General tax education, not tax advice. Publication 463 and Treas. Reg. 1.274-5T want a mileage log made at or near the time of each trip. Passenger automobiles are listed property; a Cohan-style estimate does not replace those records. A reconstruction is weaker evidence than a contemporaneous log. Nothing on this page promises a deduction, a refund, or any audit outcome. Confirm the current IRS forms for the year you file, and have a tax professional review a thin or mixed-use file.
You drove all summer. The apps paid. The notebook stayed in the glove box. Now someone — a CPA, a September 15 estimate, or a Form 1040 draft — asks for business miles by date, and you have a 1099 and a feeling.
That feeling is not a log. Reconstructing months later is painful for a boring reason: the details that make a mile useful — purpose, route, which hops were work — leave your head long before tax season. 2026 still has two business rates — 72.5¢ for miles driven January 1 through June 30 and 76¢ for miles driven July 1 through December 31 (Notice 2026-10 / Announcement 2026-11). Those cents attach to the date of the trip. A reconstructed year total times one rate is the same error as never logging at all.
The short version: contemporaneous notes (at or near the drive) are what the IRS asks for. Reconstructing months later fails because memory fades, platform histories leave empty miles, and purpose is missing. If a gap already exists, rebuild only from independent proof and write dated, conservative rows. Then capture close to each trip and run a short Friday review so uncertain drives never pile up again. A rebuild is possible. It is error-prone. It is not a promise of any tax result.
Why reconstructing months later fails
People treat a lapsed log like a puzzle they can finish in a weekend. The pieces are not sitting in a box. Three things go missing, and they go missing fast.
Memory fades
A Tuesday in June felt obvious while you were in it: grocery batch, then a personal pharmacy stop, then the app again. In September those three legs collapse into “I worked that day.” You cannot reliably split business, commuting, and personal from a calendar color two months later. The more mixed the day, the worse the reconstruction. Dual-purpose drives — work plus an errand on the same road — are where memory is least trustworthy. How to classify those while they are still fresh is in what counts as a business mile.
Empty miles
Uber, Lyft, DoorDash, and similar histories are real evidence. They are also incomplete. On-trip miles usually start at pickup and end at drop-off. The unpaid hop to the next offer, the staging loop, and the last drive that was still work often never appear. A reconstruction that copies the platform PDF and stops there understates the year. A reconstruction that then “adds 30 percent for deadhead” with no days attached overstates it. Both are guesses dressed as a log. Empty miles are why you cannot rebuild a summer from a year-in-review screenshot.
Missing context
A useful row is not only a number. It needs a date, a distance, and why that drive was work. Platform exports rarely store purpose in a form you can hand a preparer. Maps Timeline shows where the phone went, not whether the stop was a customer or a school pickup. Toll statements prove a crossing, not a business reason. When context is gone, you can sometimes prove you moved the car. You cannot honestly label every mile as business. That is the gap reconstruction cannot close with confidence.
What a useful mileage record typically needs
This is a practical recordkeeping list, not legal advice and not a complete IRS checklist for every situation. A row that is useful months later usually has:
- Date of the drive. In 2026 the date also picks the rate: 72.5¢ through June 30, 76¢ from July 1.
- Miles or distance for that trip or that continuous work block — a number you can still explain, not a rounded month total.
- Purpose / business context in a short phrase (Uber dinner shift, DoorDash batch, parts for the work car). “Work” by itself is thin.
Destination or route (first pickup area, last drop, supply run) is the field that makes the purpose believable. Year-start and year-end odometer readings help show that business miles fit inside total miles on the car. The weekly habit that keeps those fields current is how to keep an IRS-ready mileage log. The formal substantiation list is in IRS mileage log requirements.
A weekly lump — “about 3,000 miles this summer” — is a total, not a record. If a day is messy, one row per offer is cleaner; one row per continuous block with a start, end, purpose, and miles is a fallback only if you still have the platform history underneath.
The habit that replaces reconstruction: capture close, review on Friday
Contemporaneous means written at or near the time of the trip — same day, next morning, not next April. You do not need a perfect system. You need two small loops that stay small.
