24,835 Brazil Rentals: One Apartment, Three Portals, Three Prices

In four Brazilian municipalities, on 27 July 2026, 24,835 active rental ads across ZAP Imóveis, OLX, ImovelWeb, Chaves na Mão and QuintoAndar described 13,806 distinct homes. 53.8% of those homes were advertised on two or more of the five portals, and no single portal showed even half of them. The duplication in this article's title is real. The three prices mostly are not: among the homes advertised on more than one portal, 93.8% carried exactly the same rent on every portal that listed them. What the data shows is not five marketplaces pricing the same apartment differently. It's one advertiser feed republished under several brands, plus a small minority of cases where the gap between two ads for one home is large enough to change what you can afford.

Every guide to renting in Brazil, ours included, gives the same advice: before you trust a listing, look for the same home on another portal. Our own fake-listing guide states it as a fact: a real apartment for rent in Brazil is almost never listed in only one place. It's sound advice. It was also, until now, an assertion — nobody had published how often the same home actually appears twice, which pairs of sites it appears on, or how far the prices drift apart when it does. This page is that measurement.

A word on who is talking. RemRent is a rental-search aggregator being built for Brazil: one search across six portals — the five measured here plus Viva Real, from the same Grupo OLX family — with the duplicates collapsed. It isn't open, nothing is on sale here, and this measurement was only possible because the base of listings we have already collected for that product exists. We have an obvious interest in the answer being large, so the method, the sample, the date and every known bias are set out in full below, and every share we publish is a floor — a minimum. Every error our method can make pushes the number down, never up.

What we measured, and how

The method, in one paragraph. Measured on RemRent's own aggregation corpus, snapshot 27 July 2026, 24,835 active rental listings across five Brazilian portals (ZAP Imóveis, OLX, ImovelWeb, Chaves na Mão and QuintoAndar) in four municipalities: Osasco in São Paulo state, Niterói in Rio de Janeiro, and Canoas and São Leopoldo in Rio Grande do Sul. Duplicate identification is photo-based. Every listing photo is embedded with an SSCD ResNet-50 model into a 512-dimension vector. Two listings in the same municipality are treated as the same home when at least three distinct photos of one match photos of the other at a cosine similarity of 0.92 or higher (or are byte-identical files), when those matches cover at least 80% of the smaller listing's photo set and at least 25% of the larger one's, and when the advertised bedroom counts do not contradict each other. Price, address and description are never consulted in the matching decision, which is what keeps the price findings independent of the matching rather than circular.

How often the same home appears on more than one portal

The 24,835 ads collapsed into 13,806 distinct homes, roughly 1.8 ads per home. 7,433 of those homes, 53.8%, were advertised on two or more of the five portals. 1,864 homes, 13.5%, were on three or more. 179 homes, 1.3%, were on four or more. Read the other way: a renter working through one portal's results in these four municipalities is looking at a list in which more than half the homes are also sitting on at least one other site.

The share is not uniform. In Osasco it was 46.84% of 4,859 homes, in Niterói 46.71% of 2,507, in Canoas 61.66% of 3,685 and in São Leopoldo 62.21% of 2,755. That is a wide band across four ordinary municipalities, and it is the main reason we treat the combined 53.8% as a finding about this sample, not about Brazil.

The combined figure also hides the more useful result. Duplication is not spread evenly across the ten possible pairings of five portals. It is concentrated in two of them.

Each of the last two rows covers several pairings at once, so there is no single shared-homes count to print. We show the absolute number only where one pairing is large enough for it to mean something, and QuintoAndar's remaining pairings run between 0.03% and 0.10%.

ZAP and OLX are one inventory pool. 6,131 homes were on both. Of everything ZAP listed in the sample, 96.23% was also on OLX; of everything OLX listed, 92.18% was also on ZAP. For a renter, opening those two sites is close to opening one site twice. That's not a surprise once you know that the five brands are two corporate groups and one independent, and that Grupo OLX's own terms of use let an ad be republished across its sites. What the measurement adds is the size of that effect.

A third of QuintoAndar's homes are also on ImovelWeb. 1,009 homes appeared on both: 34.15% of everything QuintoAndar listed and 24.23% of everything ImovelWeb did. The two belong to the same company — Grupo QuintoAndar — so the overlap is not a surprise. What the measurement adds is how big it is. In every other direction QuintoAndar's inventory is sealed: between 0.03% and 0.10% of it reached ZAP, OLX or Chaves na Mão.

