Chuhee LeeIndependent research
All publications
Cover art: a square grid of 13 by 13 dots, one for each cell of the benchmark matrix; 11 are solid, for cells with a US measurement, and the rest are hollow, for gaps.
Horizon
Data 2016–2026
Scope
US · cars, transit, planes
Lens
What people do while traveling
Basis
Observed first, then surveys
SignalsTravelOctober 2026 · Research report

What people do on the move

Drivers, riders and flyers spend hours a week in motion or waiting. This benchmark maps what Americans do with that time, by mode and by whether the vehicle is moving or stopped, using only data from the last ten years, and rates how well each cell is measured.

Research questionWhat do people in the US do while traveling, in each mode and in motion versus stopped, and where is that behavior actually measured?

By Chuhee Lee≈8 min readData as of 21 Oct 2026One-time editionChange log

With research assistance from AI models (Claude, Gemini, ChatGPT) · Views are the author’s own

Cite asChuhee Lee, What people do on the move, October 2026 edition (data as of 21 October 2026), https://blueplanet-ai.github.io/on-the-move/, CC BY-NC 4.0.

The report in 60 seconds

Three takeaways · each with its source
  1. 4.5% vs 37% Fact

    A moment is not a month

    NHTSA observers at stopped intersections saw 4.5% of drivers handling a phone in June 2024, up from 3.0% in 2023. In AAA’s 2023 survey 37% of drivers said they had read a text or email while driving in the past 30 days. The two numbers measure different things.

    Source: NHTSA [1] [2]; AAA Foundation [3]
  2. 11 of 169 Interpretation

    With data older than ten years set aside, most of the map is blank

    The only US observations of riders on trains, buses and planes date from 2010 to 2015, so they are left out. What remains is drivers, a few airport items, one study of walkers and one of front-seat passengers, whose video frames show a device in hand 26% of the time.

    Source: UMTRI [4]; NYC DOT [5]; National Academies [6]
  3. 87.0% Hypothesis

    The largest unmeasured block is the passenger seat

    87.0% of US person trips are made in private vehicles, and household vehicles carried 1.5 people on average in 2022. The author expects much of that passenger time to go to phones and audio, but the only US measure found covers front-seat passengers in one city.

    Source: FHWA NHTS 2022 [7]; occupancy as cited in the author’s shared-cabin brief [8]

Fact: stated by the cited source. Interpretation: the author’s reading of the cited evidence. Hypothesis: expected from the evidence but not yet observed. Figures inside every takeaway are sourced.

At a glance

One-page summary

Headline findingThe map of what people do on the move is dense for drivers and mostly blank for everyone else; with data older than ten years set aside, no US measure remains for buses, trains, subways or planes in flight.

Summary

The short version

The benchmark matrix has 169 cells (13 modes and states by 13 activity measures). Only 11 have a US measurement from the last ten years, 2 are rated strong, and nearly all of them are about drivers.

Driver phone use is rare at a moment and common over a month.

4.5% of drivers handled a phone at a stop in 2024, up from 3.0%; 37% say they read a message while driving in a month. Confidence: high on each number, medium on comparing them.[1][2][3][9]

In the car, listening beats looking.

73% of adults who drive or ride use AM/FM; it is 54% of in-car audio time. Confidence: medium.[10][11][12]

Front-seat passengers hold devices more than drivers do.

A device was in hand in 26% of video frames of front-seat passengers (one Michigan city). Confidence: low; one study.[4]

Waiting is measured at airports only.

63% of flyers do something in the terminal (author’s sum) and 70% use airport Wi-Fi. Confidence: medium; one survey.[6]

Most of the map is blank.

Buses, trains, subways, platforms, planes in flight, chargers and parked cars have no US measure from the last ten years. Agency surveys checked do not ask what riders do. Confidence: high that none was found; not proof none exists.[13][14]

Who is on a device

Observed in the last ten years · not like for like

Observed · device in hand or in view
Front-seat passengersUMTRI, video frames
26%
Pedestrians crossingNYC DOT, Queens 2017
13%
Drivers at a stopNHTSA, Jun 2024
4.5%
Stated · past 30 days
Drivers who read a textAAA, 2023
37%

