Executive Summary The Numbers Insurance Trap Vehicle Surveillance Wage Theft Algorithmic Firing Externalized Costs The Math Direct Plea Sources
Compiled May 2026 · Documented Case

The Gig Delivery Trap.

A documented case against DoorDash and personal-vehicle gig work, with a direct plea to the drivers still in it.

SubjectDoorDash · Uber Eats · Grubhub · Instacart
Evidence baseState AGs · IRS · Berkeley · EPI · NBER · HRW · FTC
Net wage finding$5.49/hr after expenses (Denver-Boulder)
Executive Summary

The business model depends on a single mathematical lie.

DoorDash, Uber Eats, Grubhub, Instacart, and the rest of the personal-vehicle gig delivery industry have constructed a business model that depends on a single mathematical lie: that the cash a driver receives from the app is the same as the wage that driver actually earns.

It is not.

When the IRS's own published cost of vehicle operation, the gap in personal auto insurance, the self-employment tax burden, the absence of workers' comp, unemployment, sick leave, and health benefits, and the unpaid waiting time between orders are all subtracted from gross app earnings, the surviving figure for the median DoorDash driver is below the federal minimum wage. In several major metros it falls below half of the local minimum wage. In Denver-Boulder, a peer-reviewed Colorado study put it at $5.49 per hour.

The platforms know this. The independent researchers know this. State attorneys general have proven wage theft in court and collected multi-million-dollar settlements. Human Rights Watch has classified the industry's algorithmic management practices as a human rights concern. The IRS publishes a per-mile cost figure that, when applied honestly, shows that a substantial fraction of low-paying orders cost the driver money to complete.

The remaining piece, the one each driver has to settle for themselves, is the decision to stop participating.

This report lays out the published evidence. The closing section is a direct plea to the people still driving.

Part 1 · The Numbers

The headline numbers are a lie.

What DoorDash and Gridwise report

Gridwise, the largest gig-driver data aggregator, reports the median 2026 DoorDash gross earning at roughly $12 to $15 per hour, with the top quartile pulling $13.49 or more and the top decile crossing $15.63. These are gross figures, before vehicle expenses and self-employment tax.

DoorDash's own promotional material and Glassdoor's voluntary submissions push higher figures, in the $20 to $25 per hour range. These come from active-time-only reporting, do not deduct vehicle cost, and rely on self-selected samples skewed to drivers who have not yet quit.

What independent researchers find after deducting expenses

Peer-reviewed
CU Boulder MedLab · The Guardian
$5.49
Average wage per hour for Denver-Boulder drivers after expenses. Denver minimum wage at the time: $15.87.
2024
Berkeley Labor Center · 5 metros
$9.87
Median net hourly earnings outside California, including tips. Engaged time only, not actual hours given.
2018
Economic Policy Institute
$9.21
W-2-equivalent wage for Uber drivers after fees, vehicle expenses, and the self-employment tax penalty.
National 2022
EPI national gig survey
29%
Earned less than the state minimum wage that would apply under W-2 classification. 14% below federal floor.
National 2022
EPI national gig survey
19%
Went hungry because they could not afford to eat. 30% used SNAP in the prior month, twice the comparable rate.
NBER · Mar 2026
Seattle minimum pay floor study
2.3×
Average base pay per task when Seattle imposed a floor: $5.37 → $12.52. Measurement of how far below livable the baseline was.
2020 survey
Industry / advocacy
20%
Of gig drivers were losing money to drive given their other expenses. 45% could not handle a $400 emergency without borrowing.
2026 official
Internal Revenue Service
72.5¢
IRS-published cost per mile of vehicle operation. Fuel, depreciation, insurance, maintenance, repairs, tires, registration.

The IRS does the math the platforms refuse to do

The IRS sets the 2026 standard business mileage rate at 72.5 cents per mile. This number is not a tax dodge or an inflated estimate. It is calculated annually by an independent contractor (Motus) using a national pool of actual vehicle operating data. It reflects fuel, depreciation, insurance, maintenance, repairs, tires, registration, and the full fixed-and-variable cost of keeping a vehicle on the road.

Of that 72.5 cents, the IRS allocates 35 cents to depreciation alone, the wear-out of the asset itself.