No portal shows even half the market. Measuring each portal against all 13,806 homes in the sample gives the coverage below, together with the share of homes that appeared on that portal and nowhere else.

Data as of 27 July 2026, four municipalities, five portals, 13,806 distinct homes identified by photo matching. Shares are of the sample, not of national inventory, and each is a floor for the reasons set out further down.

The exclusive column is where the practical damage sits. The two most-visited brands add almost nothing to each other, while the two portals with the smallest share of the sample hold the largest blocks of homes that exist nowhere else: 17.41% of all homes were on ImovelWeb only, and 14.09% on QuintoAndar only, close to two thirds of everything QuintoAndar published. QuintoAndar also screens applicants the hardest, so the inventory a newcomer is least likely to be approved for is also the inventory least likely to be visible anywhere else. That is worth knowing before you plan a rental search from outside the country.

The price spread on identical homes

This is where the premise of the headline breaks. Of the 7,433 homes advertised on more than one portal, only 462, that is 6.2%, carried more than one price. The remaining 93.8% were advertised at exactly the same rent everywhere they appeared. Cross-portal duplication in Brazil is overwhelmingly republication of one advertiser's ad, not five markets competing on price for the same set of keys.

The ZAP and OLX pair makes the mechanism visible: of the 6,131 homes on both, 6,027 carried an identical rent. The same holds on the full monthly cost, not just the headline rent. Portals differ in whether they lead with the rent alone or with rent plus condominium fee plus property tax, so we ran the comparison on that combined figure as well, and the middle case was a gap of exactly zero. The median is the middle case once every gap is lined up from smallest to largest, which is not the same as the average. The headline rent behaves the same way.

Where prices do differ, though, they differ hard. Among those 462 homes the median gap between the cheapest and the most expensive ad was R$140, or 6.1% of the lower rent. A quarter of those gaps were 14.29% or wider; the widest 5% of gaps all ran above 71.65%, and the single widest hit 228%. A house in Niterói was advertised at R$2,600 on two portals and R$4,500 on a third, a difference of 73.1% for the same home. A one-bedroom apartment in Osasco appeared on all five portals at once, at R$1,900 on three of them, R$2,125 on the fourth and R$3,110 on the fifth, a spread of 63.7% between the cheapest and the most expensive ad for one home.

The direction of those gaps is not random either. Where ImovelWeb and QuintoAndar disagreed on a shared home, ImovelWeb was the more expensive one in 134 cases against 34 the other way. Chaves na Mão came in below OLX and ZAP between three and four times as often as it came in above them. So the disagreements, rare as they are, lean the same way each time — what you would expect from stale copies and differing fee conventions, not from genuine competition.

For a renter the useful read is this. Checking a second portal will confirm the price you already saw about fifteen times out of sixteen. The sixteenth is worth the two minutes on its own, because that's where the four-figure differences live.

How to check a listing yourself in two minutes

Everything above was done by machine, across our whole collected set of ads. The same check works by hand on a single listing, and the measurement suggests how to spend the two minutes.

• Choose the second portal deliberately. Opening ZAP and OLX for the same home is close to opening one site twice, because 96.23% of ZAP's homes were also on OLX. Pair one of those two with ImovelWeb, QuintoAndar or Chaves na Mão instead, and the check actually reaches different inventory.
• Search on what travels between portals. Ad text is rewritten from site to site; the photographs and the neighborhood stay constant, which is why the matching here is photo-based. Search the neighborhood plus the size and the approximate price, then compare the images.
• Compare totals, not headlines. Portals differ in whether the number in large type is the rent alone or the rent with the condominium fee and the property tax. On the combined figure the duplicate ads in our sample agreed almost perfectly, so if two copies of one home disagree on the total, that is a real difference, not a formatting one.
• Do not treat a single copy as a warning by itself. 46.2% of the homes in this sample were on exactly one portal. Appearing in only one place is the normal condition of nearly half of genuine listings, so absence elsewhere is a reason to look harder, not a verdict.
• When you find copies, expect them to agree. Roughly fifteen times out of sixteen they'll show the same rent. A copy that disagrees is worth a direct question to both advertisers, and the answer is usually a stale ad or a different definition of what the price includes.

