Skip to content
Email Deliverability

Blocklists explained: Spamhaus, SURBL, and how delisting really works

The bounce message is usually the first anyone hears about it:

550 5.7.1 Service unavailable; Client host [203.0.113.47] blocked
using zen.spamhaus.org; https://check.spamhaus.org/query/ip/203.0.113.47

At which point somebody pastes the IP into one of those free “check 87 blocklists” sites, gets back a page with three angry red rows, and asks what we are going to do about all three.

The honest answer is usually: one of those three matters, one is irrelevant, and one is a list that stopped existing in 2024. Getting that triage right is most of the work. Blocklists are not interchangeable — they index different things, list for different reasons, and have wildly different delisting processes ranging from a self-service button to a fortnight of correspondence to nothing you can do at all.

First, work out what has been listed

This is the step people skip, and skipping it wastes days.

Blocklists come in two broad families, and they answer different questions:

IP blocklists index sending IP addresses. They answer “should I accept a connection from this host?” and are consulted at SMTP connection time, before your message content is ever seen. Spamhaus ZEN, Barracuda’s list and the various policy lists live here. A listing means your server is refused.

Domain and URI blocklists index domain names found in message bodies — links, image hosts, tracking domains. They answer “does this message point at something bad?” and are consulted during content filtering, after the connection was accepted. SURBL and Spamhaus DBL live here. A listing means your message is refused, or scored heavily, regardless of how clean your sending IP is.

The practical consequence: if your bounces cite a URI list, changing IP will not help you. Not even slightly. You can migrate to brand new infrastructure, warm it beautifully, and get exactly the same rejections, because the thing being rejected is a link in your footer. We have watched a client spend real money on a mail platform migration to escape a listing that followed them across, because nobody read the reason code.

Read the bounce. It names the list. That name tells you which family you are in and therefore which set of remedies exists.

Checking properly, and the trap that gives you a wrong answer

The correct way to check a listing is a DNS query. For an IP, reverse the octets and append the zone:

# Is 203.0.113.47 in Spamhaus ZEN?
dig +short 47.113.0.203.zen.spamhaus.org

# Is example.com in the Spamhaus domain list?
dig +short example.com.dbl.spamhaus.org

# Is example.com in SURBL?
dig +short example.com.multi.surbl.org

No answer means not listed. An answer in 127.0.0.x means listed, and the last octet tells you which sub-list — this is why ZEN is useful, since it combines SBL, CSS, XBL and PBL into a single query and tells you in the return code which one caught you.

Now the trap, and it is a good one.

Spamhaus’s free public mirrors have a fair use policy that excludes queries arriving via large public resolvers. If your dig goes out through Google’s 8.8.8.8, or through the resolvers at AWS, DigitalOcean, Hetzner or Microsoft, you will not get a clean “not listed” answer. You will get:

127.255.255.254

That code does not mean “listed”. It means “your queries are being refused because of the resolver you used”. If you read it as a listing you will spend an afternoon delisting an IP that was never on the list. And there is a considerably worse version of the same mistake: if a mail server queries through a public resolver and its filtering software has not been written to handle the 127.255.255.25x range as an error rather than a hit, it can start rejecting every inbound message. That failure has taken down inbound mail for people who changed nothing except their upstream resolver.

Two ways to be correct here. Either query from a host with its own resolver and proper forward and reverse DNS, or sign up for Spamhaus’s free Data Query Service, which gives you a key to put in the zone name and works from anywhere. If you run a mail server that consults Spamhaus, check your return code handling today; this is not a hypothetical.

The web checkers at check.spamhaus.org and the lookup on SURBL’s own site sidestep the resolver problem entirely and are the right tool for a one-off check. They are also where the delisting forms live.

Spamhaus, list by list

Spamhaus is the one that matters most. It is consulted by an enormous share of receiving infrastructure, including plenty of systems whose operators could not name it, because it is baked into the default configuration of common filtering software. A Spamhaus listing is not a nuisance, it is an outage.

