channel has its own rules: WhatsApp's messaging tiers and 24-hour window, Telegram's rate limits and Discord's message and intent rules. We keep separate statistics pages for WhatsApp Business, Telegram, Discord and Messenger and Instagram, and one for live chat on websites.
The metrics behind these statistics
Many chatbot statistics look contradictory because they measure different things. These are the terms used on this page and in the reports it cites.
| Metric | What it means and how to read it |
|---|---|
| Deflection rate | Share of conversations that never reached a person. High deflection can mean the bot solved the problem, or that the customer gave up. Always pair it with satisfaction. |
| Containment rate | Share of conversations the bot handled end to end inside the chat. Usually lower than deflection, and closer to real resolution. |
| Resolution rate | Share of conversations where the customer's problem was actually solved. Vendors define it differently; check whether it is self-reported or confirmed. |
| First response time | Time to the first reply. Bots make this near zero, so it matters most for the human queue behind the bot. |
| CSAT | Customer satisfaction, usually the share of positive ratings. Few customers rate chats, so small samples swing widely. |
| Escalation rate | Share of bot conversations handed to a person. Too low can mean customers are trapped; too high means the bot is not answering enough. |
| Average handle time | How long an agent spends per conversation. AI can lower it by summarizing and drafting replies. |
| Customer effort score | How hard the customer felt it was to get help. The repeat-yourself problem shows up here first. |
What changed between 2026 and 2026
The dated sources on this page tell a short story when you put them in order:
- February 2024. Klarna reports its AI assistant handled two-thirds of service chats in its first month. It becomes the reference case for AI service.
- July 2024. Gartner publishes a December 2023 survey: 64% of customers would rather companies did not use AI for service.
- October 2024. Salesforce finds 30% of consumers would choose an AI agent over a person for faster service, and nearly 75% want to know when it is AI.
- March 2025. Gartner predicts agentic AI will resolve 80% of common issues without a person by 2029.
- October 2025. Meta reports more than a billion daily active threads with business accounts. Gartner finds only 20% of service leaders have cut headcount because of AI.
- 2026. Salesforce finds agentic AI use at 66% of service organizations, up from 39%. Gartner finds half of customers now say GenAI makes service easier, while 87% insist on a human option. Meta says over a million businesses use its Business Agents weekly.
In two years the debate moved from whether customers will accept AI to how well it is deployed. The condition customers attached to acceptance, a way to reach a person, did not change at all.
Benchmarks to aim for with a new chatbot
Published averages are a starting point, not a target you will hit in month one. Based on the sources above, a realistic progression for a small business bot looks like this:
| Measure | First month | After tuning | Where the number comes from |
|---|---|---|---|
| Share of questions answered by the bot | 30% to 50% | 60%+ | Salesforce: teams estimate AI handles 30% of cases today and expect 50% by 2027 |
| Bot CSAT versus human CSAT | Below | About equal | LiveChat: 64.7% for bot chats versus 64.2% overall |
| Time to resolve a routine question | A few minutes | Under 2 minutes | Klarna: under 2 minutes, down from 11 |
| Customers who can reach a person | 100% | 100% | Gartner: 87% say it is essential |
These ranges are our reading of the cited data, not a promise. Track your own numbers from day one with chatbot analytics, and see our guide to chatbot analytics metrics for what to watch week by week.
Widely quoted numbers we left out, and why
Statistics pages copy each other, so the same unsourced figures travel from site to site. These are numbers we came across while researching this page and chose not to use.
| Figure you may see elsewhere | Why it is not in our table |
|---|---|
| ChatGPT at 900 million weekly users (February 2026) | Widely reported as an OpenAI announcement, but we could not open a first-party OpenAI page stating it, so we used Pew and the AI Index for usage instead. |
| Chatbot market figures attributed to Grand View Research that do not match its current page | The earlier version of this page cited a 2026 market size we could not find on Grand View's report page. We replaced it with the figures the page shows today. |
| Round adoption claims such as "X% of businesses will use chatbots by year Y" with no report linked | No traceable source. Where we found a real forecast, it is in the forecasts table with its publisher and date. |
| Savings totals in billions with no method | Without the method and base year, a savings total cannot be checked or compared. |
If you know the primary source for one of these, the contact page reaches us, and we will add it with a link.