Capture close to the trip. When the drive ends, or when you park after the last drop, mark the trip while the stop is still in your head: date (already there if the phone stamped it), miles, and a purpose you would still recognize in January. Automatic GPS capture helps with the number. It does not invent the purpose. Classification is the part reconstruction cannot fake later. What to write in that purpose line — and when a next-morning note still counts as near — is same-day mileage notes vs April reconstruction.
Run a short recurring review. Pick a standing slot — Friday after the last shift is the one that survives a messy week; Sunday night works if Fridays are still on the road. Ten minutes:
- Open the log and the platform histories for the week.
- Fill any trip that has miles but no purpose, or a purpose but no date.
- Mark personal and commuting trips as such. Do not upgrade a grocery stop because the app was still open.
- If you multi-apped, confirm you did not count the same stretch of road twice.
- If you cannot place a hop, leave it out. Uncertain miles that sit until October become another reconstruction.
The Friday pass is not a second reconstruction. It is how you keep uncertain trips from stacking. A weekly review still sits “at or near” the drives. A spreadsheet filled the night before you file does not. Digital timestamps help show the habit; a notebook dated as you go does too. What neither replaces is inventing a summer in September.
Honest limits of after-the-fact reconstruction
If tracking already lapsed, an honest rebuild can still be better than a blank year. It is not the same strength as a live log. It does not become contemporaneous because you labeled the spreadsheet carefully. It does not create records that never existed.
Passenger automobiles are listed property. For listed property, the Cohan rule (the old court idea that some expenses can be estimated when you clearly spent something) does not rescue a missing vehicle log. Examiners and courts treat a year-end guess — “I drove about 18,000 business miles” — as inadequate substantiation, not a close-enough number.
What reconstruction can do: assemble the best remaining evidence of trips that already happened, write dated rows, stay conservative, and label rebuilt months as rebuilt. What it cannot do: promise that those miles will be allowed, produce a refund, or predict an audit. If a fire or flood ate a log you actually kept, say so and rebuild from the same third-party sources. If you never kept one, do not pretend you did.
If the gap is a prior year and you already filed without the miles, reconstruction is only the record you would attach to an amendment — if the window is still open and the proof is thick enough to bother. That decision is how to amend a return for missed mileage. Form 1040-X does not upgrade a guess into substantiation, and this page does not tell you to amend.
If you already have a gap: rebuild from proof, not memory
A reconstruction lives or dies on sources that existed before you sat down to invent mileage. Stack more than one. Memory is the weakest layer.
| Source | What it usually proves | What it usually misses |
|---|---|---|
| Uber / Lyft / DoorDash / Instacart trip history | Date, time, on-trip distance, paid work | Deadhead, hotspot staging, home bookends, multi-app overlap |
| Calendar, dispatch texts, blocked shifts | Which days you intended to work and where you aimed | Actual miles; personal stops on the same day |
| Google Maps Timeline or similar location history | Routes and stops the phone already recorded | Business purpose; you still have to label each drive |
| Toll / EZ-Pass / SunPass / FasTrak statements | Dated crossings that match work routes | Miles that never hit a gantry; personal crossings mixed in |
| Dated odometer photos | A reading that existed on a real day | Which of those miles were business |
| Oil-change, inspection, or repair invoices | A shop-recorded odometer and a date | The split between work and personal since the last visit |
Export or screenshot each platform’s history for the gap. Use on-trip miles as a floor, then add only the between-trip miles you can support from Timeline, tolls, or a mapped route you can still describe. Do not add a round “deadhead factor.” Where a day is thin, keep the platform figure and skip the unpaid miles you cannot place. Rounding down is the point. The shorter summer version of this pull is in the Q3 estimated-tax prep guide.
Write a log, not a paragraph. Each reconstructed row needs the same fields a same-day log would have had: date, destination or route, business purpose, miles, and a source note (which export, which calendar, which toll line). Do not classify first-and-last home drives as business unless the commuting rules actually allow it. If two apps show the same physical mile, count it once.