It is also several different lists with genuinely different rules, and treating them as one thing is why delisting attempts fail.

SBL — the manual one

The Spamhaus Blocklist proper lists IPs and netblocks associated with spam, phishing and malware operations. Listings are made by Spamhaus researchers on evidence and they are not self-service. There is no button.

Getting off the SBL means submitting a removal request that explains what happened and what has changed, and having a human agree. If your account was compromised and used to send phishing, the request needs to say so, say how it was compromised, and say what you did — closed the account, rotated credentials, patched the application, removed the mail from the queue. A request that says “please remove, we are a legitimate business” gets nowhere, because everyone says that.

Realistic timescale: hours to a few days if your explanation is complete and credible, considerably longer if it is not and you have to go back and forth. The single biggest determinant is whether you fixed the cause before asking.

CSS — the automatic one that expires on its own

Combined Spam Sources is an automatic list, generated from observed spam sources rather than researched. It behaves quite differently from the SBL: CSS listings normally expire on their own about three days after the last spam is seen from that IP.

That changes the calculus. If you have genuinely stopped the source, doing nothing is a legitimate strategy — it will clear itself over a long weekend. There is also a self-service removal form if you cannot wait, which is the right call if the IP is carrying business-critical mail.

What you must not do is request removal while the source is still emitting. It will simply be relisted, and repeated cycles of list, self-remove, relist are noticed.

XBL — compromised machines

The Exploits Blocklist covers hijacked machines: hosts running open proxies, worms, trojans and similar. Like CSS, XBL listings expire automatically once the attacks stop, and like CSS there is a self-service removal option if you need it sooner.

An XBL listing on your mail server is a security incident, not a deliverability incident. Treat it as one. Something on that host, or something with credentials to it, is being used to send. Delisting before you have found it is pointless — you will be back within the day.

PBL — not an accusation

The Policy Blocklist is the one people misread most often, and it is the one that is least about you being bad.

PBL lists IP ranges that should not be sending mail directly — residential and dynamic address space, IoT devices, ranges an ISP has declared as end-user. It is a policy statement, contributed largely by the network operators themselves. Being on it does not mean you have spammed. It means the address you are sending from is in a range where direct-to-MX mail is not expected.

Two consequences worth being clear about. If you are trying to send business mail directly from a home broadband connection or a cheap VPS in end-user space, PBL is doing exactly what it was designed to do and the fix is to send through a proper relay, not to fight the listing. But if you have a legitimate static IP that has been swept up in a range listing, there is a self-service exclusion process. End-user single-IP exclusions from PBL expire after one year, and are reversed immediately if spam is seen from that IP in the meantime. ISPs that manage their own PBL entries can set shorter periods and manage exclusions in bulk.

That one-year expiry catches people out. An exclusion you set up and forgot about in March will quietly lapse the following March, and mail that has worked for a year will start bouncing for no visible reason. Put it in a calendar.

DBL — domains, not IPs

The Domain Blocklist covers domains rather than addresses, and is consulted at the content-filtering stage. It works on a self-service basis through the checker at check.spamhaus.org, which is a genuine mercy given how often a DBL listing is a hacked website rather than a spam operation.

ZEN — the combined zone

ZEN is SBL, CSS, XBL and PBL in a single query. If you are configuring a mail server, query ZEN and nothing else from Spamhaus; there is no benefit to also querying the component lists separately, and doing so multiplies your query volume for no additional data. If you are diagnosing, use ZEN’s return code to work out which component list is responsible, because the delisting route depends entirely on which one it is.

One thing that is never in doubt: Spamhaus does not charge for delisting, ever. If you find yourself on a page asking for money to be removed from something calling itself Spamhaus, you are on a scam site. This happens often enough to be worth saying out loud.

SURBL and the problem of being listed by association

SURBL indexes domains that appear in message bodies — “domains of malicious or abused web sites”, used to filter or tag messages based on the links they contain, regardless of sender IP. Its public dataset is multi.surbl.org, a bitmasked zone combining several categories: phishing, malware, and abuse or spam sites.

