What the IRS actually wants in a mileage log

By Sabett, founder of MileGravy ·September 6, 2026 ·6 min read

I kept my mileage in a spiral notebook on the passenger seat for years. It wasn't elegant. It would have survived an audit, which is more than I can say for most of the systems I see drivers using.

Here's what the record actually has to do.

The four things every entry needs

Publication 463 requires your records to establish the time, place and business purpose of your travel. For car expenses that comes down to four fields per trip:

1. Date. The actual calendar date you drove.
2. Miles. The business miles for that trip.
3. Destination. Where you went — a street, a business name, a town.
4. Business purpose. Why. "DoorDash delivery," "pickup at Kroger," "repositioning to zone."

That's the whole requirement. Not a GPS trace, not a photograph of the odometer, not a notarized anything. Four fields, per trip, written down.

The date field carries more weight in 2026 than it used to. The business rate changed mid-year — 72.5¢ through June 30, 76¢ from July 1 — so a log without dates can't be converted into a deduction at all. I wrote about how to split a 2026 year separately.

"Timely kept" is the word that does the work

Publication 463 talks about timely kept records. The idea is that a record made at or near the time of the trip is credible in a way that a record made months later isn't. Your memory of Tuesday is good on Tuesday. It is not good in April.

In practice, timely means:

• Written down during the shift, at the end of the shift, or at worst that night
• Not batched up weekly from memory
• Definitely not assembled at tax time from whatever you can piece together

A log written the day of, with odd numbers and real destinations, reads as true. A log where every week is exactly 600 miles reads as invented, because it is.

Why "I'll reconstruct it from the app history" fails

This is the most common plan I hear, and it's the one that falls apart fastest.

The platform's history is not your mileage log, for three reasons.

It doesn't have your miles. DoorDash knows the distance it paid you for — usually a straight-line or route estimate from pickup to drop-off. It does not know the miles you drove to get to the pickup, the miles you drove repositioning between orders, or the route you actually took. Those are a large share of your real business miles, and they're the ones you'd be dropping.

It's not yours and it doesn't last. Platforms trim history. Accounts get deactivated. Apps get rewritten. If an examiner asks about 2023 in 2026, "let me log in and check" is not a plan — it's a hope about a company's data retention policy.

It isn't contemporaneous, and that's visible. A spreadsheet you built in April from app screenshots is a reconstruction. It may even be accurate. It still doesn't meet the standard Publication 463 describes, and an examiner can tell the difference at a glance.

Use the app history to cross-check your log. Don't use it as the log.

What a failed log actually costs

Here's the arithmetic drivers don't run until it's too late. Take a driver with 25,000 business miles who claims them all at 76¢ and has no usable log.

The accuracy-related penalty is "20% of the portion of the underpayment of tax," and it applies for negligence or disregard of rules as well as substantial understatement. Interest accrues on top, from the original due date.

Seven thousand dollars, because of a notebook you didn't keep. And notice what the log would have cost you: two minutes a day.

What "adequate records" means

Publication 463 describes adequate records as an account book, log, diary, trip sheet or similar record, together with documentary evidence such as receipts. Note what's on that list: a notebook counts. A spreadsheet counts. An app counts. The IRS has never cared what your log is made of — it cares that it exists, that it was kept as you went, and that it contains the four elements.

What actually separates the systems is whether you'll keep using them at 10pm in a parking lot. The best log is the one that's still complete in December.

• Paper notebook. Free, instant, never crashes, no battery. Has to be transcribed to total it up, and it lives in a vehicle that can get broken into.
• Spreadsheet. Totals itself, easy to split by date for the 2026 rate change. Requires you to actually open it, which is where most people fail.
• An app that records as you drive. Contemporaneous by construction, which is exactly the standard being asked for. Make sure it exports — a log you can't get out of an app is a log you don't really own.

Can you log part of the year and extrapolate?

There is a narrow provision for this. Publication 463 allows records covering part of a year to prove business use for the entire year, but the conditions are tight: the records have to have been kept as required, and the period covered has to be representative of the year.

People read that and hear "I only have to track three months." That's not what it says, and for delivery driving it's a bad bet. Our volume swings hard with weather, holidays, school terms and promotions. A representative three months is difficult to establish and easy for an examiner to challenge — my October and my February don't look anything alike.

Treat it as a lifeboat for the year your records got destroyed, not as a tracking strategy.

What an examiner actually does with your log

They don't read all of it. They cross-check it.

Expect your log to get held up against the 1099s the platforms filed — do the days you claim to have driven line up with the days you were paid? Against your odometer, from oil change and inspection records — does your annual total fit inside the miles the vehicle actually turned? Against the calendar — did you log 200 business miles on Christmas Day?

This is why a real log survives and a manufactured one doesn't. A log you kept as you drove will agree with all of that evidence automatically, because it came from the same reality. A log you built backward agrees with whichever source you built it from and nothing else.

It's also why recording your odometer at the start and end of the year matters. It gives you the total-miles figure that makes your business-miles figure believable, and it's two numbers a year.

Keep it after you file

The log is not done when the return goes in. The IRS's guidance on how long to keep records ties the retention period to the period of limitations for the return — generally three years from filing, longer in several situations.

Practically: keep at least three years, and keep them somewhere that isn't only in the vehicle. Photograph the notebook pages. Export the spreadsheet. If your log lives in one app on one phone, it is one dropped phone away from not existing.

What a good entry looks like

9/6/2026 — 14.3 mi — Kroger, Carpenter Rd → customer, Golfside Dr — DoorDash delivery

9/6/2026 — 6.1 mi — Golfside Dr → downtown Ypsilanti — repositioning to zone, online

Date, miles, destination, purpose. Two lines, ten seconds. Do that for every trip and your deduction is a fact instead of an argument.

A few habits that make a log hold up:

• Don't round. 14.3 is believable. 15 every time is not.
• Note your odometer at the start and end of the year. It lets you show total miles alongside business miles, which is what makes a business-use percentage credible.
• Keep the log where the miles happen. A system you have to remember to open later isn't a system.
• Back it up. A notebook that lives in a van that gets broken into is a log you no longer have.

If you want to know what those logged miles are worth before you go to the trouble, the mileage deduction calculator will tell you in about fifteen seconds. It's usually a bigger number than drivers expect, which is the best argument for keeping the log I know of.

I drive for a living and I read the IRS pages carefully, but I'm not an accountant and this isn't tax advice. Your return is yours — check anything that matters with a preparer who knows your situation.