The rows count different behaviors: a device in hand in a video frame (about four minutes each), a pedestrian looking at a device while crossing, a driver visibly handling a handheld device at a stopped vehicle, and a driver saying they did something once in a month. Read as the size of the contrast, not a ranking.[4][5][1]

What cuts against the thesis. The evidence does not show screens taking over the journey. (1) Even for front-seat passengers, about three in four video frames show no device in hand.[4] (2) Drivers’ main in-car activity is radio.[12] (3) Flyers say they pass time by sleeping (60%) about as often as by Wi-Fi entertainment (62%), in a global vendor survey.[15] (4) Missing recent data is not missing behavior; device use on trains, buses and planes may well have risen since the last observations in 2015.
Chapter 1 of 3
The matrix

What is measured, and what is not

Of 169 cells, 11 have a measure from the last ten years and 2 are strong: the map is dense for drivers and blank for most other modes.

How it is built. Rows are 13 combinations of mode and state: moving or stopped, driver or passenger, in a vehicle or waiting. Columns are an “any device” measure plus 12 activity categories (operating, audio, video, calls and messages, social and web, games, work, reading, rest, eating and drinking, companions, errands). Each cell holds the headline figure, shaded by how well it is measured.

Rating rules. Only data collected in the last ten years is used. Strong: a US-wide observation or probability sample from the last three years. Medium: a US-wide survey or dataset with a stated sample or definition; self-selected samples are flagged. Thin: a regional or small sample, or one city. Gap: nothing found. Most figures are prevalence; the only time-based figures are for driver audio and passenger video frames.

StrongMediumThinGap (none found)hover a cell for what it shows
Mode · stateAny device, observed1 Operate or monitor2 Audio3 Video4 Calls, messages5 Social, web6 Games7 Work8 Reading, learning9 Rest10 Eat, drink11 Companions12 Errands, purchases
Car · driver · moving53%task73%·36–37%11%·······
Car · driver · stopped at intersections4.5%···1.9%········
Car · passenger · moving26%··········46%·
Car · stopped to charge·············
Car · parked or waiting·············
City bus·············
Intercity bus·············
Rail (commuter, intercity)·············
Subway and tram·············
Stop, platform, station·············
Plane · in flight·············
Airport · gate, terminal·····70%······63%
Walk, bike, ferry13%············

Cell values are percentages of the people, trips or video frames sampled; units differ by source, so read across a row, not down a column. Sources for every cell are in the table below.

What each row rests on

Mode · stateWhat the sources showSource and sampleRating
Car · driver · moving73% use AM/FM, 48% online audio, 37% podcasts; 36% held and talked on a phone and 37% read a text or email in 30 days; 11% scrolled social media; screen interaction on 52.7% of trips and 116 seconds per driving hourEdison, 2026; AAA, 2023 and 2024; vendor telematics, 2024[10][3][9][16]Medium
Car · driver · stopped4.5% handling a handheld device, 1.9% phone to ear, 0.5% headset; 8.7% handling among ages 16–24NHTSA roadside observation, June 2024, stopped vehicles only[1]Strong
Car · front-seat passengerDevice in hand in 26% of video frames; conversing in about 46%. No US measure of back-seat passengersUMTRI, 306 passengers, 2,733 trips, one Michigan city[4]Thin
Airport · gate, terminal37% go straight to the gate, 45% go to the gate and then backtrack, 18% explore first; 70% use airport Wi-FiPanel of 1,200 flyers, Sep–Oct 2023[6]Medium
Walking13% looking at a device while crossingNYC DOT, 783 pedestrians at three Queens intersections, Nov 2017[5]Thin
All other rowsNo US measure from the last ten years: charging, parked or waiting, city and intercity bus, rail, subway and tram, platforms, planes in flight, bike, ferry—Gap
Chapter 2 of 3
Driving

Rare at a moment, common over a month

At a stop 4.5% of drivers handle a phone; over a month 37% say they read a message while driving; the main thing drivers do is listen to the radio.

The observed rate is low and rising. NHTSA’s roadside survey counts drivers at randomly selected intersections in daylight in June. In 2024 it found 4.5% visibly handling a handheld device, up from 3.0% in 2023 (a significant rise), and 1.9% holding a phone to the ear.[1]

The stated rate is high. In AAA’s 2023 survey 37% of drivers said they had read a text or email while driving in the past 30 days, 27% had typed one and 36% had held and talked on a phone.[3] The 2024 edition reports 37% reading and 11% scrolling social media.[9] Both can be true: they measure a moment and a month.