Then fit the total. Schedule C Part IV (or Form 4562 Part V) still wants total miles on the vehicle and the business / commuting / other split. Any two dated readings — January inspection at 41,200, August oil change at 52,400 — give you a ceiling for all driving in between. If reconstructed work miles overflow that ceiling, the reconstruction is wrong. Cut it. How to keep those bookends on purpose is the odometer photo log. Do not photograph today’s odometer and back-date the file.
If you use the 2026 standard mileage rates, split reconstructed rows at June 30. Multiply first-half business miles by 72.5¢ and second-half business miles by 76¢. Do not average the two rates. Do not apply 76¢ to a January trip because you rebuilt the log in September. The rate follows the road date, not the spreadsheet date. Illustration only, not a promise: 1,200 reconstructed business miles through June 30 and 2,400 from July 1 is $870 + $1,824 = $2,694 of standard mileage if those 3,600 miles are dated, sit inside the odometer delta, and are not a second copy of the same road. That arithmetic is not a deduction you are owed.
What to do next
If you still have a gap, finish the conservative rebuild above and write on a cover sheet which months are reconstructed. Then stop reconstructing.
- Start the live habit tonight: capture close to the trip, Friday review so blanks do not stack. The full weekly loop is the IRS-ready mileage log pillar.
- When a preparer asks for the year, export dated rows — not a year total. That packet is how to export a mileage log for your CPA. Do not dress a reconstruction up as a contemporaneous export.
- If classification is the part that keeps slipping, keep what counts as a business mile next to the Friday review.
TaxMiles: Mileage Tracker is a recordkeeping tool. It auto-detects trips and lets you classify them the same day so date, miles, and purpose sit on the actual drive. It does not rewind a summer you already drove, prepare a return, or promise any tax result.
Rebuild once. Log the rest in real time.
TaxMiles: Mileage Tracker auto-detects trips and lets you classify them the same day so 72.5¢ and 76¢ attach to the road date, not a September spreadsheet. Recordkeeping tool, not tax advice. Not MileIQ. Not Everlance. Not Mileage Tracker for Taxes (id 6758426140). Free plan is 40 trips a month; Pro is $5.99/month, $39.99/year, or $79.99 lifetime.
Download TaxMiles FreeFrequently Asked Questions
Can I reconstruct a mileage log after I stopped tracking?
Yes, as a weaker fallback. Stack independent proof and write dated rows. A round estimate is not a log. Reconstruction does not guarantee a deduction, a refund, or any audit outcome. See what a contemporaneous log is supposed to contain.
Why does reconstructing a mileage log months later fail?
Memory fades, platform histories leave empty miles, and purpose is missing. You can sometimes prove the car moved. You cannot honestly relabel every hop as business from a September feeling.
What does a useful mileage record typically need?
Date, miles or distance, and a short business purpose, plus destination or route when you have it. The weekly habit is the IRS-ready mileage log.
Does the Cohan rule let me estimate vehicle miles?
No. Cars are listed property. Cohan-style estimates do not replace Publication 463 / Treas. Reg. 1.274-5T substantiation for vehicle use.
What weekly habit keeps me from reconstructing again?
Capture close to the trip, then a ten-minute Friday (or Sunday) review so uncertain trips do not pile up. When it is time to hand the year to a preparer, use the CPA export checklist.
Related reading: IRS-ready mileage log, export for your CPA, what counts as a business mile, odometer photo log, mileage log requirements, Q3 estimated-tax prep, amend a missed mileage deduction, switch mileage apps without losing the log, and TaxMiles on the web.
This article is general tax education for U.S. federal Schedule C filers, not legal, tax, or accounting advice and not a guarantee of any deduction, refund, or audit outcome. Publication 463, Treas. Reg. 1.274-5T, listed-property rules, Schedule C instructions, and IRS mileage rates can change; read the current IRS forms (Publication 463, Notice 2026-10, and Announcement 2026-11) or work with a licensed professional. TaxMiles: Mileage Tracker (App Store id 6758579463) is a mileage recordkeeping tool, not a tax preparer. It is not affiliated with the IRS, and is not MileIQ, Everlance, or Mileage Tracker for Taxes (id 6758426140).