$
A typical DoorDash delivery is around three to five miles of paid distance plus unpaid positioning and return. If a driver puts down ten total miles to earn a $7 order, the IRS-published cost of those miles is $7.25. The driver has done forty-five minutes of work to lose a quarter, then owes self-employment tax on whatever paper profit the bookkeeping shows.
IRS 2026 standard mileage rate, applied honestly

A full-time DoorDash year of 30,000 to 45,000 miles produces an IRS deduction of $21,750 to $32,625. That deduction is not a bonus. It is the federal government's published estimate of what those miles actually cost the driver in fuel, depreciation, and operating expense, recouped against income tax. The deduction does not put cash back in the driver's hand. The vehicle still wore out by exactly that much.

Part 2 · The Insurance Trap

This is the part the platforms work hardest to obscure.

The livery exclusion

Every standard personal auto policy contains a livery exclusion. It voids coverage when the vehicle is used to carry people or property for a fee. The moment a driver taps "accept" on a DoorDash order, that exclusion fires.

Per a 2026 LegalClarity analysis, the gap is real and financially devastating. The personal insurer can deny the claim, cancel the policy, and decline to renew. The driver then carries a non-renewal mark that follows them across insurers and raises their rate for years.

Approximately 90 percent of DoorDash drivers operate without commercial coverage or a rideshare endorsement, per industry reporting. The vast majority are one accident away from a catastrophic uncovered loss.

What DoorDash actually provides

DoorDash advertises a $1 million liability policy. The fine print:

01
It is excess coverage. The driver must first file with their personal insurer and be denied. The denial itself is the trigger for DoorDash coverage to engage. The driver must produce a denial letter.
02
It only covers third-party damages. It pays for what the driver did to someone else's car or body. It pays zero for the driver's own vehicle, the driver's own injuries beyond a narrow occupational accident policy, or lost income.
03
It only applies during "Delivery Service" period, meaning food is in the car. The "Delivery Available" period, when the driver is logged in and waiting, is uncovered in most states. The driver hit at a red light while waiting for an order has no DoorDash coverage and a personal policy that just voided.
04
The denial requirement creates an adversarial loop. To access DoorDash's policy, the driver must first burn their personal coverage on a denied claim, which the insurer then uses as grounds to non-renew.

A rideshare or delivery endorsement, where available, runs $15 to $50 per month additional, which most low-earning drivers cannot absorb. Full commercial auto insurance runs $100 to $300 more per month, which is impossible at these earnings.

Part 2.3 · Vehicle Surveillance

The vehicle itself is now an informant.

Since roughly 2018, and at near-universal saturation in model years 2020 and newer, virtually every passenger vehicle sold in the United States has shipped with an embedded telematics module in constant cellular communication with the manufacturer. OnStar (GM), FordPass, Toyota Connected Services, Hyundai Bluelink, Kia Connect, Honda Link, Stellantis Uconnect, and the equivalent systems at every other major automaker generate a continuous stream of behavioral data: hard braking events, sharp acceleration, cornering force, speed, geolocation, time of day, route, and trip duration. For a DoorDash driver, this is a near-perfect surveillance record of every order taken, the long-distance no-tip runs included.

The data does not stay with the manufacturer.

A March 2024 New York Times investigation, followed by enforcement action from the Federal Trade Commission, established that the automakers were packaging this data and selling it to two specialized insurance industry data brokers, LexisNexis Risk Solutions and Verisk Analytics. Those brokers in turn sold individualized driving-behavior reports to insurance carriers, who used the reports to raise premiums, decline applications, and non-renew existing policies. In the GM case alone, as many as eight million drivers had data routed through this pipeline.

"
The named plaintiff in the leading class action, Romeo Chicco, was rejected by seven consecutive insurers after his 2021 Cadillac XT6 generated a LexisNexis driver-behavior report containing 258 individual recorded "driving events," including sharp accelerations, hard braking, and speed flags. He had not knowingly consented to any of it.
Chicco v. GM / LexisNexis · ArentFox Schiff coverage, May 2024

The FTC reached a proposed settlement with GM and OnStar in January 2025; GM paid $12.75 million in a follow-on settlement in 2026; the Texas Attorney General has separately sued Allstate's data subsidiary Arity, naming Toyota, Lexus, Mazda, Chrysler, Dodge, Fiat, Jeep, Maserati, and Ram as data sources in that pipeline.

