Live Chat Statistics 2026
By Axis Intelligence Research
Co-author: Elena Rodriguez | Last updated: October 22, 2026 | License: CC BY 4.0
AI now opens three live chats in four, but humans still answer two in three. On Comm100’s platform of more than 220 million chats, AI agents handled 75.3% of conversations in 2025 but fully resolved only 44.8% of those. Axis Intelligence Research calculates that 66.3% of chats still reached a human agent, almost exactly the 66.2% of 2024.
Quick Answer
The average live chat wait time is 22.8 seconds and the average reply time inside a chat is 44.6 seconds, with satisfaction steady at 4.1 out of 5 (Comm100, 2025 data). AI agents now engage first in 75.3% of chats, yet Axis Intelligence Research estimates that 66.3% still reach a person. Consumers prefer live chat with a person (16%) over chatbots (7%) by 2.3 to 1 (Qualtrics, 2025).
Key Findings
- According to Axis Intelligence Research, 66.3% of live chats on AI-enabled Comm100 accounts still reached a human agent in 2025, unchanged from 66.2% in 2024, even as AI handling rose from 73.8% to 75.3%.
- Axis Intelligence Research estimates the direct U.S. wage cost of a human-answered live chat at $3.11, from BLS’s $21.53 median hourly wage and Comm100’s 1,201 chats per agent per month.
- Consumers prefer live chat, email or messaging with a person 2.3 times more often than a chatbot (16% vs 7%), per Axis Intelligence Research analysis of Qualtrics’ Q3 2025 study of 20,001 consumers.
- The most-used live chat tool depends on the tier: LiveChat leads the top million sites, the Zendesk family leads the top 10,000 with 395 detections, and WhatsApp click-to-chat widgets make up 18.5% of all detections, per Axis Intelligence Research analysis of BuiltWith data (October 2026).
- Axis Intelligence Research’s Human Handoff Demand Index (HHDI) reads 84.3 out of 100, built from Gartner’s February–March 2026 survey of 3,566 customers.
What Is the Average Live Chat Response Time in 2026?
Customers wait an average of 22.8 seconds for an agent to accept a chat, and 44.6 seconds for each reply once the chat starts. Those figures come from the Comm100 AI Live Chat Benchmark Report 2026, based on more than 220 million live chat interactions, comparing 2025 with 2024. Comm100 defines wait time as the gap before an agent accepts the chat, and response time as the average wait for each message reply.
Both numbers improved slightly: wait time from 23.6 to 22.8 seconds, response time from 44.8 to 44.6 seconds. The average chat lasted 8 minutes 50 seconds in both years.
Live chat wait and response time by team size
| Team size | Wait time 2024 | Wait time 2025 | Response time 2025 | Source |
|---|---|---|---|---|
| 1–5 agents | 24.6 s | 29.2 s | 50.1 s | Comm100 Benchmark 2026 |
| 6–10 agents | 18.2 s | 18.2 s | 44.2 s | Comm100 |
| 11–25 agents | 18.0 s | 15.8 s | 39.2 s | Comm100 |
| 26+ agents | 45.5 s | 28.4 s | 41.3 s | Comm100 |
| Average | 23.6 s | 22.8 s | 44.6 s | Comm100 |
Wait time, response time and CSAT by industry (2025)
| Industry | Wait time | Response time | CSAT (1–5) | Source |
|---|---|---|---|---|
| Consumer Services | 16.3 s | 35 s | 4.4 | Comm100 |
| Banking & Finance | 17.9 s | 50 s | 4.0 | Comm100 |
| Travel & Hospitality | 19.1 s | 1 min 59 s | 3.9 | Comm100 |
| iGaming | 20.2 s | 40.6 s | 4.1 | Comm100 |
| Insurance | 21.9 s | — | 4.3 | Comm100 |
| Transportation | 25.1 s | 1 min 7 s | 4.5 | Comm100 |
| Education | 30.6 s | 45.5 s | 3.7 | Comm100 |
| Government | 53.6 s | 56 s | 4.3 | Comm100 |
| Telecommunications | 2 min 28 s | — | 3.6 | Comm100 |
| Technology | — | 1 min 11 s | 4.4 | Comm100 |
| Real Estate | — | — | 4.9 | Comm100 |
Elena Rodriguez: The big story is in the team-size table, not the average. Teams of 26 or more agents cut wait time by 37.5%, from 45.5 to 28.4 seconds, while teams of one to five got slower, from 24.6 to 29.2 seconds. Two years ago the large team was the slow one. Automation at intake is now a scale advantage: a bot that triages the queue helps a 40-seat floor far more than a founder who answers chats between sales calls. Telecommunications is the cautionary case. Wait time of 2 minutes 28 seconds and the lowest CSAT in the dataset, 3.6, despite bots touching 99.9% of chats.
