Email deliverability issues show up in four ways: a bounce, a deferral, a message that lands in spam, or silence from people who should have replied. Each has a different cause, and each leaves evidence in a different place. A bounce carries a status code that the receiving server wrote. A deferral carries a retry signal. A filtered message carries an Authentication-Results field you can read. This guide sorts the symptoms first, then walks each family of causes with the rules and error text that Google, Yahoo, Microsoft, Spamhaus and Amazon publish.
Every page cited here was read in November 2026. We have no connection with any of these companies, and providers change their requirements often, so recheck the linked page before you act on a number. The sample logs and headers are made up for illustration. The code ran under Python 3.15 and dig on 21 November, and the DNS answers are live. BillionVerify is an email verification service, and it appears only where a bad address list causes the damage.
Key facts (checked November 2026):
- Google's Gmail SMTP errors and codes page gives a separate error for each failed requirement: 5.7.26 for an unauthenticated sender, 5.7.27 for SPF, 5.7.30 for DKIM and 5.7.40 for a missing DMARC policy.
- Google's FAQ on the sender rules asks for a user-reported spam rate below 0.1%, never reaching 0.3%, and restores delivery help to a bulk sender after 7 consecutive days below 0.3%.
- Yahoo's SMTP error codes page allows a retry after a 4xx reply, but advises against retrying a 5xx reply and tells operators to remove that address.
- Microsoft's NDR reference for Exchange Online lists 5.7.509 for a DMARC reject and 5.7.511 for a banned IP address, with a delist route for the second.
- Amazon's SES enforcement FAQ puts an account under review at a 5% bounce rate or a 0.1% complaint rate, and may pause sending at 10% or 0.5%.
- RFC 7208, section 4.6.4, limits an SPF check to ten terms that trigger DNS queries, and a permanent error follows beyond that.
Sort the Symptom Before You Change Anything
Start with what the receiving server did, because the first digit of its reply already narrows the cause. RFC 5321, section 4.2.1, defines a 4yz reply as transient: the command was refused, the condition is temporary, and the client should try again. A 5yz reply is permanent, and the client should not repeat the exact request. The enhanced codes in RFC 3463 add detail. A 4.X.X code marks a persistent transient failure, a 5.X.X code marks a permanent failure, X.1.1 stands for a bad destination mailbox address, and X.7.XXX covers security or policy status.
That split describes the reply, not the fault. A 5.7.x code means a policy blocked the message, and the policy may target your domain, your IP address or a single link in the content. A 4.7.x code asks you to slow down, and an unresolved sender problem can harden into a block if you keep pushing. Most email deliverability issues fall into five groups, and the table maps each symptom to the evidence you hold and the likeliest cause to check first.
| Symptom | Where the evidence is | Cause group to check first |
|---|---|---|
| Bounce with a 5.1.x code | Bounce message, sending platform log | List quality |
| Bounce with a 5.7.x code | Bounce message, dashboards from the mailbox providers | Authentication, reputation or a blocklist |
| Deferral with a 4.7.x code | Queue log, delivery error charts | Volume, rate, DNS or reputation |
| Accepted, but sent to the junk folder | Raw headers, complaint data | Authentication, complaints or content |
| Accepted, but no engagement | Replies, clicks, a seed mailbox | Placement, or an audience that does not want the email |
The last row needs care. Google's sender guidelines state that Google does not track open rates, cannot verify the figures third parties report, and that a low open rate is not necessarily an accurate sign of a deliverability or spam problem. Treat a falling open rate as a prompt to look at the other evidence, never as a diagnosis on its own.
Deferrals explain a common false alarm. The same RFC, in section 4.5.4.1, requires an SMTP client to queue mail it cannot deliver at once and retry it later, with a retry interval of at least 30 minutes and a give-up time that generally needs to be at least 4 to 5 days. A deferred message is therefore not lost yet. Review how your platform retries before you resend by hand, or the same email may arrive twice.
Read the Reply Text Before Anything Else
The reply text is written by the receiving server, and it usually names the cause. Each large provider publishes its own reference, and the three below cover most consumer inboxes you will ever write to. Read the code first and the wording second, because a single code can carry several different causes.