For a gig delivery driver, the implications compound the insurance trap already described:

01
The behavioral signature of delivery work is exactly the signature insurers price as high-risk. Hard braking at apartment complexes. Sharp acceleration to make the next pickup window. Late-night driving. Frequent short trips with cold-engine starts. Repeated stops at restaurant addresses. The algorithm cannot tell the difference between aggressive driving and delivery driving, and would have no reason to, since both correlate with higher claims frequency.
02
The data is sold regardless of whether the driver has opted into a telematics insurance program. Progressive Snapshot and State Farm Drive Safe & Save are voluntary discount programs the driver knows about. The OnStar-to-LexisNexis-to-Geico pipeline is not. Many of the affected drivers had no idea their car was reporting on them, and could not have, because the disclosures were buried in app terms-of-service flows or activated by "dark patterns" at the dealership.
03
The mark stays attached to the driver, not the vehicle, for up to seven years on a LexisNexis profile. A driver who sells the car and quits delivery still carries the report. Future insurers see it. The premium increase persists.
04
The driver who is then denied coverage, or non-renewed after a delivery accident, attempts to shop for replacement coverage and finds that the LexisNexis file precedes them at every carrier. The combination of a non-renewal plus a telematics-derived high-risk score is, for many drivers, the end of personal vehicle ownership at any affordable rate.
The structural picture: the driver's own car, financed in the driver's name, insured at personal-policy rates, is acting as a continuous evidence-gathering device against the driver. The data is being sold to brokers who sell it to insurers who use it to raise the driver's rates or deny coverage entirely. The platform does not pay for any of it. The platform did not even have to set it up. The automaker built the surveillance, the broker built the market, the insurer built the pricing model, and the gig platform reaps the entire benefit of the labor while the driver absorbs every cost across the chain.
The compound trap

A driver who started delivering in 2020 in a brand-new car may now, five years in, be functionally uninsurable in the personal market because of telemetry their own car generated while they were working. There is no analog for this in W-2 employment. A pizza delivery driver employed by Domino's is covered by Domino's commercial fleet policy. A DoorDash driver in a personally-titled 2021 vehicle is the policy, the surveillance subject, and the financial loser, all at once.

The structural result

The driver is paying full commercial cost of vehicle operation, generating commercial-grade revenue for the platform, but is uninsured at commercial standards and is being surveilled by the vehicle itself. The platform externalizes the insurance risk onto the driver and the driver's family. The automaker externalizes the data-collection cost onto the driver while monetizing the data stream. The broker takes a margin. The insurer raises the rate. When the accident happens, both parties involved can be financially destroyed, with the platform legally insulated and the data file still active on the driver's name for seven years.

Part 3 · Wage Theft

This is not allegation. This is settled fact in two jurisdictions.

State enforcement actions on the record

SETTLED
Washington, D.C. · 2019-2020 · $2.5 million
D.C. Attorney General v. DoorDash
Sued under the Consumer Protection Procedures Act, alleging the company misrepresented to customers that tips would go to the workers. From 2017 to 2019, customer tips were instead used to subsidize DoorDash's guaranteed base pay obligation. The platform pocketed the difference. DoorDash paid $2.5 million in settlement, including $1.5 million in restitution to drivers.
New York State · February 2025 · $16.75 million
NY AG Letitia James v. DoorDash
Same model. Between May 2017 and September 2019, DoorDash used customer tips to offset, not supplement, the guaranteed base pay it had promised drivers. The customer-facing message read "Dashers will always receive 100 percent of the tip." The tip was technically delivered, on paper, by being routed first to cover the wage DoorDash had already promised. The base pay obligation was eliminated by the customer's generosity. Sixty-three thousand New York drivers became eligible for restitution.

Pattern, not incident

These were two state AG enforcement actions, in jurisdictions with active labor enforcement bureaus. The model operated nationally during the same period. Drivers in the other forty-eight states received no restitution. The pay model was changed in 2019, but the underlying business approach, of treating algorithmic wage opacity as a feature, continues. The Tech Policy Press analysis published January 2026 identifies algorithmic wage determination, surveillance-based behavior scoring, gamified bonus structures, and algorithmic deactivation as ongoing extraction mechanisms with no resolution.