How Much of Live Chat Does AI Actually Resolve?
AI agents engaged first in 75.3% of chats on AI-enabled Comm100 accounts in 2025, but resolved only 44.8% of those without a human. Comm100 counts a chat as AI-handled whenever the bot engages first, whether it resolves the question or transfers the customer. The report notes that these AI figures cover only accounts that adopted AI chatbots.
Axis Intelligence Research multiplied the two averages to estimate where every chat ends up.
Where a live chat ends up on AI-enabled accounts (Axis estimate)
| Path | 2024 | 2025 | Formula |
|---|---|---|---|
| Fully resolved by AI | 33.8% | 33.7% | handled × resolved |
| Bot first, then a human | 40.0% | 41.6% | handled − fully resolved |
| Human only, no bot | 26.2% | 24.7% | 100 − handled |
| Reached a human (total) | 66.2% | 66.3% | 100 − fully resolved |
Inputs: Comm100 handled 73.8% → 75.3%, resolved 45.8% → 44.8%. Multiplying two platform averages gives an estimate, not a measured count.
Elena Rodriguez: This is the number the AI rollout plans miss. Bots took on more of the front door in 2025, but they resolved a slightly smaller share of what they touched, so the human workload share did not move: 66.2% to 66.3%. More chats now arrive at an agent after a bot has tried and failed, which is a harder conversation than a fresh one. That explains Comm100’s other figures: workload per agent fell 5.8%, while chat duration stayed flat at 8 minutes 50 seconds.
AI handling and resolution by team size and industry (2025)
| Segment | Chats handled by AI | AI-handled chats resolved by AI | Source |
|---|---|---|---|
| 1–5 agents | 54.3% (from 20.9%) | 89.0% | Comm100 |
| 6–10 agents | 45.3% (from 65.3%) | 69.6% | Comm100 |
| 11–25 agents | 92.5% | 47.8% | Comm100 |
| 26+ agents | 67.5% | 41.2% | Comm100 |
| Insurance | 99.9% | — | Comm100 |
| Telecommunications | 99.9% | 63.9% | Comm100 |
| Banking & Finance | 97.1% | 75.2% | Comm100 |
| Government | 94.1% | 67.6% | Comm100 |
| Education | 90.4% | 75.9% | Comm100 |
| iGaming | 75.6% | 38.1% | Comm100 |
| Technology | 54.8% | 67.3% | Comm100 |
| Consumer Services | 27% | — | Comm100 |
What vendors report
Other vendors publish a single resolution figure, each with its own definition. Intercom says on its Fin product page that Fin resolves 76% of conversations on average across more than 12,000 customers, about 2 million a week. Text S.A. reported a 74% AI agent resolution rate in an April 2026 investor release. These figures are company-reported and are not directly comparable with Comm100’s 44.8%, which is an average across all AI-enabled accounts on its platform.
Are Customers Satisfied with AI in Live Chat?
Satisfaction with AI chatbots rose to 49.3% in 2025 from 45.2% in 2024, while satisfaction on chats handed from a bot to a human reached 92.6%, up from 86.7% (Comm100). Overall live chat CSAT held at 4.1 out of 5.
That gap is the case for the hybrid model in one line. Customers rate bot-only experiences at roughly one in two and bot-to-human handoffs at better than nine in ten. Real Estate (4.9) and Construction (4.8) post the highest CSAT; Telecommunications (3.6) and Education (3.7) the lowest.
