Why your emails are going to spam—and what to test first.

Do not change twelve things at once

When an email lands in spam, the first instinct is often to rewrite the subject line. That can be the least useful first move. Mailbox providers consider the sender, domain, IP, authentication, audience behavior, sending pattern, links, message structure, and recipient-specific history. A cleaner subject cannot rescue a compromised server or a list full of people who never asked to hear from you.

Start by confirming the problem. Send the exact production message through the exact production route to a controlled mix of Gmail, Outlook, Yahoo, and business inboxes. Keep the sender and tracking links intact. If the problem clusters at one provider, investigate provider-specific reputation and policy. If spam placement is broad, start with identity, list quality, and infrastructure.

Test first. A placement result narrows the search before you destroy the evidence.

Causes 1–3: authentication and identity

Cause 1 is missing authentication on the real message. SPF or DKIM may exist in DNS but fail on the route your ESP actually used. Cause 2 is DMARC misalignment: SPF passes for a provider-owned return path or DKIM passes for a provider domain, while neither aligns with the visible From address. Cause 3 is an identity recipients do not recognize, such as a new From domain, a free webmail address used for bulk mail, or an abrupt sender-name change.

Inspect delivered headers, not only the ESP settings page. Find `spf=pass`, `dkim=pass`, and `dmarc=pass`, then check which domains were evaluated. If alignment fails, configure the provider's branded sending domain, custom return path, or DKIM signing domain. Keep the visible From identity stable enough that recipients know who is speaking.

Causes 4–6: reputation, complaints, and list quality

Cause 4 is weak domain or IP reputation. There is no single reputation number shared by every mailbox provider. Use provider dashboards, delivery errors, complaint history, blocklists, and repeated placement tests as evidence. Cause 5 is a complaint rate that tells a provider recipients did not want the mail. Cause 6 is poor list quality: invalid, recycled, purchased, scraped, or long-dormant addresses can create bounces, traps, and negative engagement.

The fix is operational, not cosmetic. Stop questionable acquisition sources. Suppress hard bounces, complaints, and unsubscribes everywhere, not just in one ESP. Send first to people with a recent relationship and a clear expectation. If the sending IP or account was compromised, contain it before requesting delisting or trying to repair reputation.

Causes 7–9: volume, infrastructure, and shared risk

Cause 7 is an unexpected volume spike. A domain that normally sends hundreds of messages can look very different when it suddenly sends tens of thousands. Cause 8 is unstable or misconfigured infrastructure: missing PTR records, formatting problems, repeated deferrals, or a new route that does not match the tested setup. Cause 9 is shared infrastructure risk. Another sender can affect the reputation of a shared IP, while your own audience and domain reputation still remain your responsibility.

Increase volume gradually to engaged recipients, watch provider responses, and keep marketing and transactional streams separate when their risk differs. If you use a managed ESP, ask which parts of the route are shared and which reputation signals are yours. Moving to a dedicated IP is not automatically an upgrade; it gives you control and the full burden of building and maintaining the IP's reputation.

Causes 10–12: links, content, and provider differences

Cause 10 is a risky link path: public shorteners, unfamiliar tracking domains, compromised destinations, or long redirect chains. Cause 11 is a message that is difficult to trust or parse—one giant image, misleading urgency, hidden identity, broken unsubscribe handling, malformed HTML, or content that differs sharply from what the audience expected. Cause 12 is provider-specific filtering. Gmail, Outlook, Yahoo, and private business gateways do not use one shared model.

Open every post-tracking URL and make the message understandable without images. Replace avoidable shorteners with direct or branded links. Do not hunt for a magical banned word list; read the whole message as a skeptical recipient. When one provider is the outlier, combine its placement result with that provider's reputation dashboards and delivery codes instead of flattening the problem into a global score.

A repair sequence that preserves what you learn

First, fix hard authentication failures and any security incident. Second, stop harmful audience sources and suppress invalid or complaining recipients. Third, stabilize the route and volume. Fourth, investigate provider-specific reputation and blocks. Fifth, simplify suspicious links or message structure. Then send the same representative email again and compare the same provider segments.

Keep a short change log. Record the sender, subject, provider, seed segment, and one change made between tests. If placement improves, you have a plausible cause worth monitoring. If it does not, do not pile on five more edits; return to the evidence and choose the next strongest hypothesis.

What a placement test can—and cannot—prove

A seed test shows where controlled inboxes placed one specific message at one point in time. It is useful for provider patterns, broad failures, and before-and-after comparisons. It cannot reproduce every subscriber's engagement history or guarantee the folder outcome for a whole audience. Missing results may also mean delay, rejection, or a matching problem; they are not automatically spam.

Use the result as a decision instrument. If important providers show clean placement and the setup evidence is sound, you have more confidence to proceed. If one provider shows a clear spam cluster, you have a smaller investigation. That is much better than rewriting the campaign because an online scanner disliked three words.