Example Study

Health search behavior in a research study

20 participants were asked to research intermittent fasting using their normal browsing habits. LaBrowser captured every search query, click, page visit, and ChatGPT prompt as structured event data. Here's what the data reveals.

Study Overview

20

Participants

3,324

Total events

116

Search queries

16.4 min

Avg session

7 of 20

ChatGPT users

15 of 20

Multi-tab users

Search Behavior

Participants averaged 7.1 queries each, ranging from focused 3-query sessions to exhaustive 8-query deep dives. The most common queries targeted direct health benefits and specific fasting schedules.

Activity timeline

Each row is a participant. Dots show events over time — hover for details.

TAB_OPENED
NAVIGATE
INPUT_SUBMIT
0s6m15s12m30s18m45s25mP01P02P03P04P05P06P07P08
NAVIGATE
INPUT SUBMIT
TAB OPENED

Most common queries

Health benefit and schedule queries dominated.

intermittent f...11intermittent f...7intermittent f...716:8 fasting s...7intermittent f...6autophagy fast...6is intermitten...6fasting and in...5intermittent f...5intermittent f...5

Clicks per search query

Most queries led to 2-3 result clicks before the next search.

07142128012345+

Queries per participant

Most participants issued 4-6 queries. A few stopped at 3; one reached 8.

025791-23-45-67-89-1011+

Source Analysis

Healthline was the most-clicked domain, but PubMed articles had the longest dwell times — participants spent 84 seconds on average reading research abstracts, compared to 52 seconds on popular health sites.

Clicks by domain

Popular health sites got the most traffic.

verywellfit.com34scholar.google...34healthdata.org31medicalnewstod...28webmd.com27health.harvard...26healthline.com25dietdoctor.com24mayoclinic.org24nhs.uk23

Avg dwell time by domain (seconds)

Academic sources were read more carefully.

health.harvard...61healthdata.org57pubmed.ncbi.nl...54scholar.google...53nhs.uk51medicalnewstod...50mayoclinic.org49ncbi.nlm.nih.gov49

Tab usage

Most participants stayed in a single tab. Those who used multiple tabs tended to have longer sessions and more diverse source consumption.

024571 tab2 tabs3 tabs5 tabs6+ tabs
Want the setup behind this search behavior study? Read how LaBrowser parsers work or browse the product decision example for a contrasting task.

AI Usage

7 of 20 participants (35%) used ChatGPT alongside Google Search. They issued fewer search queries and visited fewer sources, suggesting AI served as a partial substitute for traditional web searching.

ChatGPT users vs. non-users

AI users had shorter, more focused sessions with fewer queries.

ChatGPT users
Non-ChatGPT users
Avg. search qu..5.65.9Avg. session (..15.816.8

ChatGPT prompts

All 10 prompts captured across 6 participants. Prompts ranged from broad summaries to specific safety questions.

ParticipantPrompt
P09What does the latest research say about intermittent fasting and longevity?
P17How long does it take for autophagy to start during fasting?
P05What should I eat to break a fast for maximum benefit?
P05Is there evidence that fasting improves gut microbiome?
P05Is 16:8 fasting safe for someone with diabetes?
P04How long does it take for autophagy to start during fasting?
P04Summarize the key findings on intermittent fasting from PubMed
P16Is 16:8 fasting safe for someone with diabetes?
P16Is there evidence that fasting improves gut microbiome?
P19Explain autophagy in simple terms
P19How does fasting affect muscle mass?
P14How does fasting affect muscle mass?
P14Can intermittent fasting help with inflammation?

Participant Strategies

Three distinct research strategies emerged: deep divers who read few sources thoroughly, wide scanners who skimmed many sources quickly, and AI-assisted researchers who used ChatGPT to guide their search.

Research strategies

Queries issued vs. average dwell time per source (seconds).

Deep diver
Wide scanner
AI-assisted
014.328.542.85734.35.56.88Queries issuedAvg dwell time (s)

Session durations

Most sessions lasted 12-13 minutes. Deep divers tended toward longer sessions.

012230-22-44-66-88-1010-1212-1414-1616-1818-2020-2222-2424-2626-28

Download & Reproduce

All charts above were generated from the same export files you'd download from the LaBrowser Study Console. The sample dataset and full analysis code are available on GitHub.

technologylab-ai/labrowser-example-health-search

Sample data files (events.json, google_search_v1.json, chatgpt_session_v1.json) plus a Jupyter notebook and Python script reproducing every chart on this page.

View on GitHub

Or clone locally

git clone https://github.com/technologylab-ai/labrowser-example-health-search
cd labrowser-example-health-search
pip install -r requirements.txt
jupyter lab analysis.ipynb