Part 4 · Algorithmic Termination

Fired by software, with no human review, no appeal that matters.

Human Rights Watch published "The Gig Trap: Algorithmic, Wage and Labor Exploitation in Platform Work in the US" in May 2025, covering Amazon Flex, DoorDash, Favor, Instacart, Lyft, Shipt, and Uber. HRW's central finding: these platforms use algorithmic systems to extract labor while evading the legal obligations that would attach to employment.

Deactivation is the apex of the leverage. A driver who has spent two years building rating, completion rate, and tenure can be terminated by an automated decision based on a customer complaint or a pattern flag, with no human review, no opportunity to face the accuser, and no meaningful appeal. The appeal, when it exists, is to the same opacity that produced the decision.
Human Rights Watch · May 2025

University of Washington research cited by Cornell ILR found that drivers of color were reactivated at higher rates than their peers after deactivation, which the researchers read as evidence that the original deactivations were producing discriminatory outcomes, whether by customer-rating channels or by algorithmic proxy.

New York City has proposed just-cause protections (Int. 276 and Int. 1332) modeled on the city's 2021 fast-food workers' law. As of this writing, no national framework exists. The driver remains terminable on no notice, by software, with their income gone the same day.

For a worker carrying car-loan debt, fuel costs, and the lingering insurance non-renewal from an earlier accident, this leverage is total. The platform sets the terms, the platform changes the terms, and the platform terminates without explanation. The driver bears every risk.

Part 5 · The Externalized Costs

What an employer pays. What the platform doesn't.

A W-2 employee's labor cost to an employer includes, by federal and state law:

Federal mandate
Employer half of FICA
7.65% Social Security and Medicare. The DoorDash driver pays both halves: 15.3%.
State mandate
Unemployment insurance tax
No unemployment coverage if deactivated. Income disappears the same day.
State mandate
Workers' compensation
None, beyond a narrow occupational accident policy during active delivery.
State / municipal mandate
Paid sick leave
None. Driver carries the cost of every sick day directly.
Federal / state mandate
Rest break time
None paid. Waiting time between orders is unpaid by design.
State mandate (varies)
Disability insurance
No employer contributions. No coverage.
ACA framework
Employer health coverage
None. Driver carries individual market premium or is uninsured.
Operational
HR, benefits admin, compliance
Fixed overhead the platform avoids by classifying the worker as a vendor.

The Berkeley Labor Center quantified the value of the mandated benefits package the driver does not receive. The result: a properly costed delivery hour, if treated as employment, would carry several dollars per hour of additional employer cost that the platform retains as margin.

=
DoorDash is profitable because it pays no benefits. The drivers are insolvent because they pay all the costs.
The whole business in one sentence
Part 6 · The Survival Math, Worked Honestly

A representative full-time week, on paper.

Mid-range figures. Forty engaged hours. Eight hundred business miles. IRS 2026 standard rate.

DASHER
Weekly · 40 hrs engaged
Gross app earnings (40h × $12) +$480.00
Business miles driven 800 mi
IRS-published cost (800 × 72.5¢) -$580.00
Self-employment tax on net N/A
Health insurance (no employer) $300-700/mo
Insurance gap exposure CATASTROPHIC
Unpaid waiting time UNCOUNTED
WEEKLY NET -$100.00
Paper loss after IRS-published vehicle cost. The cash deposit hides this because the deduction is taken at tax time, not at the fuel pump.

The accounting collapses when the vehicle dies.

The driver in this scenario has worked forty hours, taken on commercial risk, run depreciation against an asset that will need replacement, and finished the week at a paper loss. The cash deposit hides this because the IRS deduction is taken at tax time, not at fuel-pump time.

A replacement vehicle, financed at subprime rates because the driver's credit has eroded from working for cash flow rather than profit, locks in another five years of payments against the same broken math.

This is the trap. The cash flow looks survivable until the asset wears out, the insurance gap fires, the algorithm deactivates, or the body fails. Then it collapses all at once.

Part 7 · A Direct Plea

To the driver reading this.

If you are still driving for DoorDash or any of the others, I am writing this to you specifically.

You already know something is wrong. You have known it for a while. The cash that comes in does not match the work you put out, and you have spent enough nights doing the arithmetic in your head between orders to feel the shape of it, even if you have not written it down.

