Yandex Mail spam filter: how it thinks and how to stay out of it

Yandex Mail is the second-largest consumer email platform in CIS after Mail.ru, and the filter is noticeably more aggressive. Yandex engineers have talked publicly about their ML-first approach since the mid-2010s, and the architecture shows — domain reputation dominates the decision, content is a secondary signal, and the filter adapts fast to patterns that Gmail takes weeks to learn. If your campaign reaches Inbox on @mail.ru but lands in Spam on @yandex.ru, you're seeing this in action.

Yandex's share in the CIS market

Mail.ru is still #1 across CIS, but Yandex has eaten into its share steadily — especially among under-35 users and among professionals using Yandex.360 (the equivalent of Google Workspace). As of 2026, estimates place Yandex at roughly 30% of Russian consumer email volume, plus a growing slice of corporate mailboxes on @yandex-team.ru and custom-domain Yandex.360 tenants. If you're sending B2C, Yandex is ~30% of your addressable audience. If you're sending B2B to CIS SMEs, it's often more.

How the ML filter scores

Yandex's spam model is heavier on ML than rule-based filters like SpamAssassin. From public statements and observed behaviour, the weighting roughly is:

Domain reputation (largest single factor)

Built from aggregate recipient engagement across all Yandex users who have received mail from your domain. Opens, replies, moves out of Spam, complaints, deletions without open — all fed into a per-domain score. New domains start at a mediocre default and move from there. A single month of bad sending drags this score hard.

Per-recipient engagement

For each recipient, Yandex tracks whether they've engaged with mail from your domain. First sends to a given recipient are weighted pessimistically; repeated engaged recipients get a strong positive signal. This is why cold lists to Yandex users almost always get Junked on the first attempt.

Content classification

Standard ML content features: subject length, image ratio, link count, language match, HTML structure anomalies. Secondary to reputation but enough on its own to Junk a message if it looks outright spammy.

Yandex-specific features

A few signals Yandex weighs that other providers don't, or weigh less:

• Yandex.Advisor blacklist — a Yandex-maintained blocklist of domains flagged for fraud, tied to Yandex Browser's safe- browsing warnings. A hit here is near-instant Junk.
• Russian-language presence check — mail sent to a mailbox that predominantly uses Russian is scored more suspiciously if the content is English-only.
• Domain age on Yandex — not global domain age. How long Yandex has been seeing mail from your domain specifically.

postoffice.yandex.ru: setup and what it shows

Setup

1. Go to postoffice.yandex.ru and sign in with any Yandex account.
2. Add your sending domain. Verify via DNS TXT or a control-file at your domain root.
3. Enable the FBL via the Settings panel; Yandex delivers ARF-format complaints to an address you provide.

What it shows

• Domain reputation score, 30 / 90 / 365-day trend.
• Spam rate (user-reported complaints as %).
• Delivery rate and error breakdown by SMTP code (good detail here).
• Per-IP reputation if you send from multiple IPs.
• DKIM signing coverage — % of mail signed.

What it hides

• Folder placement — Yandex will not tell you what % landed in Spam folder vs Inbox. Complaint rate is the only proxy.
• Content-level scores — the ML classifier's decision is opaque. No "your content scored X" readout.
• Per-message diagnostics — you cannot look up a single message's path through the filter.

The five quickest fixes

1. DKIM strictness

Yandex's DKIM parser is strict on selector format and key length. 2048-bit RSA, simple selector name (no underscore prefix, no starting digit), record that resolves cleanly on dig without truncation. Every DKIM-signed message should show dkim=pass in the Yandex-received-headers section of a test delivery.

2. SPF alignment

Yandex treats SPF softfail as a strong negative signal even when the rest of authentication passes. Keep Return-Path aligned to From, publish SPF at the root, end with -all once your inventory is clean.

3. Warm up specifically on Yandex

A "warmed" domain in Gmail's sense is a fresh domain to Yandex. Include engaged Yandex recipients in your warm-up runs — not just Gmail addresses. 2–3 weeks of low-volume, engagement- positive sending to @yandex.ru specifically.

4. Complaint minimisation

Yandex uses a similar 0.2–0.3% ceiling to Gmail and Yahoo, but the complaint button is more prominent in the Yandex Mail UI — effective complaint rates tend to run 1.5–2x Gmail for the same list. Aggressively suppress anyone who has opened less than once in 6 months.

5. Cyrillic encoding

Same issue as Mail.ru: Cyrillic subject lines must be correctly MIME-encoded (=?UTF-8?B?...?=). Plain UTF-8 subject headers often pass Gmail and get Junked on Yandex. Test a Cyrillic subject send to a seed mailbox before scaling.

Yandex.Advisor blacklist

Yandex maintains a domain blocklist called Advisor that powers safe- browsing warnings in Yandex Browser and also feeds directly into the mail filter. A listing here is effectively terminal for Yandex deliverability until resolved. Check your sending domain and any tracking domains at yandex.com/support/safe-browsing/. Delisting requests are possible but slow — weeks, not days.

Free placement test for Yandex

The inbox-check tool includes @yandex.ru as a default seed in every test, with a real screenshot of the mailbox as a Yandex user would see it. Authentication results, SpamAssassin + Rspamd scores, full headers. No signup, no daily cap. Pair it with postoffice.yandex.ru for the complete picture — placement test tells you where your message landed; postoffice tells you what Yandex thinks of your domain over time.

Frequently asked questions