How we sourced these statistics
Every figure on this page was read on its source page in November 2026. Our rules:
- Primary sources only. A company's own filing, press release or documentation, an analyst firm's published release, or a research organization's report. Where we could only reach a news report of a company statement, the source column says "reported by".
- Dated. Every row shows when the source was published. Older survey data is labeled with its fieldwork year.
- No orphan numbers. If we could not open the source and find the number, we left it out. That removed several figures that circulate widely, including user counts we could only find on aggregator sites.
- Vendor data is labeled. Freshworks, Salesforce, Intercom, LiveChat and Klarna figures describe their own customers or operations, not the whole market.
Sources used on this page:
- Stanford HAI 2026 AI Index (Apr 2026)
- Pew Research Center (Jun 2025)
- Meta Q3 2025 earnings call (Oct 2025)
- Gartner survey of 3,566 customers (Aug 2026)
- Gartner survey of 5,728 customers (Jul 2024)
- Salesforce State of the AI Connected Customer (Oct 2024)
- Zendesk CX Trends 2026 (Nov 2025)
- Gartner survey of 187 service leaders (Dec 2024)
- Salesforce State of Service: AI Agents Edition (3,075 pros) (2026)
- Salesforce State of Service, 7th ed. (6,500 pros) (2025)
- Intercom 2026 Customer Service Transformation Report (2,400+ pros) (2026)
- Meta Q2 2026 earnings call (Jul 2026)
- Freshworks 2026 Customer Service Benchmark (2026)
- Klarna press release (Feb 2024)
- LiveChat Customer Service Report (2024 data) (2024)
- Gartner survey of 321 service leaders (Dec 2025)
- Gartner survey of 321 service leaders (Apr 2026)
- Gartner prediction (Feb 2026)
- Gartner poll of 163 service leaders (Jun 2025)
- Gartner prediction (Mar 2025)
- Gartner prediction (Jan 2026)
- Gartner prediction (Mar 2026)
- Grand View Research (2026)
- MarketsandMarkets (2025)
- Gartner forecast (Sep 2026)
- Meta Q1 2025 earnings call, reported by Nairametrics (Apr 2025)
- Mark Zuckerberg, reported by The Star / Reuters (Sep 2025)
- Pavel Durov, reported by TASS (Mar 2025)
- Telegram Bot Platform page (2026)
- Discord company page (Q4 2025 data)
Spotted a newer figure or an error? The CSV above is the same data, so corrections are easy to check.
What the numbers mean if you are building a chatbot
Put the data together and a clear brief emerges for anyone launching a bot this year:
- Answer from your own knowledge. Customers judge the bot against general AI assistants, so it needs to understand plain questions. Train it on your help content with an AI knowledge base rather than writing every answer by hand. Our guide to training a chatbot on your knowledge base shows how.
- Make the human path obvious. 87% of customers expect it. Offer it early and pass the full conversation along.
- Say it is a bot. Most consumers want to know, and in some places the law requires it.
- Go where your customers already are. The billion-plus daily business threads on messaging apps are a reason to put the same bot on WhatsApp and your site, not just one.
- Measure resolution and satisfaction together. Deflection alone rewards bots that make people give up.
Conferbot is built around that brief: a no-code flow builder, AI answers grounded in your content, live chat handover and one inbox across the channels you connect. Channels are available on every plan, including Free, and you can start from a ready-made template or see the full plan comparison on the pricing page. If you are still comparing tools, our roundup of the best AI chatbot builders and the free chatbot builder page are good next reads.
Start free with Conferbot - the free plan includes 600 conversations a month and needs no credit card, so you can have a bot answering on your busiest channel this afternoon.
Was this article helpful?
Build and deploy in 10 minutes. No coding needed.
50+ Chatbot Statistics for 2026 FAQ
Everything you need to know about chatbots for 50+ chatbot statistics for 2026.
About the Author
The Conferbot team writes about building, deploying, and improving AI chatbots.
View all articles