Shopping with a Platform AI Assistant: Who Adopts, When in the Journey, and What For
Authors: Se Yan, Han Zhong, Zemin, Zhong, Wenyu Zhou
Organizations: Zachary · Guanghua School of Management, Peking University, Beijing 10086, China. · Rotman School of Management, University of Toronto, Toronto, Ontario M5S 3E6, Canada. · International Business School, Zhejiang University, Haining, Zhejiang 314400, China.
This paper provides some of the first large-scale descriptive evidence on how consumers adopt and use platform-embedded shopping AI in e-commerce. Using data on 31 million users of Ctrip, China's largest online travel platform, we study "Wendao," an LLM-based AI assistant integrated into the platform. We document three empirical regularities. First, adoption is highest among older consumers, female users, and highly engaged existing users, reversing the younger, male-dominated profile commonly documented for general-purpose AI tools. Second, AI chat appears in the same broad phase of the purchase journey as traditional search and well before order placement; among journeys containing both chat and search, the most common pattern is interleaving, with users moving back and forth between the two modalities. Third, consumers disproportionately use the assistant for exploratory, hard-to-keyword tasks: attraction queries account for 42% of observed chat requests, and chat intent varies systematically with both the timing of chat relative to search and the category of products later purchased within the same journey. These findings suggest that embedded shopping AI functions less as a substitute for conventional search than as a complementary interface for exploratory product discovery in e-commerce.
Figures & tables
Figure 1: AI chat adoption rates by demographic and account characteristics
Figure 2: Density of event types over the course of the customer journey
Chat-Before-Search
Chat-Between-Search
Chat-After-Search
Chat-Only
Panel A: Prevalence
Share of journeys (%)
15.076
30.916
12.411
41.596
Panel B: Journey characteristics
Median journey length (hours)
51.809
94.837
51.970
18.405
Mean login sessions
11.313
28.625
10.538
5.582
Mean chat sessions
1.377
1.566
1.358
1.411
Table 1: Journey Characteristics by Relative Timing of Chat and Search
Figure 3: Temporal distribution of orders relative to chat sessions by chat intent
Appendix figures & tables23 assets
Supplementary material from the paper’s appendix.
Appendix
AI chat adopters
AI chat non-adopters
Difference
Variable
N
Mean
S.D.
N
Mean
S.D.
Diff.
t -stat
Panel A: Demographics and account status
Female
1,904,368
0.516
0.500
29,237,985
0.440
0.496
0.076
204.492
Age group
1,904,368
2.616
1.096
29,237,985
2.186
1.180
0.430
521.849
New user indicator
1,904,368
0.430
0.495
29,237,985
0.516
0.500
− 0.085
− 230.466
iOS device indicator
1,904,368
0.379
0.485
29,237,985
0.490
0.500
− 0.112
− 306.670
Appendix
Table A1 : Summary statistics by AI chat adoption status
Figure A1 : Estimated marginal effects on AI chat adoption from linear probability, logit, and probit models
Figure A2 : Permutation-based variable importance from a random forest model of AI chat adoption
Specification
LPM R2
RF AUC
Demographics only
0.014
0.635
Demographics + Engagement
0.036
0.962
Appendix
Table A2 : Predictive Performance of Adoption Models
Figure A3 : Illustrative customer journey timelines for ten randomly selected users
Variable
N
Mean
S.D.