Customer surveys agree. Only 7% of customers used a chatbot in their most recent service interaction, and only 27% would try one again after a bad experience, according to a Gartner survey of 3,566 customers in February and March 2026. In the August 2026 Gartner Q&A, 87% called access to a human agent essential when generative AI is used. Gartner’s July 2026 release found customers about three times more likely to turn to third-party GenAI tools than to a company’s own chatbot.
Do Customers Prefer Live Chat or Chatbots?
Live chat with a person wins: 16% of consumers prefer live chat, email or messaging with a person, against 7% for a chatbot or AI assistant. The figures come from the Qualtrics XM Institute Data Snapshot Consumer Preferences and Priorities, 2026, based on 20,001 consumers in 14 countries surveyed in Q3 2025, averaged across eight common service tasks. Qualtrics groups live chat with email and person-to-person messaging in one category.
Overall, 64% of consumers prefer a human channel (messaging, phone or in person). Self-service on a website or app is the single most preferred channel at 28%; automated phone systems are last at 2%.
Channel preference by age
| Age group | Self-service | Chatbot / AI | Live chat, email or messaging (person) | Phone (person) | In person | Source |
|---|---|---|---|---|---|---|
| Under 30 | 27% | 11% | 20% | 19% | 21% | Qualtrics XM Institute, Q3 2025 |
| 30–44 | 31% | 8% | 19% | 21% | 20% | Qualtrics |
| 45–59 | 30% | 5% | 14% | 27% | 23% | Qualtrics |
| 60+ | 25% | 5% | 13% | 28% | 28% | Qualtrics |
| All | 28% | 7% | 16% | 24% | 23% | Qualtrics |
Channel preference by task
| Task | Chatbot / AI | Live chat, email or messaging (person) | Phone (person) | Source |
|---|---|---|---|---|
| Get order status | 9% | 26% | 19% | Qualtrics |
| Get technical support | 9% | 22% | 32% | Qualtrics |
| Resolve a billing issue | 6% | 20% | 46% | Qualtrics |
| Select a phone plan | 6% | 13% | 22% | Qualtrics |
| Schedule a medical appointment | 6% | 13% | 38% | Qualtrics |
| Book an airplane ticket | 6% | 11% | 16% | Qualtrics |
| Apply for a bank account | 6% | 11% | 15% | Qualtrics |
| Buy a TV | 5% | 8% | 8% | Qualtrics |
Elena Rodriguez: Under-30s are the only group where chatbot preference reaches double digits, 11%, and even they prefer chatting with a person nearly twice as often, 20%. Order status is live chat’s home ground: at 26%, chat with a person beats the phone for that task. Billing is where chat loses to voice, 20% against 46%. Route accordingly: put the bot and the chat widget on the order-tracking page, and a phone number next to the invoice.
How Much Does a Live Chat Conversation Cost?
Axis Intelligence Research estimates the direct U.S. wage cost of a human-answered live chat at $3.11.
Formula: median hourly wage × 2,080 hours ÷ 12 months ÷ chats per agent per month.
- Median hourly wage of U.S. customer service representatives: $21.53, May 2025 (Bureau of Labor Statistics).
- Chats per agent per month: 1,201 in 2025, down from 1,275 (Comm100).
$21.53 × 2,080 ÷ 12 ÷ 1,201 = $3.11 per chat.
| Segment | Chats per agent per month | Wage cost per human chat |
|---|---|---|
| Teams of 26+ agents | 1,590 | $2.35 |
| iGaming | 1,540 | $2.42 |
| All industries | 1,201 | $3.11 |
| Consumer Services | 829 | $4.50 |
| Teams of 6–10 agents | 810 | $4.61 |
| Technology | 299 | $12.48 |
| Banking & Finance | 214 | $17.44 |
| Insurance | 121 | $30.84 |
| Telecommunications | 107 | $34.88 |
| Education | 62 | $60.19 |
Axis Intelligence Research calculation. Applies the U.S. median wage to a global platform average; excludes benefits, software, management and time agents spend on other channels.
Elena Rodriguez: The spread is the point. In insurance and telecom, bots touch 99.9% of chats, so agents see only about a hundred a month and each one carries $30 or more in wages. That is not inefficiency so much as specialization: the agent has become the exception desk. In education, low volume and seasonal staffing push the cost past $60. The business case for AI in those sectors is not removing agents; it is making sure the chats that reach them are worth that much.