Gmail Replies
Gmail appends gsmtp to every error, and gcdp when a custom rule from a Google Workspace administrator caused it. Its error table was last updated on November 10, and it separates reputation, authentication, rate and infrastructure faults into distinct codes. The table condenses the rows senders meet most, with the first action each calls for.
| Code | What Gmail reports in brief | First action |
|---|---|---|
| 550 5.1.1 | The account does not exist | Suppress the recipient |
| 421 4.7.28 | An unusual rate of email from your sending IP address or domain | Stop for 10 minutes, then resume on one connection |
| 421 4.7.0 | Very low reputation of the sender, by IP address or by domain | Review the reputation data, slow down |
| 550 5.7.26 | Unauthenticated sender, or blocked by the domain's DMARC policy | Fix SPF, DKIM and alignment |
| 5.7.27 and 5.7.30 | A failed SPF result or a failed DKIM result, for bulk senders | Repair the failing protocol |
| 5.7.40 | No DMARC record, or no policy inside it | Publish a record |
| 451 4.7.23, 550 5.7.25 | No PTR record, or forward DNS that does not match | Correct reverse DNS |
| 4.7.29 and 5.7.29 | The email was not sent over TLS | Turn on TLS |

Screenshot: the table in Google's error reference that covers the authentication, rate and TLS blocks, viewed in Nov 2026.
Two points need care. The 4.7.x versions of the authentication codes are rate limits, and Google's FAQ explains that since November 2025 non-compliant traffic meets temporary and permanent rejections. Read each 4.7.x authentication code as the early form of its 5.7.x twin. The 4.7.28 code also hides which quota you hit. When the message does not name it, the guidelines advise assuming that your DKIM, SPF and IP quotas are all affected, and the IP quota is shared by every sender on that address.
Yahoo Replies
Yahoo sorts its errors by theme instead of by number. A 421 or 451 reply is temporary, and Yahoo lists the usual reasons: odd traffic from your sending server, a spam-like message, complaints from Yahoo users, busy servers, a name server that was briefly down, or authentication that could not be decided because of a temporary error. A 553 or 554 reply is final, and the reasons include an invalid address, a failed DMARC or DKIM policy, content Yahoo will not accept, suspicious behavior and a Spamhaus listing.
Three of its categories help with diagnosis. Unsolicited volume can mean you set up the IP recently and did not raise traffic slowly. Excessive unknown recipients points to a large share of invalid addresses, and Yahoo warns that such a server may be an open relay or an open proxy. An unresolvable domain in the MAIL FROM or From header is checked with an SOA query, so a subdomain used there needs an SOA record as well as an A or MX record.

Screenshot: Yahoo's error code page, where the reasons behind 553 and 554 errors and the advice not to retry them are listed, as of Nov 2026.
Microsoft Replies
Microsoft documents its non-delivery reports for Exchange Online, and the same codes turn up when you write to company mailboxes hosted there. A 5.1.1 report means the recipient mailbox does not exist. A 5.7.509 report states that the sender's domain in the 5322.From address does not pass DMARC and enforces reject. A 5.7.511 report means the IP address was banned, and Microsoft gives an email route for requesting removal. A 5.7.513 report means the recipient's company put your server on a custom block list, so only that company can undo it.
Two more codes point at the sender. A 5.7.506 reply tells you your server introduced itself with the receiving server's name instead of its own fully qualified domain name, which Microsoft calls typical spambot behavior. The 5.7.501 to 5.7.503 reports mean a sending account was banned for spam activity, and Microsoft advises resolving the account issues and resetting its credentials. For consumer Outlook.com addresses, the announcement for high-volume senders names the rejection text 550; 5.7.515 Access denied, sending domain [SendingDomain] does not meet the required authentication level.

Screenshot: Microsoft Learn's NDR reference at the 5.7.509, 5.7.511 and 5.7.513 rows, taken in Nov 2026.
Group Your Bounce Log by Cause
A single reply is an anecdote. A hundred replies grouped by cause and provider show you where to look. A pattern that crosses providers points at your own setup, a pattern at a single provider points at that provider's rules, and a pattern inside the address group points at the list.