Here is the shape.

You are not making minimum wage. You are running the depreciation of your car against the gross of your earnings, and the math is not closing. The IRS knows it. Their published cost-per-mile is the federal government's own number for what your driving costs you. When you compare that number to your gross app earnings divided by your miles, you are watching yourself work for free, and sometimes for less than free.

You are uninsured, in a real sense, every hour you are on the app. The $1 million liability the platform advertises will not pay until your personal insurer has denied your claim, which they will, and at the moment they deny it they will also non-renew you, which means your next policy will be a substandard policy at a punitive rate, if you can get one at all. You are one accident from being unable to legally drive.

If your car is a 2020 or newer model, the car itself has been reporting on you the whole time. Every hard brake, every late-night run, every sharp acceleration is being logged by the telematics module the manufacturer installed, then sold to data brokers like LexisNexis and Verisk, then sold to insurance companies. Eight million GM drivers were caught up in just one pipeline. The FTC has settled with GM, the Texas AG has sued Allstate's data broker subsidiary, and a class action plaintiff found 258 driving events on his personal LexisNexis report he never knew existed. The behavior pattern that delivery work produces, frequent stops, hard braking at apartments, fast lane changes to make pickup windows, looks exactly like the pattern insurers price as high-risk, and the file follows you for up to seven years. The car you are driving for them is the same car being used to make you uninsurable.

You have no recourse if you are deactivated. The platform can terminate your income today, with no warning, no human reviewer, and no explanation that you can challenge. There is no unemployment for you on the back side. The flexibility the company sells you is real only until the algorithm decides otherwise, at which point the flexibility is theirs and the consequences are yours.

You are paying the employer half of your own payroll tax. You are paying your own health insurance, or you are uninsured. You are paying your own workers' comp, which is to say, you have none. The thing the platform calls a "1099" is the legal mechanism by which it has externalized every cost of employment onto your back while keeping the revenue.

You are not lazy if you stop. You are not weak. You are not failing.

The math is the math. It was rigged before you got there. The settlements the New York and D.C. attorneys general won against DoorDash are evidence on the record that the company has, in plain English, stolen wages from drivers. Human Rights Watch has named the industry as a human rights problem. The Economic Policy Institute has documented that something close to a third of gig workers are below their state's minimum wage. Twenty percent of drivers were losing money to drive even before the inflation of the last few years. You are not imagining what you feel.

There is no version of this that gets better by working harder. Acceptance rate manipulation, top dasher status, peak pay, none of it closes the gap between gross app earnings and the real cost of the work. Drivers who cherry-pick orders do somewhat better than drivers who take everything, but the ceiling is still set by a model designed to keep you below it.

If you are doing this because you have no other option

That is a real situation and it deserves a real answer, not a slogan. The answer is not more hours behind the wheel. The answer is anything that returns one of three things to you: a wage rather than a fee, your time outside the app, or an asset that does not depreciate while you sleep.

That can be a W-2 job at minimum wage, which after employer-side benefits and the absence of vehicle depreciation will pay you more than full-time DoorDash, in real dollars, almost everywhere. It can be skilled trades training through a union apprenticeship, where you are paid to learn. It can be temp-to-perm warehouse work, which is brutal but at least pays its own overhead. It can be a part-time job that meets your minimum need plus enough rest to plan a next move.

It can be applying for the benefits you have already paid into and have been told you are not eligible for. SNAP. Medicaid. Disability if you qualify. Unemployment if you were misclassified. These are not handouts. They are the social insurance system the platforms have specifically constructed their model to avoid contributing to, with money you and people like you have already paid in.

The flexibility the app sells you is the same flexibility you have walking out of any low-wage job in America. The app did not invent it. The app just used it to make you feel like leaving was the same as quitting on yourself. It is not.

The car you are driving is the most valuable asset most working people own. The platform is consuming it on a schedule designed to keep you running long enough to wear it out, then walking once it dies, then watching you come back with a financed replacement.

That is the engine of the business. Your asset is the fuel.

Stop being the fuel.

If you can stop today, stop today.

