菜单
Understand what determines inbox vs spam placement. Use when emails land in spam, investigating Gmail Promotions tab, diagnosing deliverability issues, or optimizing engagement signals.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Diagnose email problems and plan email infrastructure. Use when emails land in spam, bounce rates spike, deliverability drops, open rates decline, providers throttle sending, you need to set up email for a new product, or you're not sure which email skill to use.
Choose and configure an agent-native inbox provider for AI agents that send and receive email. Use when giving an AI agent its own email address, setting up two-way agent email communication, comparing agent inbox platforms, or adding guardrails to autonomous email workflows.
Build dashboards, alerts, and monitoring systems for email sending operations. Use when setting up deliverability monitoring, configuring alert thresholds, checking blocklists, building email metrics dashboards, or responding to deliverability incidents.
Classify inbound email replies by intent and route them appropriately. Use when building reply handling for outreach, support, or transactional email - detecting interested, not interested, OOO, bounce, unsubscribe, forwarded, and question replies.
Maintain email conversation context across messages using threading headers. Use when building thread reconstruction, linking replies to conversations, detecting thread hijacking, stripping quoted content, or providing thread context to AI agents.
Receive, parse, and process incoming email via provider webhooks. Use when setting up inbound email handling, parsing MIME messages, extracting content from replies, detecting threads, filtering spam on inbound, or routing incoming messages.
| name | inbox-placement |
| description | Understand what determines inbox vs spam placement. Use when emails land in spam, investigating Gmail Promotions tab, diagnosing deliverability issues, or optimizing engagement signals. |
| license | MIT |
Understand the factors that determine whether your email lands in the inbox, spam folder, or a secondary tab - and how to control them.
domain-authentication - SPF, DKIM, DMARC setup (a prerequisite for inbox placement)sender-reputation - building and monitoring the reputation signals that drive placementemail-warmup - ramping volume on new domains/IPs to establish reputationspam-filter-avoidance - deep dive on content patterns that trigger filtersbounce-handling - managing bounces that degrade your placement over timesuppression-lists - keeping your list clean to protect engagement metricsInbox placement is the outcome of a scoring system. Every major mailbox provider runs incoming email through a multi-signal classifier that produces a disposition: inbox, spam, or a category tab (Gmail Promotions, Outlook Other). The classifier weighs several signal categories, and their relative importance has shifted significantly since 2023.
Here is the rough weighting that major providers apply, from most to least influential:
| Signal | Weight | What it means |
|---|---|---|
| Sender reputation | ~40% | Domain and IP history, complaint rates, bounce rates |
| Engagement signals | ~25% | Opens, clicks, replies, deletes, moves-to-spam per recipient |
| Authentication | ~15% | SPF/DKIM/DMARC pass/fail, alignment |
| Content analysis | ~10% | Spam phrases, formatting, link patterns, HTML quality |
| Infrastructure | ~10% | IP age, domain age, DNS configuration, TLS |
These percentages are approximate - providers don't publish exact weights, and they vary by provider. But the ranking is consistent: reputation and engagement matter far more than content tricks. A sender with excellent reputation can use the word "free" in a subject line without consequences. A sender with poor reputation will land in spam even with perfect content.
This is the single most important thing to understand about inbox placement: you cannot content-hack your way into the inbox if your reputation or engagement signals are bad.
Authentication is table stakes, not a differentiator. Without it, you're filtered or rejected before any other signal is evaluated.
| Requirement | Gmail | Microsoft | Yahoo |
|---|---|---|---|
| SPF pass + alignment | Required for bulk | Required for bulk (May 2025) | Required for bulk |
| DKIM pass + alignment | Required for bulk | Required for bulk (May 2025) | Required for bulk |
| DMARC (p=none minimum) | Required for bulk | Required for bulk (May 2025) | Required for bulk |
| TLS for SMTP | Required | Required | Required |
| Valid PTR records | Required | Recommended | Required |
"Bulk" means 5,000+ messages per day to that provider's users.
p=reject, the message is bounced. With p=quarantine, it goes to spam. With p=none, it's delivered but with a negative signal attached.Since November 2025, Gmail has moved beyond soft filtering for non-compliant bulk senders - emails may be rejected at the SMTP level with 4xx/5xx errors rather than silently filtered.
See the domain-authentication skill for full setup details.