The script below applies the codes from the three pages above to a made-up log of 14 replies. The rule set is short on purpose, so extend it with the codes your own platform reports.
import re
from collections import Counter
# (pattern on the enhanced status code, cause, what to do), from the Gmail, Yahoo and Microsoft pages
RULES = [
(r"[45]\.1\.[123]$|5\.2\.1$", "list", "suppress the address"),
(r"[45]\.7\.(26|27|30|32|40|515)$|5\.7\.509$", "authentication", "fix SPF, DKIM, DMARC alignment"),
(r"[45]\.7\.(23|24|25|29)$|5\.7\.506$", "infrastructure", "fix PTR, HELO, SPF terms or TLS"),
(r"[45]\.7\.28$|4\.2\.1$|5\.7\.508$", "rate", "slow down, spread the sends"),
(r"5\.7\.51[13]$", "blocklist", "find the listing, fix the cause, request removal"),
(r"[45]\.7\.[01]$", "reputation", "check Postmaster Tools, SNDS, list source"),
(r"[45]\.2\.2$", "recipient", "retry later, suppress if it repeats"),
]
def classify(reply: str):
match = re.match(r"(\d{3})[ -](\d\.\d{1,3}\.\d{1,3})", reply)
if not match: # some replies carry no enhanced code, so fall back to the wording
if re.search(r"doesn't have a .* account|does not exist", reply):
return "list", "suppress the address"
return "unreadable", "read the raw text"
enhanced = match.group(2)
for pattern, cause, action in RULES:
if re.match(pattern, enhanced):
return cause, action
return "other", "read the full reply text"
LOG = """gmail.com|550 5.1.1 The email account that you tried to reach does not exist.
gmail.com|550 5.1.1 The email account that you tried to reach does not exist.
gmail.com|550 5.1.1 The email account that you tried to reach does not exist.
gmail.com|421 4.7.28 Gmail has detected an unusual rate of email originating from your IP address.
gmail.com|421 4.7.28 Gmail has detected an unusual rate of email originating from your IP address.
gmail.com|550 5.7.26 This email has been blocked because the sender is unauthenticated.
outlook.com|550 5.7.509 Access denied, sending domain does not pass DMARC verification.
outlook.com|550 5.7.509 Access denied, sending domain does not pass DMARC verification.
outlook.com|550 5.7.511 Access denied, banned sender
outlook.com|550 5.1.1 Bad destination mailbox address
yahoo.com|554 delivery error: dd This user doesn't have a yahoo.com account [-5]
yahoo.com|421 4.7.0 Try again later, closing connection.
example.org|451 4.7.23 The sending IP address for this message doesn't have a PTR record.
example.org|452 4.2.2 The recipient's inbox is out of storage space."""
counts = Counter()
for line in LOG.splitlines():
domain, reply = line.split("|", 1)
cause, action = classify(reply)
counts[(cause, domain)] += 1
total = sum(counts.values())
for (cause, domain), n in sorted(counts.items(), key=lambda kv: (-kv[1], kv[0])):
print(f"{n} of {total} {cause:15}{domain}")
# 3 of 14 list gmail.com
# 2 of 14 authentication outlook.com
# 2 of 14 rate gmail.com
# 1 of 14 authentication gmail.com
# 1 of 14 blocklist outlook.com
# 1 of 14 infrastructure example.org
# 1 of 14 list outlook.com
# 1 of 14 list yahoo.com
# 1 of 14 recipient example.org
# 1 of 14 reputation yahoo.com
Read the output as shares. Five of the 14 replies are address faults spread over three providers, which blames the list. The authentication replies come from both Gmail and Outlook.com, so they point at DNS and the From domain, not at a single provider's mood. The lone 4.7.0 reply from Yahoo has no cause in its wording, so the script only flags it as a possible reputation signal, and you should read the full text before acting on it.
Authentication Failures: SPF, DKIM and DMARC
Authentication is the most mechanical group of causes, because the receiver ran three tests and wrote the results into a header. Consumer Gmail inboxes expect either SPF or DKIM from every sender, and anyone who passes 5,000 messages in a day also needs DMARC, with the From domain aligned to either authenticated domain. The guidelines add that unauthenticated messages might be marked as spam or rejected with a 5.7.26 error. The sections below follow the order in which a receiver reads the evidence.
Read the Authentication-Results Header
RFC 8601 defines the header. It begins with the identifier of the server that ran the tests, then lists each method with its verdict and the property it tested. The DKIM verdicts are none, pass, fail, policy, neutral, temperror and permerror, and the property header.d names the signing domain. In Gmail you open the message, choose More, then Show original, and copy the full header, as Google's header tracing page describes. An email header analyzer reads a pasted block and separates the three verdicts for you.