Moving traffic is seen only through telematics. The roadside survey sees only stopped vehicles. A 2024 telematics report finds phone-screen interaction on 52.7% of trips and 116 seconds per driving hour, about 3.2% of drive time by the author’s arithmetic, down from 134 seconds in 2022. Its drivers are enrolled in insurance programs and no driver count is given.[16] So a few percent of the time, on most trips, can sit beside 37% in a month.

Listening is the main activity. Edison’s 2026 survey finds 73% of adults who drive or ride use AM/FM, 48% online audio and 37% podcasts.[10][11] In time, AM/FM took 83% of ad-supported in-car audio in the fourth quarter of 2025, and 54% once ad-free listening is counted.[12] Phones carry 29% of in-car listeners.[17]

Audio sources used in the car

% of US adults who drove or rode in a car in the past month · Edison Infinite Dial · Jan 2026

AM/FM radio
73%
Online audio
48%
Podcasts
37%
SiriusXM
24%

Sources are not mutually exclusive, so shares sum to more than 100%. The base is worded as drivers and riders in the slides and as drivers in the trade report, so the passenger share cannot be separated.[10][11]

Why it matters. Audio needs neither hands nor eyes; the part of driver attention that does, handling a phone, is growing.[1]

Reading the evidenceHigh confidence that phone handling at stops rose from 2023 to 2024; medium confidence on the stated rates; low confidence for moving traffic, where the only recent source is insurance-program telematics.
Chapter 3 of 3
Riding, waiting and the gaps

Where the hands are free and the data are missing

Passengers, riders and people waiting are the least measured: one video study of front-seat passengers, one survey of airports, and nothing recent for buses, trains, subways or planes in flight.

Car passengers: one study. University of Michigan researchers filmed front-seat passengers in 75 vehicles, coding 13,638 frames from 2,733 trips by 306 passengers, each about four minutes of travel. A device was in hand in 26% of frames and passengers were conversing in about 46%. The drivers were volunteers in one small city and the collection year is not stated (the report is dated February 2020).[4] Private vehicles carry 87.0% of US person trips, so this is the largest block of travel time.[7]

Buses, trains, subways: nothing from the last decade. The last US observations of riders found, in Chicago from 2010 to 2015, are older than the ten-year limit and left out. Two recent agency surveys checked, NYC Ferry (2022) and four North Carolina bus systems (2022–23), ask about trip purpose and apps, not what riders do on board.[13][14]

Airports: most explore. In a 2023 panel survey of 1,200 frequent flyers, 37% went straight to the gate and 63% did something in the terminal first (the author’s sum), and 70% used airport Wi-Fi.[6] In the air, an 11-country Viasat survey has flyers saying they would use Wi-Fi for entertainment (62%) or sleep (60%); it has no US breakdown, so it is not in the matrix.[15]

Walkers: one city. A New York City DOT observer counted 13% of 783 pedestrians at three Queens intersections as looking at a device in 2017.[5] National time-use statistics do not help: the American Time Use Survey folds travel into each purpose category.[18]

What would close the gaps (the author’s suggestion). Add a “what did you do during the trip” question to travel diaries and transit on-board surveys, or repeat observation on buses, subways and platforms.

GapWhy it mattersWhat would measure it
Back-seat and other car passengersLargest block of travel time; 87.0% of person trips are in private vehicles[7]An activity question in NHTS-style diaries; in-car observation with consent
Bus, rail, subway, platformSeated riders with free hands; the last observations are from 2015Repeat observation; activity items in agency on-board surveys
Plane in flight, charging, parkedLong dwell times with no US-only activity figuresAirline and charger-network data with stated definitions
Time share, all modesOnly driver audio and passenger video frames have time-based figuresTime-use diaries with a travel-activity module
Reading the evidenceLow confidence for car passengers (one study); medium for airports (one survey); high that the other rows have no recent US measure in the sources searched.

Method, limits and flags

Read before citing

Question. What do people in the US do while traveling, in each mode and in motion versus stopped, and where is that behavior measured? The report builds a matrix, fills each cell from the best US source found, and rates the evidence rather than estimating the missing cells.