Engagement is the signal category that has grown most in influence. Providers track per-recipient and per-sender engagement patterns over time, and these directly affect future placement.
| Signal | Weight | Notes |
|---|---|---|
| Replies | Very high | Strongest positive signal. A reply tells the provider this is a wanted conversation. |
| Moves from spam to inbox | Very high | Explicit user correction - tells the provider they got it wrong. |
| Opens | Medium | Tracked via pixel, but less reliable due to Apple Mail Privacy Protection and image blocking. |
| Clicks | Medium | Indicates content relevance. But excessive link tracking can itself trigger filters. |
| Adding sender to contacts | High | Strong trust signal, especially at Gmail. |
| Starring or flagging | Medium | Indicates importance to the recipient. |
| Signal | Weight | Notes |
|---|---|---|
| Mark as spam | Very high | The single most damaging action. Google requires complaint rates below 0.1% for healthy delivery, and starts filtering at 0.3%. |
| Delete without opening | Medium | Repeated pattern indicates unwanted email. |
| Ignore (no interaction) | Low-medium | Accumulates over time. If a recipient ignores 10 consecutive emails, future sends are increasingly likely to be filtered. |
| Unsubscribe via header | Low | Better than a spam complaint. Providers view List-Unsubscribe usage as a positive infrastructure signal even though the recipient is leaving. |
Even if a recipient never complains, persistent non-engagement degrades your placement over time. Here is how it works in practice:
This is a death spiral. The fix is to stop sending to disengaged recipients before the provider starts penalizing you. A fatigue scoring system helps - tracking sends, opens, and time since last engagement per recipient, and suppressing or reducing frequency when the score crosses a threshold.
A practical fatigue model scores contacts on these factors:
Contacts scoring above 70 on a 0-100 scale should stop receiving email. Between 40 and 70, reduce frequency. Below 40, safe to send.
Content filtering is less influential than reputation, but it can still tip borderline emails into spam. Modern filters use ML classifiers, not just keyword matching, but certain patterns reliably trigger negative scoring.
These are the phrase patterns that consistently trigger filters across providers:
Urgency/pressure language:
Financial promises:
Engagement bait:
Deceptive calls to action:
Health/pharmaceutical:
Context matters more than individual words. Modern filters evaluate patterns, tone, and how words work together. "Free shipping on orders over $50" in a transactional receipt is fine. "FREE!!! Click here for your FREE gift!!!" is not. The combination of trigger words, urgency, excessive formatting, and lack of established sender relationship is what pushes emails into spam.
ALL CAPS in subject lines or body text. Occasional emphasis is fine. Full sentences in caps score negatively.
Excessive exclamation marks. One is fine. Three in a row ("!!!") is a spam signal. This applies to both subject lines and body content.
Colored or oversized text. Large red text, especially combined with urgency language, mimics classic spam formatting.
Misleading subject lines. Starting with "Re:" or "Fwd:" when it's not a reply or forward. Providers detect this and penalize it.
Single-image emails. An email that is just one large image with no text is a strong spam signal. Spam filters can't analyze image content as easily as text, so spammers historically used image-only emails to evade keyword filters. Providers learned to flag this pattern.
Link-to-text ratio. Keep it reasonable - 2 to 3 links per email is standard. An email with 15 links and minimal text looks like a phishing attempt.
URL shorteners. Services like bit.ly, tinyurl, and similar shorteners are heavily scrutinized because they obscure the destination. Use full URLs or your own branded short domain.
Mismatched anchor text and URLs. If the visible link text says "google.com" but the href points to a different domain, that's a phishing signal. Filters catch this.
HTTP links (not HTTPS). Insecure URLs in email are a negative signal. All links should use HTTPS.
Too many tracking redirects. Every click-tracking system adds a redirect. Multiple layers of redirects look suspicious.
The general guideline is 60% text, 40% images or less. However, research shows that emails with 500+ characters of text are not meaningfully penalized regardless of image count. The real risk is emails with very little text and one or more large images - this pattern strongly correlates with spam.
Poorly structured HTML is correlated with lower inbox placement. Emails failing HTML validation are 18-25% more likely to land in spam at Gmail and Outlook. Common issues:
Gmail uses three main destinations for email, and each has different implications.
Reserved for personal, conversational email. To land here:
Gmail routes marketing and commercial email here. Triggers include:
List-Unsubscribe headers (signals marketing email)The Promotions tab is not spam. As of 2025, 54% of Gmail users check their Promotions tab daily. Open rates in Promotions (~19%) are only slightly lower than Primary (~22%). The real impact on total campaign performance is often less than half a percentage point.
In September 2025, Google changed the Promotions tab to sort by relevance rather than recency by default. This means emails from brands the recipient engages with most appear at the top. High engagement actually benefits you more in Promotions than it used to.
Promotions tab annotations let you add product carousels, deal badges, and expiration dates to your promotional emails. These increase visibility within the tab rather than trying to escape it.
Content, reputation, or authentication failures land you here. Unlike Promotions, spam placement is actively harmful - it degrades your reputation further and creates the engagement death spiral described above.
When a user drags an email from Promotions to Primary, Gmail learns that preference for that sender. Enough users doing this for a given sender can shift the default placement. The reverse is also true - users moving emails from Primary to Promotions or spam trains the classifier against you.
Microsoft Outlook uses a similar but distinct system.
