Home › Blog › Research › SURBL Study

Original research11 min read

Does a SURBL listing actually hurt cold email? We tested 2,817 sending domains

43.3% of our live sending domains are listed on SURBL. We measured what that costs, two independent ways, and the answer is not what the listing rate suggests.

WarmInboxes original research·Fieldwork: 30 days to 29 July 2026·Published 7 October 2026
Share this

The data behind this study

Every figure on this page is computed from these files. Fieldwork 30 days to 29 July 2026. Licensed CC BY 4.0: reuse with attribution to WarmInboxes.com.

The short version

2,817

Sending domains tested, every one that sent in the window

43.3%

Listed on SURBL (1,220 domains)

1.14% → 1.16%

Reply rate before and after listing. Flat

0.025%

Of all sends lost to blocklist bounces

A SURBL listing did not change reply rate or booked conversations. It did raise bounce rate, by about half. Those two facts sit together more comfortably than they look: a listing costs you a slice of delivery volume, and the slice is small enough that it never reaches the outcome anybody is paid on.

Where the domains came from mattered far more than any of it. Infrastructure we build and run ourselves came back 26.9% listed. Bought-in stock came back 63.0% listed — 2.3× worse for the same job.

Why we ran it

Clients running campaign reviews kept asking the same question: bounces had crept up on particular domains, and was a blocklist the reason? It is a fair question and we could not answer it, because we were reading blocklist status out of bounce messages.

That turns out to be a bad instrument. Only about one listed domain in five ever produces a bounce that names the listing — you can see what a bounce does and does not tell you by pasting one into our bounce analyzer. Four in five are invisible to anyone watching bounces, which means every previous answer we had given on this subject was built on a fifth of the picture.

The question got louder in June, when SURBL started mass-listing .info domains on Microsoft and Azure and operators across the industry watched whole estates go red at once.

There is a second reason the question is awkward to answer from the outside. SURBL's listing criteria reach patterns of domain registration and usage, not only message content, so infrastructure built for cold email attracts listings at rates that would look alarming on a corporate mail server. A high listing rate is therefore expected. Whether it matters is a different question, and one nobody in this industry had published a number for.

So we stopped inferring and tested every domain directly.

Method

PopulationAll 2,817 domains that sent during the window. Not a sample.
TestDirect DNS query per domain against multi.surbl.org — the same query our free bulk SURBL checker runs, so you can reproduce this on your own domains in a browser tab.
Why not bounce dataOnly ~1 listed domain in 5 ever produces a bounce that reveals the listing. Our bounce analyzer classifies the ones that do.
Validation5 domains known to be listed all returned positive; 3 known-clean controls all returned clean
Query failuresNone. Every domain returned a definite answer
PlatformsTwo sending platforms, pooled — the question is about domains, not tooling

What the sweep found

Estates are lettered rather than named. A and B within a provider are separate pools of infrastructure, not a split by any property we tested.

EstateDomainsListedSendsBounce share
Azure, estate AOurs5249.5%(50)1,118,1820.0145%
Azure, estate BOurs49916.4%(82)912,5400.0187%
Google, estate AOurs41554.5%(226)53,8290.0093%
Google, estate BOurs9955.6%(55)29,8620.0536%
Google, bought-in reseller stockBought in1,08256.6%(612)493,4280.0764%
Third-party SMTPBought in19898.5%(195)268,2180.0000%
All domains2,81743.3%2,876,0590.0254%

Listing rate on its own is a misleading number. Our Google estate A is 54.5% listed and has the lowest real impact of anything in the table at 0.0093% of sends, because each of those domains carries only about 130 sends a month. A listed domain that barely sends costs almost nothing.

Bought-in reseller stock is the worst performer on impact, not just on listings. It accounts for 0.0764% of its own sends in blocklist bounces, about 4.7× the rate of our own Azure estates, which carry 2,030,722 of the 2,876,059 total sends at 0.0164%.

One row is a measurement problem rather than a result. Third-party SMTP came back 98.5% listed — 195 of 198 domains — on roughly 1,355 sends per domain per month, and reported exactly zero blocklist bounces. Zero is not a good score here. It means that channel does not surface blocklist rejections in a form anyone downstream can read, so its true cost is unmeasured rather than low. That gap is the reason we no longer rely on bounce messages anywhere.

Who built the domain matters most

SourceDomainsListedBounce share
Built and run by us1,53726.9%0.0167%
Bought-in stock1,28063.0%0.0495%

This is the finding with the clearest practical consequence, and it is the one we did not set out to test. Domains we register, configure and warm ourselves came back 2.3× cleaner than domains bought ready-made. Same blocklist, same window, same sweep. If you are buying inbox capacity from a reseller, this is the number worth asking them for.

It is the same argument our monthly inbox placement log makes from the other direction: every batch of prewarmed domains and inboxes we build is seed-tested before it ships, and only what clears 95% is deployed. Nobody runs that test on stock you buy ready-made.

Does being listed hurt? Two tests, same answer

Run two ways on purpose. Each has a weakness the other does not share, so agreement between them is worth more than either alone.

Test 1 · Same domains, before vs after their first listing

106 domains with at least 200 contacts on each side of the listing date. Each domain is its own control.

MetricBeforeAfterChange
Reply rate1.14%1.16%flat
Bounce rate1.41%2.16%+53%

Test 2 · Listed vs clean domains over the same period

Measured on sends needed per interested reply, which is the number a campaign is actually judged on.