The important property is in that phrase regardless of sender IP. SURBL does not care who sent the message. It cares what the message points at. Which means you can be perfectly clean and still be hit.

Here is the version we see most often. An agency client runs a newsletter. The footer contains a link to a case study on a microsite the agency built four years ago and has not updated. That microsite’s WordPress install gets compromised and quietly starts serving a phishing page from /wp-content/uploads/. The microsite domain gets listed. Every newsletter containing that footer link is now scored as if it were phishing, from an unlisted IP, with perfect SPF, DKIM and DMARC.

Nothing in Postmaster Tools will explain that. Nothing about your sending infrastructure is wrong. The only signal is a rise in content-based rejections and a fall in delivery, and the only fix is to find and clean the listed domain.

The same applies to shared infrastructure you may not think of as yours: link shorteners, shared click-tracking domains provided by an ESP, a shared image CDN host. If you use a generic shortener in marketing mail, you have outsourced part of your deliverability to whoever else is using it that week. That is a genuinely bad trade, and it is the main reason to insist on a branded tracking domain rather than accepting a platform’s shared default.

Delisting from SURBL runs through the removal form linked from its lookup result, and turnaround is typically 24 to 48 hours. The request needs to describe what caused the listing and what you did about it — cleaned the compromise, removed the injected files, patched the platform, rotated credentials. A request submitted while the phishing page is still being served will not be granted, and quite reasonably so.

Barracuda

Barracuda’s reputation list (b.barracudacentral.org) is the other IP list you will meet regularly, particularly with recipients running Barracuda appliances, which is common in UK SMEs and education.

Delisting is a self-service form at barracudacentral.org/rbl/removal-request that asks for the server IP, an email address, a phone number, and a reason. The phone number is not decorative — requests with obviously invalid details are ignored, and so are repeated submissions of the same request. Processing is generally quick, often within about twelve hours.

Barracuda is a reasonable list, run reasonably. Fix the cause, submit once, wait.

The ones to ignore, and the one that no longer exists

Not every listing deserves your attention, and pretending otherwise is how agencies burn a day on nothing.

UCEPROTECT operates three tiers. Level 1 lists individual IPs that have sent spam, which is fair enough. Levels 2 and 3 escalate to listing entire netblocks and, at Level 3, entire autonomous systems — so a single abusive customer somewhere on your provider’s network can list every IP that provider operates, including yours. Their delisting model offers paid express removal starting around $100.

My position on this is not neutral. Do not pay. Two reasons. First, paying to remove a listing you did not cause funds the model that produced it, and the model is widely regarded in the industry as coercive rather than protective. Second, and more practically, it buys you almost nothing: the major receivers do not use UCEPROTECT Level 2 or 3 precisely because the false positive rate of listing entire ASNs is unacceptable. You would be paying to clear a listing that is not affecting your delivery.

If a client’s monitoring flags UCEPROTECT L3, the correct response is to explain what it is and move on. If UCEPROTECT L1 lists you specifically, that one is worth a look, because it means something actually sent spam.

SORBS no longer exists. Proofpoint decommissioned it on 5 June 2024, and the service holds no reputation data. It had been a genuinely respected list since the early 2000s and was used by a great many systems.

This matters more than a historical footnote, because a number of “check 87 blocklists” tools still query it and still report on it, sometimes as a hit. If a monitoring report tells you that you are listed on SORBS, the report is wrong, and you should be suspicious of everything else on it too. Those multi-list checkers are the source of most unnecessary panic in this area: they weight a dead list, a vanity list and Spamhaus identically, then present a scary-looking count. Three listings out of 87 tells you nothing. Which three tells you everything.

If you use one of these tools, use it to discover which lists to look at, then confirm each one directly against that list’s own checker before acting.