Prioritizes emails that seem important based on:
Routes lower-priority email here:
Outlook's Focused Inbox continuously reevaluates placement based on recipient behavior. If a user starts opening newsletters from a specific sender more frequently, future emails from that sender may move to Focused. If they start ignoring emails from a previously-engaged sender, those emails migrate to Other.
Microsoft announced requirements for high-volume senders to Outlook.com and Hotmail, effective May 5, 2025:
Your list quality directly affects inbox placement through bounce rates, spam trap hits, and engagement ratios.
| Bounce rate | Impact |
|---|---|
| < 2% | Healthy. No placement impact. |
| 2-5% | Yellow zone. Providers start watching more closely. |
| 5-10% | Active filtering likely. Some providers throttle delivery. |
| > 10% | Severe. Expect spam placement or outright blocking. |
Hard bounces (address doesn't exist) are worse than soft bounces (temporary failures). Every hard bounce should immediately suppress that address so it's never tried again.
Spam traps are email addresses operated by providers and blocklist operators specifically to catch senders with poor list hygiene. There are two types:
Pristine traps: Addresses that were never used by a real person. They exist solely to catch scraped or purchased lists. Hitting one is strong evidence of bad list acquisition.
Recycled traps: Addresses that once belonged to real users but were abandoned and repurposed. They catch senders who don't clean their lists - if the address has been inactive for a year and you're still sending to it, your list hygiene is poor.
Hitting spam traps has an outsized impact on reputation. There's no way to identify trap addresses in your list - the only defense is good acquisition practices (never buy lists) and regular list cleaning (suppress addresses that haven't engaged in 6+ months).
A high rate of "unknown user" rejections (550 5.1.1 responses) tells providers you don't verify addresses before sending. This is especially relevant for AI agents that construct recipient addresses dynamically - guessing "contact@company.com" without verification produces bounces that count against you.
These are baseline factors that affect placement before any email is sent.
email-warmup skill).Domains registered less than 30 days ago are treated with suspicion by most providers. Domains less than 7 days old are frequently blocked outright for bulk sending. Register your sending domain well in advance of your first campaign.
Inbox placement testing (seed testing) measures where your emails actually land across providers before you send to your real list.
Seed testing shows how a neutral, new subscriber would receive your email. It does not account for:
Seed testing is best used as a baseline diagnostic, not as a guarantee of real-world placement. Run seed tests when:
Services like Validity Everest, GlockApps, MailMonitor, Mailtrap, and InboxMonster maintain seed lists and provide placement dashboards. Some ESPs (Mailgun, SendGrid) have built-in placement testing.
Inbox placement is the result of all these signals evaluated together. Here is a practical framework for diagnosing placement issues:
dig on your SPF, DKIM, and DMARC records. Send a test email and check the headers for dmarc=pass, spf=pass, dkim=pass. If any fail, fix them before investigating further.This is a different problem from spam placement and usually isn't worth fighting. But if Primary/Focused placement matters for your use case:
Obsessing over spam trigger words while ignoring reputation. "Free" in a subject line won't send you to spam if your reputation is good. A perfectly worded email will land in spam if your reputation is bad. Fix reputation first, content second.
Trying to game the Promotions tab. Marketers spend enormous effort trying to escape Gmail's Promotions tab. The actual open rate difference is small (~3%), and the September 2025 relevance-based sorting means high-engagement senders benefit from Promotions placement. Focus on making your Promotions-tab emails worth opening instead.
Sending to your entire list regardless of engagement. Every email to a disengaged recipient hurts your placement for everyone else. Segment by engagement. Stop sending to contacts who haven't opened or clicked in 90 days. Re-engage them with a specific win-back campaign, and if that doesn't work, suppress them.
Not monitoring complaint rates. Google's threshold is 0.1% for healthy delivery, with problems starting at 0.3%. Many senders don't check until they're already in trouble. Set up Google Postmaster Tools and feedback loops from your ESP on day one.
Assuming authentication equals deliverability. SPF/DKIM/DMARC passing is necessary but not sufficient. It gets you past the front door. Reputation, engagement, and content determine which room you end up in.
Using URL shorteners in email. bit.ly, tinyurl, and similar services are heavily penalized because they obscure link destinations. Use full URLs or set up a branded redirect domain.
Sending image-only emails. A single large image with no text content is one of the strongest spam signals. Always include substantive text content alongside images.
Ignoring Microsoft. Many deliverability guides focus exclusively on Gmail. Microsoft (Outlook.com, Hotmail, Office 365) has its own filtering stack, its own reputation system, and as of May 2025, its own bulk sender requirements. Test and monitor placement at both.
Buying or scraping email lists. Purchased lists contain spam traps, invalid addresses, and people who never consented to hear from you. The bounce rates and complaint rates from a purchased list can destroy a domain's reputation in a single send. There is no shortcut to list building.
Not suppressing bounced addresses immediately. Every subsequent send to a known-bad address compounds the reputation damage. Automate suppression on hard bounce - the address should be blocked before the next send attempt even reaches your provider.