SpamAssassin is an open-source spam filter that runs hundreds of tests on an email and adds up a score. Each rule that matches adds points (or subtracts them for good signs), and if the total reaches the threshold, 5.0 by default, the message is marked as spam. Your report lists every rule that fired, so once you know what each one means, you can fix the cause rather than guessing.
How SpamAssassin scoring works
SpamAssassin does not make one big decision. It checks lots of small things: the headers, the structure of the message, the text, the links, and DNS-based reputation lists. Each test is a named rule with a score attached. The scores are added together:
- Positive scores point towards spam, for example an HTML-only message or a link to a blocklisted domain.
- Negative scores point towards legitimate mail, for example a valid DKIM signature or a sending IP on a reputable allowlist.
The default threshold is 5.0. A message scoring 5.0 or more is tagged as spam; below that it passes. Server administrators can change the threshold, and many do, so a score of 4.5 might pass on one server and fail on another. That is why you should aim for a score well below the threshold rather than just under it.
Scores also differ by installation. Default scores are set by the SpamAssassin project and can change between rule updates, administrators can override any of them, and many add third-party rule sets. Treat the scores in your report as a guide to what matters rather than fixed universal values.
Reading the report
SpamAssassin adds headers to messages it scans. They look like this (the scores here are illustrative):
X-Spam-Status: No, score=1.9 required=5.0 tests=DKIM_SIGNED,DKIM_VALID,HTML_IMAGE_RATIO_02, MIME_HTML_ONLY,SPF_PASS autolearn=no
The tests= list is the part to focus on. Rules starting T_ are test rules still being evaluated, usually with tiny scores. Now for the rules that turn up most often and what to do about them.
Common rules and how to fix them
HTML_IMAGE_RATIO_02, _04, _06, _08
These fire when the message has a lot of image area compared with the amount of text. The number suffix indicates the ratio band. Spammers have long hidden their message in a single image to avoid text filters, so an email that is mostly pictures looks suspicious.
Fix: add real text. A newsletter that is one large banner should have the headline, the offer and the call to action written in HTML text, with images as support. This also helps everyone whose email client blocks images by default.
MIME_HTML_ONLY
The message has an HTML part but no plain-text alternative. Legitimate email software almost always sends both.
Fix: send multipart/alternative with a text version. Most email platforms generate one automatically if you enable it. If your website sends mail with a library such as PHPMailer, set the AltBody property (or your library's equivalent).
MISSING_MID
The message has no Message-ID header. Every properly formed email should have one, and its absence suggests hand-rolled or broken sending software.
Fix: this almost always comes from a website script or device sending mail directly. Send through a proper mail server or a library that sets the header. A correct one looks like:
Message-ID: <20261027101500.3f9a1c@mail.example.co.uk>
URIBL_ and SURBL rules
Rules such as URIBL_BLACK or URIBL_DBL_SPAM mean a domain linked in your email appears on a URI blocklist. This includes links in your signature, tracking domains and image hosts, not just your main website.
Fix: find which domain is listed (your Mail Tester UK report checks links separately), and remove or replace the link if it is not yours. If it is your own domain, find out why it was listed, fix the cause, and request delisting. Avoid public URL shorteners, which are frequently abused and listed.
URIBL_BLOCKED is different. It means the scanning server's query to the blocklist was refused, normally because the server sends too many queries. It scores zero and says nothing about your links.RCVD_IN_ rules
Rules starting RCVD_IN_ check the IP addresses in your message's Received headers against DNS lists. Some are bad, such as RCVD_IN_SBL or RCVD_IN_XBL, which mean an IP on your path is listed for spam or as a compromised machine. Others are good: RCVD_IN_DNSWL_ rules indicate an allowlisted sender and subtract points.
Fix: for negative listings, see our blacklist removal guide. If you are on shared hosting or a shared IP at an email platform, raise it with your provider.
BAYES_ rules
Bayesian filtering learns from mail that the server's users have marked as spam or not spam. BAYES_00 means the text looks very much like legitimate mail and subtracts points; BAYES_99 and BAYES_999 mean it looks very much like spam and add a lot. Results in between are milder.
Fix: Bayes is trained per server, so you cannot control it directly. What helps is writing like a person: avoid heavy sales language, excessive capitals and exclamation marks, and phrases commonly found in spam. If a high Bayes score is consistent across tests, rewrite the copy.
KAM_ rules
Rules beginning KAM_ come from a widely used third-party rule set rather than SpamAssassin's core rules. They target patterns seen in current spam campaigns, such as particular phrasing, suspicious link structures or lookalike formatting. Their names usually hint at what they detect.
Fix: read the rule name, look at the matching part of your email, and change it. Because these rules track live spam trends, they can fire on legitimate mail that happens to resemble a current campaign.
FREEMAIL_ rules
FREEMAIL_FROM fires when you send from a free consumer address such as a Gmail or Outlook.com account. On its own it carries a small score. FREEMAIL_FORGED_REPLYTO is more serious: it means the From and Reply-To addresses are different free mail accounts, a common scam pattern.
Fix: send business mail from your own domain, with authentication in place. Our guides for Google Workspace and Microsoft 365 cover the setup.
Authentication rules: SPF, DKIM and DMARC
SpamAssassin also has rules for authentication: SPF_PASS, SPF_FAIL, SPF_SOFTFAIL, DKIM_VALID, DKIM_INVALID, DKIM_VALID_AU (valid and aligned with the From domain) and so on. Passing results subtract a little; failures add points. If any failure rules appear, fix authentication first, because it affects every mailbox provider, not just SpamAssassin. See our guides to SPF and DKIM.
How Mail Tester UK avoids double-counting
Mail Tester UK runs its own dedicated checks for SPF, DKIM and DMARC, and looks up your sending IP and domains on 20 blacklists. Each of those has its own section and its own effect on your score out of 10.
SpamAssassin has overlapping rules for the same things. If both counted, a single SPF failure or blacklist listing would cost you twice. To keep the score fair, the SpamAssassin portion of your result leaves out the rules covering authentication and IP blacklists that our own checks already score. You still see SpamAssassin's verdict on content, structure, headers and links, which is where it adds the most value.
Managed UK hostingdijihostManaged UK hosting with Plesk, Cloudflare DNS, daily backups and a domain included, for £1 a day + VAT. Old sites welcome.Host with dijihostA sensible target
Do not chase a perfect zero by stripping all personality out of your emails. Aim for:
- No rules indicating broken structure (
MISSING_MID,MIME_HTML_ONLY, malformed headers). - No blocklist hits on your IPs or links.
- Authentication passing, so the negative-scoring rules work in your favour.
- A total comfortably below 5.0, ideally below zero.
Remember that SpamAssassin is one filter among many. Gmail and Microsoft rely heavily on their own reputation and engagement data. A clean SpamAssassin score is necessary but not sufficient; our guide to why emails go to spam covers the rest.
Frequently asked questions
What is a good SpamAssassin score?
Lower is better. The default spam threshold is 5.0, so anything under that passes a default installation, but aim for zero or below to leave headroom for servers with stricter settings.
What does a negative SpamAssassin score mean?
Some rules subtract points because they indicate legitimate mail, such as a valid DKIM signature or a listing on a sender whitelist. A total below zero means the positive signals outweigh the negative ones.
Why do my SpamAssassin scores differ between servers?
Each server can change rule scores, add third-party rule sets, adjust the threshold, and train its own Bayes database. The same email can score differently on two systems.
What does URIBL_BLOCKED mean?
It means the receiving server's DNS lookup to the URIBL blocklist was refused, usually because of query limits. It is informational, scores nothing, and says nothing about your links.