The script below reads a made-up header that shows the most confusing pattern: SPF and DKIM both pass, yet DMARC fails. It uses only the standard library.
from email import message_from_string
RAW = """Authentication-Results: mx.example.net;
dkim=pass header.d=mailer.example.net header.s=k1;
spf=pass smtp.mailfrom=bounce@bounces.example.net;
dmarc=fail header.from=example.com
From: Billing <billing@example.com>
Subject: Your invoice
Hello
"""
def aligned(a: str, b: str) -> bool:
"""Relaxed alignment for registered domains: same name or one is a subdomain of the other."""
return a == b or a.endswith("." + b) or b.endswith("." + a)
def read_auth_results(raw: str):
value = message_from_string(raw)["Authentication-Results"]
parts = [p.strip() for p in " ".join(value.split()).split(";")]
results = {}
for part in parts[1:]: # parts[0] is the authserv-id, the server that wrote the verdicts
method, _, rest = part.partition("=")
words = rest.split()
props = dict(w.split("=", 1) for w in words[1:] if "=" in w)
results[method] = (words[0], props)
return parts[0], results
server, results = read_auth_results(RAW)
print("checked by:", server)
for method, (verdict, props) in results.items():
print(f"{method:6} {verdict:5} {props}")
from_domain = results["dmarc"][1]["header.from"]
dkim_domain = results["dkim"][1]["header.d"]
spf_domain = results["spf"][1]["smtp.mailfrom"].split("@")[1]
print("DKIM domain aligned with From:", aligned(dkim_domain, from_domain))
print("SPF domain aligned with From: ", aligned(spf_domain, from_domain))
# checked by: mx.example.net
# dkim pass {'header.d': 'mailer.example.net', 'header.s': 'k1'}
# spf pass {'smtp.mailfrom': 'bounce@bounces.example.net'}
# dmarc fail {'header.from': 'example.com'}
# DKIM domain aligned with From: False
# SPF domain aligned with From: False
Both tests passed, but for the sending platform's domains, and neither domain matches the From domain, so DMARC has nothing to align with. The fix belongs at the platform: sign with a DKIM key on your own domain, use a return-path under your own domain, and publish the DNS records it hands you. The Google FAQ explains that only a single protocol has to align for direct mail, although it recommends aligning both, and that forwarded or mailing-list messages are exempt. The helper above treats any parent or child name as aligned, which suits ordinary registered domains. A production check needs the public suffix list.
SPF Failures and the Lookup Limit
RFC 7208 defines the results you will see. A softfail is a weak statement that the host is probably not authorized, which is what a ~all ending produces. A temperror is a transient DNS problem, and a retry may succeed. A permerror, in section 2.6.7, means the published records could not be interpreted and definitely needs the DNS operator to act. Three faults produce it: two or more SPF records on a name (section 3.2), more than ten lookup-causing terms in a single evaluation, and more than two void lookups, which are lookups that return nothing (both in section 4.6.4).

Every marketing platform, CRM and help desk that sends in your name tends to add an include, so the count grows without anyone deciding to grow it. Fix an over-limit record by pruning: remove services you no longer use, and publish a single corrected record in place of the old ones. A free SPF checker reads the live record, walks every nested include and counts the lookups against the limit of ten. Gmail's 4.7.24 and 5.7.24 replies report an SPF record with one or more suspicious entries, which makes a stray or mistyped include the first thing to look for.
DKIM Failures
RFC 6376 describes the two ways a signature fails. In section 6.1.2, the verifier looks up the public key from the d= and s= tags, and if the key record does not exist it reports a permanent failure, no key for signature. In section 6.1.3, it recomputes the body hash and compares it with the bh= tag, and a mismatch returns body hash did not verify. The first usually means a selector was never published or was published under the wrong name. The second means something changed the message after it was signed, such as a footer added later.
Two Google rules sit on top of the standard. Delivery to Gmail inboxes calls for a DKIM key of 1024 bits or longer, and Google recommends 2048 bits when your DNS host supports it. The Postmaster dashboards page adds that some dashboards cover only DKIM-authenticated messages, so email signed only with a platform's domain may not appear under yours.