If you cannot stop today, set the date you stop. Pick one. Write it on something you will see. Use the time between now and then to apply for everything else, to file every complaint that is owed, to claim every settlement payment you are eligible for, to document every mile and every accident report and every deactivation appeal that has ever been answered by silence.

Then walk.

The work you do after you walk will be hard. Most work is. The difference is that the hours you spend on it will be hours that pay you, not hours that subsidize someone else's billion-dollar valuation with the metal of your own car.

You are worth more than the algorithm has been pricing you at. Everyone running these numbers, in the academic studies, in the AG offices, in the human rights reports, has come to the same conclusion. You are not failing the gig. The gig was the failure. It was designed that way.

Walk.

The work you do after you walk will be hard. Most work is.

The difference is that the hours you spend on it will be hours that pay you.

Sources

Gridwise · 2026 How Much Do DoorDash Drivers Make in 2026? Aggregated data from 115,771 tracked drivers.
gridwise.io/blog/how-much-do-doordash-drivers-make
CU Boulder MedLab New Study Reveals Conditions among Colorado Rideshare Drivers. $5.49/hr after expenses, Denver-Boulder.
colorado.edu/lab/medlab
Berkeley Labor Center · 2024 Gig Passenger and Delivery Driver Pay in Five Metro Areas.
laborcenter.berkeley.edu
Economic Policy Institute · 2018 Uber and the Labor Market, Lawrence Mishel. $9.21/hr W-2 equivalent.
epi.org/publication/uber-and-the-labor-market
Economic Policy Institute · 2022 National Survey of Gig Workers Shows Poor Working Conditions and Low Pay.
epi.org/press/national-survey-of-gig-workers
Human Rights Watch · May 2025 The Gig Trap: Algorithmic, Wage and Labor Exploitation in Platform Work in the US.
hrw.org/report/2025/05/12/the-gig-trap
NBER · March 2026 Impact of Minimum Pay Rules on Gig Delivery Drivers. Seattle base pay doubled with floor.
nber.org/digest/202603
NY AG · February 2025 AG James Secures $16.75 Million from DoorDash for Cheating Delivery Workers Out of Tips.
ag.ny.gov · 16.75 million settlement
D.C. AG · 2020 DoorDash tip diversion, $2.5 million settlement.
mselaborlaw.com · DC settlement
IRS · December 29, 2025 IRS Sets 2026 Business Standard Mileage Rate at 72.5 Cents Per Mile. IR-2025-128.
irs.gov/newsroom/IR-2025-128
LegalClarity · April 2026 What Insurance Companies Cover Delivery Drivers? Livery exclusion analysis.
legalclarity.org/delivery-drivers
Insureon · December 2025 DoorDash Driver Insurance: Auto Insurance & More. Coverage gap and denial requirements.
insureon.com/doordash-drivers
Cornell ILR / Carow Policy Center Just Cause NYC Gig Workers Provides Human Review of Algorithmic Firings.
ilr.cornell.edu/carow/just-cause
Tech Policy Press · January 2026 Why Gig Platform Wage Theft is a Governance Crisis.
techpolicy.press · governance crisis
National Institute for Workers' Rights · 2020 The Gig Economy By The Numbers. 20% losing money, 15% on public assistance.
niwr.org · gig economy numbers
FTC · January 16, 2025 Proposed settlement order with GM and OnStar for selling connected vehicle data. Coverage: WardsAuto, Jan 23, 2025.
wardsauto.com · FTC/GM proposed order
InsureMojo · 2026 GM Pays $12.75M Over Driver Data Sold to Insurers.
insuremojo.com · GM settlement 2026
Motley Rice · March 2026 GM Selling Driver Data Through OnStar. 8M drivers affected through pipeline.
motleyrice.com · GM driver data
ArentFox Schiff · May 2024 FTC Makes Statement on Connected Vehicle Data. Chicco v. GM/LexisNexis, 258 events, 7 insurer denials.
afslaw.com · FTC connected vehicle data
Lutzker & Lutzker · April 2025 Data Sold to Auto Insurers Without Drivers' Knowledge or Consent. Texas AG investigation; suit vs Arity.
lutzker.com · data sold to insurers
Insurify via Insurance News Net High Car Insurance? Your Car Maker May Be Selling Your Driving Data. Markey FTC referral, 130-page driver file.
insurancenewsnet.com · Insurify