How Many Websites Use Live Chat in 2026?
BuiltWith counts 7,419,132 live chat widget detections across the web, including 222,445 in the top million sites and 944 in the top thousand, per its Live Chat Widgets usage data, updated October 18–19, 2026. BuiltWith detects the scripts a site loads; a site running two tools counts twice.
| Traffic tier | Detections | Detections per 100 sites | Source |
|---|---|---|---|
| Top 1,000 sites | 944 | 94.4 | BuiltWith; Axis calc. |
| Top 10,000 sites | 4,361 | 43.6 | BuiltWith; Axis calc. |
| Top 100,000 sites | 26,760 | 26.8 | BuiltWith; Axis calc. |
| Top 1 million sites | 222,445 | 22.2 | BuiltWith; Axis calc. |
| Entire internet | 7,419,132 | — | BuiltWith |
The United States accounts for 3,629,737 detections, 48.9% of the total by Axis calculation, ahead of Brazil (293,507), Russia (238,866), India (231,214) and the United Kingdom (205,888).
What Is the Most Popular Live Chat Software in 2026?
There is no single leader; it depends on which part of the web you measure. Axis Intelligence Research compared BuiltWith’s three rankings side by side.
| Rank | Entire internet (Oct 19) | Top 1 million (Oct 18) | Top 10,000 (Oct 19) |
|---|---|---|---|
| 1 | LeadConnector — 591,785 | LiveChat — 21,182 | Zoom* — 674 |
| 2 | WhatsApp Me — 586,645 | Zoom* — 17,981 | WebEx* — 664 |
| 3 | Click to Chat for WhatsApp — 560,644 | WebEx* — 17,620 | Intercom — 278 |
| 4 | tawk.to — 517,492 | tawk.to — 14,651 | Zendesk Embeddables — 275 |
| 5 | LiveChat — 421,606 | Telegram* — 13,391 | Salesforce Live Agent — 229 |
| 6 | Chaty Premio — 344,548 | Zendesk Embeddables — 10,639 | Drift — 130 |
| 7 | WhatsApp Chat WP — 226,268 | Intercom — 9,958 | HubSpot Messages — 125 |
| 8 | Kakao* — 224,824 | HubSpot Messages — 7,512 | Zopim — 120 |
| 9 | Tencent QQ* — 212,577 | Zopim — 5,739 | tawk.to — 89 |
| 10 | Tidio — 185,467 | HubSpot Conversations — 4,901 | HubSpot Conversations — 80 |
*Messaging or video tool rather than a support chat widget. Source: BuiltWith.
Three patterns stand out from the Axis analysis:
- WhatsApp buttons rule the long tail. WhatsApp Me, Click to Chat for WhatsApp and WhatsApp Chat WP total 1,373,557 detections, 18.5% of the whole category.
- Enterprise suites rule the top. In the top 10,000, the Zendesk family (Embeddables plus Zopim) leads support chat with 395 detections (9.1%), ahead of Intercom (278) and Salesforce Live Agent (229). LiveChat, first in the top million, has 77 detections in the top 10,000 (1.8%).
- The category is noisy. Zoom, WebEx and Line.me make up 32.2% of BuiltWith’s top-10,000 live chat category.
Elena Rodriguez: Read this as three different markets. Small sites install a free WhatsApp button and move the conversation to the customer’s phone. The million-site middle buys a standalone widget, and LiveChat wins there. The top 10,000 buy chat as one module of a support suite that already holds the tickets and the customer record, which is where Zendesk, Intercom and Salesforce sit. A vendor’s install count tells you which of the three markets it serves, not how big it is.
Is the Live Chat Software Business Growing?
Standalone live chat is shrinking inside the vendors that sell it, while bots and help desks grow. Text S.A. reported that its LiveChat product earned PLN 275.5 million in fiscal 2025/26, down 12.4% from PLN 314.3 million, while ChatBot revenue rose 9.7% and HelpDesk 81.7%, per its Q2 2026/27 KPI presentation. LiveChat still produced 83.3% of group revenue. Group monthly recurring revenue (MRR) was $7.29 million at the end of September 2026, up 3.3% year on year, and customers paying $500 or more a month generated 56% of it, up from 50%.
