Most people don't know the listing photos already graded the car
A wholesale buyer on AutoScout24 has two hundred lots to triage before the auction. The titles look fine. Mileage is plausible. The only way to know which Golf is a rust project is to open the gallery.
So they open the gallery. Sill corrosion on both sides. No interior. A hero shot from twenty metres. They close the tab, open the next one, and do it again until the shortlist is a memory of the last twelve cars they can still picture.
You are re-doing a job the pixels already did
Auction desks and dealer sourcing teams already know the wrecks photograph like wrecks. They still click every listing because the portal gives them price, year, and kilometres — not a condition grade. The damage is in the photos the seller uploaded. It is not in a field you can sort.
The workaround is a junior buyer with thirty tabs, a Slack dump of "this one's rough," and a habit of sending someone to look at cars that were obviously rotten in image three. Stock photos slip through. Watermarked galleries slip through. A clean car shot from one distant angle gets the same attention as a twelve-photo walkaround.
Most people don't know the listing photos already graded the car. The dents, the rust, the cracked light, the missing cabin shots — that is the condition file. Opening two hundred galleries is transcription.
Grade the gallery, flag the gaps
The useful output is not another photo zip. It is a grade you can rank on, a score, a panel-by-panel list of what is actually visible, and an honest note when the evidence is thin: no interior, dirty, stock, wheels not in frame.
That is the Car Listing Condition Grader API. You paste AutoScout24 listing or search URLs. Search pages expand into cars. Each vehicle comes back with an A–E grade, a 0–100 score, visible damage by panel, interior wear and tyre condition when the photos allow it, and flags for the shots that limit how much you should trust the number.
Three beats:
- Point it at a search or a list of listings you already intended to click.
- It reads the live gallery and inspects the photos.
- You sort the wrecks from the rest, then inspect the shortlist like a person.
I built it so the first pass happens on the photos the seller already posted, not in a buyer's browser history.
Who this is for, and who it will mislead
If you triage auction or wholesale lots before bidding, this is for you. If you rank trade-in and remarketing stock before you send a buyer, this is for you. If you want to join visible condition to asking price and mileage, this is for you.
If you need a mechanical inspection, this will lie to you with a straight face. It cannot see the engine, the gearbox, the underbody, accident history, or anything the seller did not photograph. Sellers shoot the good side. Missing angles show up as flags, not as invented rust. The same listing can score a few points differently on a second run — rank on the grade, do not treat the score as a lab measurement.
Listings with no usable photos still come back, ungraded, so you know the gap. Input, pricing, and the rest of the caveats live on the Apify Store README for Car Listing Condition Grader.
UK property listings hide a similar record in a different drawing. If your queue is floorplans rather than body panels, I wrote separately about reading room sizes from listing floorplans.
A tab for every lot is not a condition process
The old cost is the hour spent confirming what image four already showed, and the bid you still placed because the wreck was listing eighty-three and nobody had opened it yet.
If the listing is public and the damage is in the gallery, read the gallery.
The other listing-vision posts on rainminer start from the same habit: stop clicking through evidence the page already hosted as images.