What this means for the fake-listing checklist

The cross-portal check is the strongest single step in our guide to spotting a fake listing, and this measurement calibrates that step; it does not replace it. Two corrections come out of the data.

First, the step works, but not universally. Slightly more than half of genuine homes here could be found on a second portal, so a match is meaningful confirmation. The reverse is not: with 46.2% of real homes on one portal only, failing to find a second copy does not mark a listing as fake. Treat a no-match as a prompt to verify ownership another way, not as proof.

Second, the checklist leans harder on price differences than the data supports. Two honest ads carry the same rent 93.8% of the time, so a mismatch really is unusual — but unusual is not the same as fraudulent. The largest gaps we measured, including the Niterói house at R$2,600 on two sites and R$4,500 on a third, were between ordinary ads on established portals with no scam involved. A lower price on another site does not identify the fake. What identifies the fake is the demand for money before a viewing, and no measurement changes that.

What this data cannot tell you

Why these four municipalities. They were chosen for balance, not size: each of the five portals carried at least 295 rental listings in each of the four, which makes this the first sample where QuintoAndar, the smallest and most closed of the five, is present in enough volume to compare against the rest. Three metropolitan areas and three states are represented. São Paulo city, with more than half a million listings of its own, was deliberately left out of this round because it cannot be processed end to end on the hardware this ran on. Four municipalities are not a national sample, so the per-city figures are published alongside the combined figure instead of behind it.

How we checked the matching. Merge quality was audited three ways. Fifty groups of ads for one home — forty drawn at random and the ten largest — were checked photo by photo in ten independent review passes: 49 came back as one home, one as a mixed group, none unresolved. The mixed one turned out to be two different apartments in the same condominium, the known hard case for any photo matcher. Group sizes showed no sign of over-grouping either: 5,393 homes with two ads, 1,568 with three, then a short tail down to a single largest group of 39 ads for one home. And on 0.57% of the homes with more than one ad, the most expensive ad was more than 1.5 times the cheapest.

Every bias we know about pushes the measured duplication down, so the shares above are floors, not estimates.

• The matcher is deliberately conservative. An internal audit on 6,504 homes with multiple ads put cases where we wrongly treated two different homes as one at 0.11%, while genuine pairs we failed to match are roughly twenty times more common. A dedicated check after this run found 34 genuine pairs that met every criterion and were still never joined, about 0.3% of the joins we actually made, plus 36 pairs that share only generic condominium stock photos and were correctly left apart.
• Listings with fewer than three usable photos are invisible to it. 201 ads in the sample were excluded on that basis.
• Matching is partitioned by municipality. The same home advertised under two different city names cannot be detected at all, by construction.
• The crawls were not perfectly complete. About 98.7% of the scheduled collection runs finished cleanly; on OLX, completeness can only be verified at state level, never per municipality. A listing missing from one portal's crawl can only reduce measured overlap.
• Same-condominium twins remain the hard case. One of the fifty groups we audited by hand was two distinct apartments in one building whose photographs and bedroom counts agreed. That is the one error that would inflate duplication, and at one in fifty audited groups it is small next to the misses in the other direction.
• Four municipalities are not Brazil. The per-city band ran from 46.71% to 62.21%, and São Paulo city, the largest rental market in the country, is not in this sample.
• These are asking prices. Everything here is what advertisers published on 27 July 2026, not what tenants signed. A gap between two ads is a gap between two asks.

The thresholds are stated above so that a different choice of them can be argued with, and so the next round can be compared against this one.

Who we are, and why we have this data

RemRent is being built as a search layer over six portals — the five in this article plus Viva Real, from the same Grupo OLX family — not another portal of its own. One index of what they all publish, with duplicates collapsed, so that a home advertised on three sites appears once and its three prices sit beside each other instead of three browser tabs apart. We have gathered more than 1.2 million listings from those sources so far, and that is what made this measurement possible at all. It will carry no inventory of its own, run no credit check, and stand between no renter and no landlord.

It's not open yet. The waitlist is how you hear when it is: leave your email below and we will write to you on the day it opens in Brazil.