← All articles

When a Vinted search finds nothing, it quietly shows you popular junk - here is how to detect it

While building a Vinted scraper, my cloud test suite flagged something odd: a search for zqxwcevrbtynumz introuvable - deliberate nonsense - returned 50 items. Dresses, sneakers, a phone case. All real listings, none related to the query.

Vinted does this on purpose: when a search has no real matches, the page is filled with popular listings so the user never sees an empty screen. Fine for shoppers. Terrible for a scraper that charges per item: users would pay for junk.

The obvious signals don’t work

Each listing in the server-rendered payload carries tracking data, so my first idea was to read it:

{ "contentSource": "search", "searchScore": 1 }

contentSource stays "search" even for the fallback junk, and searchScore has the same decreasing values (1, 0.9999989…, 0.9999979…) on every page I compared - nonsense query or real one. No isFallback flag, no emptyState marker anywhere in the payload. I diffed the full set of JSON keys between a real search page and a nonsense one: identical.

Title matching backfires

Second idea: keep only items whose title contains the searched words. It kills the junk, but it also killed legitimate results. Searching gameboy on vinted.de returns actual Game Boy games - titled “Super Mario Advance 4”, “Tetris DX”, “L’âge de glace”… Sellers title listings by the game’s name, not the console’s. My filter dropped all of them. Title relevance is a trap for marketplace search.

The signature was in the pagination

The real tell is one number. Every search result page carries:

"pagination": { "current_page": 1, "per_page": 96, "total_entries": 960 }

For real searches, total_entries is the (capped) result count: 960, 234, whatever. For every no-result search I could produce - on vinted.fr, vinted.de and vinted.co.uk, with nonsense text or with a real but impossible query like gameboy advance sp ags 101 rouge - it is exactly 150. The fallback feed is a fixed-size block of popular items, and its size is the signature.

One subtlety: the flight data contains several total_entries keys (favourites lists, other blocks), so you must read the one immediately after the items array, not the first match in the page.

What the scraper does with it

total_entries == 150  → the market found nothing real:
                        keep only items that actually match the text (usually zero), bill nothing for junk
anything else         → trust Vinted's own ranking, deliver everything

The check runs per country, because a search can be empty in Germany and full in France. And a strictMatching option lets price researchers force title matching everywhere.

After the fix, the nonsense search returns 0 items and the gameboy search returns its 40 real games. The test suite is green, and nobody pays for a fallback feed.

The result, if you want to see it live: Vinted Scraper on Apify - 26 countries, prices converted with ECB rates, alerts on new listings. Input templates: github.com/EuroScrape/apify-actors.

Part of the EuroScrape actor collection — all articles.