What a listing actually costs

Worth being concrete, because “you are on a blocklist” is treated as either an apocalypse or a shrug depending on who is in the room, and neither is right. It depends entirely on which list.

Take a business sending 30,000 messages a month, split roughly as UK SMEs usually are: perhaps 45% to Microsoft 365 and Outlook, 30% to Gmail and Workspace, and 25% to everything else — smaller hosts, appliances, on-premise servers.

A Spamhaus ZEN listing hits the “everything else” bucket hardest, since that is where DNSBL rejection at connection time is most common, and reaches into the Microsoft slice too. You would expect to lose the majority of that 25% outright as hard bounces, plus meaningful damage elsewhere. Call it 8,000 to 10,000 messages a month failing to arrive, with the bounces themselves further degrading your sender metrics. If those are invoices, order confirmations or appointment reminders, the cost is not the mail — it is the support calls, the late payments and the missed appointments.

A SURBL listing distributes differently. It does not usually produce hard bounces; it produces spam-foldering across every receiver that consults it. Quieter, harder to notice, and often longer-lived for exactly that reason. Delivery reports look fine. Nobody replies.

A UCEPROTECT Level 3 listing, on the same mail, plausibly costs you nothing at all.

Same word, “listed”, three entirely different outcomes. Which is why the first job is always identifying the list rather than reacting to the count.

Delisting: the order that works

The sequence matters more than the paperwork.

Fix the cause first, and prove it to yourself. Every list, without exception, relists faster than it delists. If you cannot say precisely what sent the spam or what served the malware, you are not ready to submit anything. Find the compromised account, the open form, the injected file, the misconfigured relay.

Then check whether you need to ask at all. CSS and XBL expire by themselves once the source stops. If the IP is not urgent, three days of patience is a perfectly respectable strategy and carries no risk of a relist-loop reputation.

Submit once, with specifics. Name the cause, the date, the remediation and the date of remediation. “Compromised webmail credential on 14 March, used to send approximately 4,000 messages between 02:00 and 06:00 UTC; account disabled 06:40, password policy enforced, 2FA now mandatory, queue flushed” gets actioned. “We are a legitimate business, please delist” does not.

Do not submit repeatedly. Multiple requests for the same listing are ignored by Barracuda as policy and are unhelpful everywhere else. Once is a request; four times is a pattern.

Never pay. Nothing legitimate charges for removal.

The part nobody wants to hear

Some listings you cannot do anything about but wait.

If you are on shared hosting and the shared mail IP is listed because of another customer, you cannot delist it. You do not control the IP, you cannot submit a credible remediation account for someone else’s compromise, and the list operator will not take your word for it. The delisting is your host’s job, and your only real options are to escalate and to consider whether shared sending infrastructure is appropriate for mail that matters. The structural reasons this keeps happening are worth understanding in full — we set them out in why cPanel hosting gets your emails blacklisted.

If you have been relisted several times, the delisting gets slower each time, and that is deliberate. List operators track repeat offenders. The fourth request from an IP that has been listed and cleared three times in a month will not be handled with the same generosity as the first.

And if the listing is the automatic, evidence-driven kind and the evidence is real, no amount of correspondence shortens the wait. The clock starts when the sending stops, not when you write the email.

The way to spend less of your life on this is not better delisting technique. It is not sending from shared IPs when the mail matters, keeping your own tracking and link domains rather than borrowing someone’s, patching the small sites nobody remembers, and monitoring the handful of lists that actually influence delivery — Spamhaus ZEN, Spamhaus DBL, SURBL, Barracuda — rather than a dashboard of 87 that includes one which has not existed since 2024.


FXRM runs mail on dedicated infrastructure with correct PTR records, per-domain DKIM signing and DMARC enforcement, so a listing caused by somebody else’s compromised account is not a problem you inherit. If you are trying to work out which of your bounces matter, get in touch, or read more about our email hosting.

Avatar photo

Alison Ivers

Leave a comment

Your email address will not be published. Required fields are marked *