Text says its LiveChat and Text widgets draw more than 2 billion views a week, and that conversion to order rises 266% when customers chat with its AI agent (September 2026 presentation; company-reported).
LivePerson, the enterprise conversational platform, reported Q2 2026 revenue of $52.5 million, down 12% from $59.6 million, in its Form 10-Q. Americas revenue fell 30.8% by Axis calculation. It booked a $51.8 million goodwill impairment and agreed to be acquired by SoundHound AI, with an aggregate consideration amount of about $42.8 million before adjustments.
Budgets follow the same direction. A Gartner survey of 199 service leaders (April–May 2026) found AI spending up 38% while total service budgets rose 2%. Leaders still rank live chat and phone as their most valuable channels today and expect GenAI chatbots to overtake them within two years. Only 24% of service leaders showed positive financial returns from AI, per Gartner’s July 2026 release.
Elena Rodriguez: Text’s own revenue table is the clearest market signal on this page. The widget that leads BuiltWith’s top million lost 12.4% of its revenue in a year while the bot and ticketing products grew. Customers are not abandoning chat; they are buying it as part of a bundle and paying for the automation and the case management around it.
The Axis Human Handoff Demand Index (HHDI)
The Human Handoff Demand Index (HHDI) measures how strongly customers expect a human behind the chat widget, on a 0–100 scale. Near 100, customers route around bots and insist on people; near 0, they accept AI-only service. All three inputs come from a single Gartner survey, so they share one sample and one method.
Baseline reading, October 22, 2026: 84.3 / 100.
| Component | Gartner figure | Score | Weight |
|---|---|---|---|
| Customers who call human access essential when GenAI is used | 87% | 87 | 1/3 |
| Customers who did not use a chatbot in their latest service interaction | 7% used one | 93 | 1/3 |
| Customers unwilling to retry a chatbot after a bad experience | 27% would retry | 73 | 1/3 |
Formula: HHDI = (87 + (100 − 7) + (100 − 27)) ÷ 3 = 84.3.
The index will be recalculated when Gartner publishes its next customer channel survey. The input most likely to move it is the 7% chatbot-use figure.
Elena Rodriguez: The platform data and the survey data point the same way. Comm100 shows two in three chats still reach a person; the HHDI shows customers expect exactly that. Plan the human queue as the default and earn bot traffic one issue type at a time, starting with order status.
How Many People Work in Live Chat and Customer Service?
The U.S. employs 2,666,000 customer service representatives, the occupation that staffs most chat queues, according to the Bureau of Labor Statistics. BLS projects a 5% decline (−141,800 jobs) from 2025 to 2035, with about 289,500 openings a year from turnover.
AI raises output most for new agents. A study of 5,179 chat support agents in the Quarterly Journal of Economics (2025; NBER working paper 31161) found an AI assistant raised issues resolved per hour by 14% on average and 34% for novice and low-skilled agents.
Methodology
Scope. Live chat means the website or in-app widget where a customer types to a business in real time, answered by a person, an AI agent or both.
Freshness rule. Every benchmark on this page was published in 2025 or 2026. Where a 2026 report compares 2025 with 2024, both years are shown. Pre-2024 benchmarks were excluded.
Sources. Comm100’s 2026 benchmark report (PDF; 2025 platform data, company-reported); Qualtrics XM Institute’s Data Snapshot (PDF; 20,001 consumers, Q3 2025); Gartner’s 2026 survey releases; Text S.A.’s September and October 2026 investor presentations and investor news; Intercom’s Fin page; LivePerson’s SEC filing; BLS; NBER/QJE; BuiltWith detection counts (estimates). BuiltWith pages are bot-protected; counts were read from screenshots of the live pages captured by the Axis editorial team on October 21–22, 2026. All documents were opened on October 21–22, 2026.
Axis calculations.
- Chat outcome split: fully resolved by AI = handled × resolved; reached a human = 100 − fully resolved; bot then human = handled − fully resolved. Uses Comm100 averages for AI-enabled accounts; an estimate.
