New report: The State of Conversational AI 2026 Download free

The authority on AI + Human conversations.

What wins and what converts when AI and people talk to your customers.

Latest study, October 2026

Short messages get over 5 times the replies

Our recent study compared 140,861 real messages. Keep it under 60 words.

See the findings
Latest study, October 2026
4.5x

A form link got 4.5 times the responses of asking people to text back.

Conversational AI Labs The State of
Conversational
AI 2026
greetAI
New report, free download

The State of Conversational AI 2026

What wins and what converts, from a recent study of 270,131 real messages.

  • All 9 tests and what they found
  • What other major studies show
  • Nine rules for better conversations
Download the free report

What real conversations show.

From our recent study of 270,131 messages, to 9 October 2026.

38,194
People in conversation
33,953
Conversations
10 seconds
Typical first reply
61.6%
Response rate
8,142
Calls answered
19,493
Call minutes
26,814
Facts captured
913,920
Messages compared

What the study found

Reply rates. Real messages compared side by side, each difference over 99.9% certain.

Proof, not promises, about AI and people working together.

Ten channels, one standard

Voice, SMS, WhatsApp, email, Instagram, Messenger, X, TikTok, LinkedIn and web chat, all measured the same way.

Where AI hands over to a person

The moment a human steps in decides whether the customer stays. We measure when it should happen and how.

The Channel Index

We score each channel on speed, problems solved first time and customer satisfaction.

White papers

Clear write-ups of what we find, checked by people outside the team.

Data you can use

See how your own conversations compare, and where to improve first.

Building or launching conversational AI? Get in touch to find out how to do it properly.

Ten channels. Measured side by side.

WhatsApp Voice Email SMS Instagram Messenger Web chat X LinkedIn TikTok

Four steps before anything is published.

1

Collect

Real conversations, anonymised and shared by partner organisations, then labelled by trained reviewers.

2

Test

Fair trials that compare one approach against another, set up in advance so the result cannot be bent.

3

Analyse

Every result shows its range and sample size. We publish the results that did not work too.

4

Review

At least two outside academics check each paper. The data and code are kept for anyone to re-run.

What the wider research says.

+34%

AI helped new staff most

  • New agents+34%
  • All agents+14%
Brynjolfsson, Li and Raymond, NBER. 5,179 support agents.
4x

AI matched skilled staff on sales calls

  • AI4x
  • Skilled staff4x
  • New staff1x
Luo et al, Marketing Science, 2019. 6,200 customers.
−79.7%

Saying “I am a bot” first cut purchases

  • Not told first100
  • Told first20
Purchases indexed to 100. Luo et al, Marketing Science, 2019.
64%

Customers still want to be won over

  • Prefer no AI64%
  • Might switch53%
Gartner survey, 2024. 5,728 customers.

Building or launching conversational AI? Get in touch to find out how to do it properly.

Findings for your kind of business.

Professional services

Financial services

Healthcare

Retail

Energy and utilities

Hospitality and travel

Local government

Automotive

Book a consultation.

Building or launching conversational AI? Explore the data with us and learn how to do it properly.

Consultation

£1,500

45 minutes, plus VAT

Book a consultation
  • Explore the data for your industry
  • Learn what AI could do for your customers
  • Leave with clear next steps

Building or launching conversational AI?

Get in touch to find out how to do it properly.

Get in touch
Download the free report

Book a consultation

We only use this to reply to you. Privacy