DMARC Rejections
A DMARC rejection arrives as Microsoft's 5.7.509, as the Gmail 5.7.26 text that cites the sender's published DMARC rule, or as the Postmaster error named DMARC policy of the sender domain. For that last error, Google advises reading your daily DMARC reports to learn whether the message was genuine or a spoof, and using a policy of none when you only need to satisfy the guideline. The rejection means the domain enforces reject, and some real source of your email is not passing. The aggregate reports name that source. Microsoft's announcement recommends moving gradually from none to quarantine to reject once your legitimate sources are aligned.
Bounces Climbing: Check the List First
Hard bounces are permanent failures, and Amazon's FAQ gives the plainest example: an address that does not exist. Its bounce rate counts only hard bounces to domains you have not verified. Temporary failures such as a full mailbox, and bounces caused by a blocked IP address, stay out of the figure. Because the rate leaves those cases out, compare thresholds only after you know what each tool counts. Amazon explains that mailbox providers and anti-spam organizations use high bounce rates to spot bad sending practices, and that high rates can send email to the spam folder.

Screenshot: the Amazon SES FAQ answer with the bounce rates for review and for a sending pause, captured in November 2026.
Bad addresses reach your audience file in ordinary ways: a typo at signup, an old import, a purchased file, a contact who left a company. Spamhaus names sloppy list upkeep among the triggers for a Combined Spam Sources listing, next to unsolicited email and compromised accounts. Google wants each recipient's address confirmed before the subscription starts, and Yahoo advises prompt removal of invalid recipients plus a policy for dropping addresses that return 5xx errors. Group your hard bounces by source and import date. When a single source produces most of them, the fix is to stop using that source, not to slow your sending.
Verification removes the addresses that would bounce before they cost you reputation. For a single address you can run a quick check on an email address with no signup: it runs syntax, DNS, MX and live SMTP tests, and it reports disposable, role and catch-all results separately. For files and automation, billionverify.com applies the same layers in bulk or through a REST API. Catch-all domains come back as their own category instead of valid. Valid and invalid results cost 1 credit each, unknown results cost nothing, and a free account gets 20 credits on each day it is used, up to 600 a month, with no credit card.
A verifier cannot make people want your email, so it sits beside consent and not in place of it. Our guide to cleaning an email list turns the steps into a sequence, and the article on catch-all email detection explains the result no check can settle. When a hard bounce comes from an address you do not recognize, a reverse email lookup reads the domain, organization and routing behind it. If the same form collects a phone number, a phone number validator checks it against its country and normalizes it to E.164.
When Only One Mailbox Provider Pushes Back
A problem at one provider and not at the others means that provider's rules or data have flagged you. Each large provider runs a free dashboard on email deliverability for senders, and each shows a different slice of the story. None of them gives the full picture, so use the dashboards to confirm the cause you suspect, not to discover a cause from nothing.
Gmail and Google Postmaster Tools
The Delivery errors dashboard in Google Postmaster Tools charts the portion of email that passed at least one sender test and was then rejected or temporarily failed. It sorts the errors into named causes: rate limit exceeded, suspected spam, spammy content such as Internet links, a bad attachment, the DMARC rule of the sending domain, a poor reputation for the IP address or the domain, a listing on a public blocklist, and a bad or missing PTR record. Google's page gives a recommended action for each. The data is not real time, it normally refreshes inside a day, and it may be missing on days with low volume to protect Gmail users' privacy.
Google calls temporary failures a throttling tool and recommends exponential backoff: retry a failed request after longer and longer delays. Without slower sending after temp fails, the page predicts a period of permanent failures. That is the same sequence as the Gmail table above, where a 4.7.x code comes first and a 5.7.x block follows.
Outlook.com and SNDS
Microsoft's Smart Network Data Services portal shows data about the IP addresses you control, and its Junk Mail Reporting Program sends a report when users junk your messages. You sign in with a Microsoft account and request access to each address. The portal puts it plainly: deliverability to Outlook.com is based on your reputation, reputation is always the responsibility of the sender, and the data should help you keep mailing lists clean. The Outlook.com Postmaster site adds that since May 5, 2025, domains sending over 5,000 emails a day must meet stricter authentication standards, that non-compliant messages go to the junk folder, and that rejection will follow until the DNS records are corrected.