- Wage cost per human chat = $21.53 × 2,080 ÷ 12 ÷ chats per agent per month.
- Live chat vs chatbot preference = 16 ÷ 7 (Qualtrics).
- Detections per 100 sites = tier detections ÷ tier size × 100; U.S. share = 3,629,737 ÷ 7,419,132.
- Vendor families: Zendesk = Embeddables + Zopim; HubSpot = Messages + Conversations; WhatsApp = WhatsApp Me + Click to Chat for WhatsApp + WhatsApp Chat WP.
- HHDI = (87 + (100 − 7) + (100 − 27)) ÷ 3.
About This Dataset
live-chat-statistics.csv holds 438 rows: Comm100’s 2025 and 2024 live chat benchmarks by team size and industry; Qualtrics channel preferences by age, country and task; BuiltWith detections by traffic tier, country and tool across three rankings; Gartner’s 2026 survey results; Text, Intercom and LivePerson disclosures; BLS and NBER data; and every Axis calculation, with its formula in method_note. License CC BY 4.0. We update it when Comm100, Qualtrics, Gartner, BuiltWith or BLS publish new figures.
Cite this dataset: Axis Intelligence Research, Live Chat Statistics 2026, 2026.
Citation
- APA: Axis Intelligence Research. (2026). Live chat statistics 2026: Response times, AI resolution and who still answers the chat [Data set]. Axis Intelligence. https://axis-intelligence.com/live-chat-statistics/
- MLA: Axis Intelligence Research. “Live Chat Statistics 2026: Response Times, AI Resolution and Who Still Answers the Chat.” Axis Intelligence, 22 Oct. 2026, axis-intelligence.com/live-chat-statistics/.
- Chicago: Axis Intelligence Research. “Live Chat Statistics 2026: Response Times, AI Resolution and Who Still Answers the Chat.” Axis Intelligence, October 22, 2026. https://axis-intelligence.com/live-chat-statistics/.
Live Chat Questions Support Leaders Ask in 2026
What is a good live chat wait time?
The 2025 average is 22.8 seconds before an agent accepts a chat, and 44.6 seconds per reply (Comm100, 220 million+ chats). Consumer Services sets the pace at 16.3 seconds; Telecommunications averages 2 minutes 28 seconds.
What share of chats can an AI agent resolve without a human?
On Comm100’s platform, AI agents engaged first in 75.3% of chats in 2025 and resolved 44.8% of those. Axis Intelligence Research estimates that 33.7% of all chats end with the bot and 66.3% still reach a person.
Why do vendors quote 74% or 76% AI resolution when Comm100 shows 44.8%?
They measure different things. Intercom (76%) and Text (74%) report a company-defined average for their own AI agents; Comm100’s 44.8% is the average across all AI-enabled accounts on its platform, measured against every chat the bot touched.
Is a chatbot-to-human handoff bad for satisfaction?
No. Chats handed from a bot to an agent scored 92.6% satisfaction in 2025, against 49.3% for AI chatbots alone (Comm100). Gartner found 87% of customers call human access essential when GenAI is used.
Which customer service tasks should stay on live chat with a person?
Order status is the strongest fit: 26% of consumers prefer chatting with a person for it, against 9% for a chatbot (Qualtrics, Q3 2025). For billing, 46% prefer the phone.
How much does one live chat cost in agent wages?
About $3.11 per human-answered chat, by Axis Intelligence Research’s calculation from BLS’s $21.53 median wage and 1,201 chats per agent per month. It ranges from $2.42 in iGaming to $34.88 in telecom, where agents only see the chats bots cannot close.
Which live chat software do the biggest websites use?
In BuiltWith’s top 10,000 sites (October 19, 2026), the Zendesk family leads support chat with 395 detections, followed by Intercom (278) and Salesforce Live Agent (229). LiveChat leads the top million with 21,182.
Do younger customers prefer chatbots?
More than older ones, but still not over people. Of consumers under 30, 11% prefer a chatbot and 20% prefer live chat, email or messaging with a person, against 5% and 13% for those over 60 (Qualtrics).
Related Axis research: AI customer service statistics · chatbot statistics and the CDRI index · how many websites are active · enterprise software spending · attention span data · AI ROI calculator