P5
P25
Median
P75
P95
Panel A: All journeys
Journey length (hours)
253,209,391
34.110
79.577
0.000
0.000
7.000
43.000
138.000
Number of login sessions
253,209,391
4.235
10.717
1.000
1.000
2.000
4.000
14.000
Total login time (minutes)
253,209,391
46.951
171.683
0.000
2.000
11.000
45.000
193.000
Number of chat sessions
253,209,391
0.001
0.059
0.000
0.000
0.000
0.000
0.000
Number of orders
253,209,391
0.315
0.831
0.000
0.000
0.000
0.000
2.000
Appendix
Table A3 : Summary statistics of customer journeys
Figure A4 : Normalized temporal distribution of events relative to chat sessions
Figure A5 : Normalized temporal distribution of events relative to order placement (with-chat journeys)
Figure A6 : Normalized temporal distribution of events relative to order placement (no-chat journeys)
Chat-Before-Search
Chat-Between-Search
Chat-After-Search
Chat-Only
Panel A: Prevalence
Share of sessions (%)
10.6
12.0
10.2
67.2
Panel B: Session characteristics
Mean session length (hours)
0.8
1.2
0.8
0.4
Mean chat sessions
1.3
1.4
1.3
1.4
Mean hotel searches
11.3
35.8
11.2
0.0
Appendix
Table A4 : Session Characteristics by Relative Timing of Chat and Search
Chat-Before-Search
Chat-Between-Search
Chat-After-Search
Chat-Only
Panel A: Prevalence
Share of journeys (%)
15.049
30.804
12.411
41.736
Panel B: Journey characteristics
Median journey length (hours)
51.727
94.904
51.889
18.160
Mean login sessions
11.282
28.639
10.521
5.562
Mean chat sessions
1.377
1.565
1.357
1.411
Appendix
Table A5 : Journey Characteristics by Relative Timing of Chat and Search (48-Hour Inactivity Definition)
Chat-Before-Search
Chat-Between-Search
Chat-After-Search
Chat-Only
Panel A: Prevalence
Share of journeys (%)
15.089
30.940
12.414
41.557
Panel B: Journey characteristics
Median journey length (hours)
51.878
94.874
52.037
18.472
Mean login sessions
11.319
28.620
10.548
5.586
Mean chat sessions
1.377
1.566
1.358
1.411
Appendix
Table A6 : Journey Characteristics by Relative Timing of Chat and Search (48-Hour Inactivity, 8-Hour Order-Gap Definition)
Chat-Before-Search
Chat-Between-Search
Chat-After-Search
Chat-Only
Panel A: Prevalence
Share of journeys (%)
15.091
30.957
12.420
41.531
Panel B: Journey characteristics
Median journey length (hours)
51.938
94.876
52.049
18.526
Mean login sessions
11.324
28.617
10.553
5.590
Mean chat sessions
1.377
1.566
1.358
1.411
Appendix
Table A7 : Journey Characteristics by Relative Timing of Chat and Search (48-Hour Inactivity, 12-Hour Order-Gap Definition)
Figure A7 : Density of Event Types over the Course of the Customer Journey (48-Hour Inactivity Definition)
Figure A8 : Density of Event Types over the Course of the Customer Journey (48-Hour Inactivity, 8-Hour Order-Gap Definition)
Figure A9 : Density of Event Types over the Course of the Customer Journey (48-Hour Inactivity, 12-Hour Order-Gap Definition)
Intent
User Query (Translated)
AI Response (Excerpt, Translated)
Attraction
What are the must-visit attractions in Qingdao?
Highlights include Zhanqiao Pier, Badaguan Scenic Area, and Laoshan Scenic Area (trails from ¥120).
Hotel
Good-value hotels or guesthouses near a subway station in Beijing?
Six metro-adjacent options are compared, including a historic siheyuan courtyard hotel (255m from station, ¥394/night) and a loft apartment (¥381/night, 4.7 rating).
Travel Planning
What is the best one-day itinerary for Macau?
Start at Macau Tower, continue to the Venetian and Parisian resorts, and end with the figure-eight Ferris wheel at night.
Consumer Support
How do I register for Phoenix membership through the Trip.com app?
Please contact Trip.com customer service via in-app live chat or phone support using the link below.
Flight
Which economy-class flights departing Dali have available seats today?
Which destination city are you flying to?
Other Transportation
What is the driving route from Heihe to Wudalianchi?
The drive is approximately 247 km and takes around 3 hours; see the link below for the full route.
Appendix
Table A8 : Illustrative AI Chat Conversations by Intent Category
Variable
N
Mean
S.D.
P5
P25
Median
P75
P95
Panel A: User-level AI assistant engagement
Total number of logins
31,129,629
34.339
46.403
2.000
7.000
19.000
44.000
114.000
Average chat sessions per login among AI adopters
188,647
0.092
0.171
0.006
0.016
0.035
0.087
0.333
Panel B: Session-level summary statistics
Number of requests in session
213,460
1.313
0.936
1.000
1.000
1.000
1.000
3.000
Characters per minute
213,460
1,456.520
3,710.211
18.000
458.000
1,168.079
1,656.000
2,598.000
Appendix
Table A9 : Summary statistics for AI assistant engagement and usage patterns
User Intent
User group
Attr.
Hotel
Plan.
C.S.
Flight
Trans.
Other
Panel A: Full sample
All users
0.421
0.177
0.073
0.054
0.018
0.083
0.173
Panel B: Gender
Female
0.450
0.175
0.068
0.053
0.015
0.076
0.163
Male
0.408
0.171
0.083
0.056
0.020
0.079
0.183
Appendix
Table A10 : User intent distribution of AI assistant requests
Figure A10 : Average number of AI chat sessions per user by demographic and account characteristics
Figure A11 : Average AI chat session duration by demographic and account characteristics
Figure A12 : Average number of requests per AI chat session by demographic and account characteristics
Figure A13 : Distribution of Chat Requests by Journey Progress Quintiles and Chat Intent