2,733

sends per interested reply, listed domains

2,687

sends per interested reply, clean domains

1.7%

difference — inside normal campaign variance

Both tests say the same thing: a listing does not move replies or booked conversations. Bounce rate does rise, so deliverability is genuinely affected — this is not a claim that listings are harmless. It is a claim about where the harm stops.

A SURBL listing costs you send volume. It does not appear to cost you results.

How big is the problem, really

Every cause below is measured the same way over the same window: delivery lost as a share of all sends.

Invalid addresses in prospect lists0.420%
Recipient-side policy blocks0.260%
All SURBL listings combined0.025%

Invalid addresses in prospect lists cost roughly 16× what every SURBL listing in the fleet costs put together. Recipient-side policy blocks cost about 10×.

Over the window, blocklist bounces were 731 out of 2,876,059 sends, and 1.79% of all bounces from any cause. If you are spending time on SURBL before you have spent it on list verification, you are optimising the third-largest problem first. Start with the free email verifier and the CSV list cleaner, then read the segmentation playbook for what to do about recipient-side policy blocks.

How the listings appear to be decided

SURBL does not publish its criteria. What follows is diagnosed from the outside — from which of our domains were listed and which were not, across this sweep and the estates we have watched since. Three patterns account for nearly everything we have seen, and each has a fix.

1Bulk registration, grouped on a shared keyword

Fresh registrations appear to be scraped from public WHOIS and normalised, with domains registered in a batch grouped by the word they have in common. Ours made the pattern obvious: getupscale.com, proupscale.com and upscalego.com share the token upscale. All three were listed. The registration batch is visible, the shared token is the key, and the whole group goes together.

The fix

Register generic names with no token in common. If nothing in one domain’s name ties it to the next, there is no key to group them on. This is our naming convention for the free prewarmed domains that come with every prewarmed package, and it is the main reason those pools have stayed clean. The inboxing playbook has the full naming and sourcing rules.

2Redirects pointing at one master domain

Older domains appear to be fingerprinted off redirection. We had 200 domains redirecting to warminboxes.com. warminboxes.com and all 200 of them are now listed. That is the part worth sitting with: the master domain gets listed too. A redirect is a published, crawlable link from every burner back to the brand, and it is trivially walkable in both directions.

The fix

Do not point more than about a hundred domains at one master — honestly, keep it under ten. Beyond that, stop redirecting and serve something instead:a) A masking proxy. A small Cloudflare edge worker that serves crawlers a different page from the one a person redirects to.b) A real page on every domain, which is the safer of the two. Cloudflare Pages can programmatically publish a copy of the main site on each burner domain, booking embed and all, so there is no redirect to follow in the first place.This is simple to build. Do not pay anyone $3 to $10 per site for it — we are launching it free.

3Shared nameserver zones

Domains also appear to be grouped by the nameserver zone they sit in. Most infrastructure providers were running thousands of unrelated domains through the same Cloudflare zones, which makes every domain in the zone a neighbour of every other one. Co-tenancy at that density was always going to carry contamination. Our bulk nameserver checker shows you how your own estate is spread: paste the list and it groups your domains by the zone they resolve through.

The fix

Spread the estate across many accounts and zones so no single zone holds a meaningful share of it — which is what we do for you. With budget, dedicated DNS is better still: Route 53 or a paid DNS host costs more and is worth it at scale.

What we changed afterwards

Detection moved off bounce messages

Bounce-based detection saw about one listing in five. The DNS sweep is now scheduled across the full estate, so listing status is known rather than inferred.

Nothing was retired wholesale

233 listed domains carrying 3,173 mailboxes were still in production and producing around 155 interested replies a month. Retiring working infrastructure over a listing that does not affect replies would have cost clients real conversations to fix a number on a dashboard.

Sourcing weighting changed

Bought-in stock listing 2.3× higher than our own is the finding with the longest tail. It is a reason to build rather than buy, and we have weighted new capacity accordingly.

What this study does not show

  • One blocklist, not all of them. This tests SURBL. Spamhaus, UCEPROTECT and the rest behave differently and carry different weight with receivers. Check those separately with the Spamhaus DBL checker or run a domain against all twenty at once in the blacklist checker.
  • One fleet, one window. 2,876,059 sends over 30 days to 29 July 2026, on cold-email infrastructure. A transactional or newsletter sender has a different risk profile and should not read these numbers as theirs.
  • Observational, not randomised. Domains were not assigned to be listed. Test 1 reduces this by using each domain as its own control, but a listing is still something that happened to a domain rather than something we did to it.
  • The zero is unmeasured, not clean. The third-party SMTP row reports no blocklist bounces because that channel does not surface them, so its real cost is unknown and is excluded from any conclusion about impact.

Check your own domains

The sweep behind this study is the same query our free SURBL checker runs. You can point it at one domain or paste in a whole CSV, and it costs nothing and needs no account.

If your listing rate looks like our bought-in column rather than our own, the problem is where your domains came from, not what you are sending.

Read next

Prewarmed & Fresh Inboxes

Google, Microsoft 365 and Azure inboxes with free prewarmed domains and DNS configured before you log in. Prewarmed inboxes send the day they land; every order is delivered within 48 to 72 hours.

Get Prewarmed & Fresh Inboxes