Scope. United States only, by choice of the author, and only data collected in the last ten years (since October 2016). Modes: car (driver and passenger, including charging and waiting), bus, rail, subway and tram, plane and airport, with walking, biking and ferries as an edge tier. It covers people’s activity, not crashes or safety outcomes.

Sources. The source type follows the claim. For how people behave, government observation and statistics, then peer-reviewed studies and agency reports, then independent panels, with vendor data as support and trade press to trace the original. Each reference carries its type and date. Observed, stated and company-claimed figures are labeled and kept apart. Searches ran on 21 October 2026 across government reports, journals, transit agency surveys and trade press, and were not a systematic review; “none found” means not found, not absent.

What this report is not. Confidence levels and the strength ratings are the author’s assessment of the evidence, not statistical measures. The “what would close the gaps” suggestions are the author’s judgment, not forecasts.

Flags and open questions

  1. Older data left out. Observations of Chicago-area rail riders (2010–2015), intercity bus, Amtrak and airline passengers (2014), a Fargo campus bus survey (2010), a Northern California commuter survey (fielded 2011–12), the SHRP 2 driving video (2010–2013) and a Seattle pedestrian study (2012) were found and excluded by the ten-year rule. Including them would fill several rail, bus and plane cells; that is a choice for a later edition.
  2. Not like for like. Observation studies count a device at a moment or in a frame; AAA counts people who ever did something in a month; Edison counts reach, and one source gives minutes. Cells are comparable along a row, not down a column.
  3. Roadside survey limits. NOPUS sees only stopped vehicles at intersections in daylight in June, and counts what is visible. It is not a measure of moving traffic or of crash risk.
  4. Passenger video study. UMTRI’s drivers were volunteers who regularly carry passengers, in one small city; 72% of passengers were women against about 62% nationally; its text gives 26% in a table and 27% in the results; “device in hand” counts holding, not necessarily using. The collection year is not stated, so it may be near the ten-year edge; the author kept it, rated thin.
  5. Telematics. Cambridge Mobile Telematics covers drivers enrolled in insurance programs, gives no driver or trip count in the preview read, and counts screen touches only. It is rated medium with that caveat; the 3.2% of drive time is the author’s arithmetic.
  6. Edison in-car figures. The slide and a trade report describe the base differently (adults who drove or rode in the past month; Americans who drive), so the audio figure sits in the driver row and cannot be split from passengers. The 83% and 54% radio time shares are ad-supported audio unless stated, relayed by trade press.
  7. Airport survey. A non-probability panel weighted to the census; coverage is presumed to be the US. The source text says 45% go straight to the gate while its figure shows 45% for “go to the gate, then backtrack” and 37% for straight to the gate; the figure is used. The 63% is the author’s sum.
  8. Viasat. Commissioned and published by Viasat, 11 countries; the US sample is 1,515 but no US-only results are published, so it is quoted in the text only. The country table on page 28 sums to 10,712 against 11,213 stated, and the report does not explain the gap.
  9. Pedestrians and agency surveys. The New York study covers three Queens intersections. Other pedestrian studies found were older than ten years, from a single college campus, or could not be opened. The two transit agency surveys are examples of what agencies ask, not a census of surveys.
  10. Scoring. The strength thresholds (three years, ten years, sample type) are the author’s choice. A different cut-off would move several cells.

Change log

  • October 2026 edition
  • Added
    First edition, data as of 21 October 2026. Matrix of 13 modes and states by 13 activity measures, with US sources from the last ten years for 11 cells.

Related reports on this site

One cabin, several people sizes the passenger-hours gap that this report cannot fill with activity data. In-Car AI Assistant Benchmark covers the assistants that sit in the driver’s seat of the activities counted here.[19]

About this report

Personal research by Chuhee Lee. Research, source verification and drafting were carried out with AI models used as a research team: Claude (Anthropic), Gemini (Google) and ChatGPT (OpenAI). Every factual claim is sourced to the references listed below; topic selection, framing, ratings and conclusions are the author’s. These are personal research projects; the views are the author’s own and do not represent any employer or organization. No company named sponsors, reviews or endorses this report, and product names and trademarks belong to their owners. Licensed under CC BY-NC 4.0.

References

Continue down the map

ForcesShiftsSignalsOpportunities
Follow new editions

Each report is refreshed on a schedule. Know when it changes.

Next refresh dates are listed on the home page.