Yahoo and Spam Complaint Reports
Yahoo's Complaint Feedback Loop sends you a copy of each spam complaint in ARF format, so you can suppress that person from later campaigns. It supports only DKIM-signed email and works per domain, which is another reason to sign with your own domain. Yahoo's best practices page asks senders to keep the spam rate below 0.3% and notes that Yahoo calculates it on what reached the inbox, so a rate computed on everything you sent will not match Yahoo's figure.
Blocklist Listings
A blocklist listing can explain a sudden wall of 5.7.x rejections across many providers at once. Yahoo's error page names a Spamhaus listing among the reasons for a permanent rejection, and Postmaster Tools has separate errors for a listed address and a listed domain. Know which blocklist flagged you, because they mean different things. Spamhaus's ZEN blocklist page describes ZEN as the SBL, CSS, XBL and PBL combined, and a separate Domain Blocklist covers domain names.
| List | What it flags | First thing to check |
|---|---|---|
| PBL | End-user IP ranges that should never send email directly | Whether you send from a home or office connection instead of a mail server |
| XBL | Individual IPs showing signs of compromise, such as malware or relaying with stolen credentials | Infected devices, leaked mail credentials, software that proxies traffic |
| CSS | Low-reputation SMTP traffic, including poor list hygiene and compromised accounts | The list source and any recent import |
| SBL | Malicious network ranges | Your network provider and the range you send from |
| DBL | Domains with a poor reputation | The domains in your links, your From address and your HELO name |
Spamhaus explains that addresses on the Policy Blocklist are not necessarily bad, only addresses that should not originate email, and that end users can request an exclusion. Its Exploits Blocklist entries expire on their own once the malicious behavior stops. The ZEN page directs every listed sender to the Spamhaus reputation checker, which names the list and the steps to remediate it. Its best practices are to restrict outbound port 25 to your mail server, use double opt-in, and make the hostname, the HELO name and the reverse DNS entry agree.
Fix the cause before you ask for removal, because a compromised account, an open relay, a bad import or a volume spike will list the address again. Microsoft's NDR page routes a 5.7.511 ban to a delist address and to the Office 365 delist portal, which has you submit the IP and confirm an email address. A 5.7.513 report is different: the recipient's company blocked you, and only that company can lift it. An email blacklist check from our site queries four public lists for an IP or a domain: SpamCop, Barracuda, PSBL and UCEPROTECT Level 1. It leaves Spamhaus out, so use Spamhaus's own checker for that list.
Spam Folder Placement With Clean Authentication
A message that is accepted and then filed as junk is the hardest case, because the server gives you no code. The Postmaster page defines the spam rate as the share of messages delivered to the inbox that recipients then mark as spam. That definition hides a trap. If Gmail already sends much of your email to spam, fewer people see it in the inbox, and the dashboard figure can look low or even fall to zero. Google lists this under its own troubleshooting items: a spam rate that seems very low, and a high figure that suddenly changed to 0.
A low reading is therefore not proof of health. Read it next to the delivery errors and a seed mailbox you can inspect, and compare it with replies. A high rate, in turn, has a direct source. The guidelines name the practices to avoid: purchased addresses, email to people who never signed up, opt-in boxes checked by default and promotions mixed into receipts. They also warn against deceptive display names, subject lines that start with Re: or Fwd: when the message is not a reply, and HTML or CSS that hides content.
Unsubscribe handling belongs in this review. Bulk senders to Gmail must support one-click unsubscribe on marketing email, and the FAQ adds that a link broken for a long time will not trigger spam marking by itself but costs the sender any help with delivery problems. To see what a mailbox provider would make of a real email, the free email deliverability test sends a message to a private mailbox and scores authentication, DNS, spam filters, blacklists, headers, content and placement. It needs a free account, and it describes only the message you send, so use your real template and sending platform. Our email deliverability guide covers the written rules behind these checks.
Deferrals After a Volume Jump or a New Domain
Deferrals that begin after a launch, a platform switch or a first big import usually come from volume, not from a broken record. Google's Top 10 Gmail sender issues page warns that raising volume too quickly causes delivery problems, and it gives a recovery routine. Send nothing for 15 minutes, then send a single message and check it arrives. If it does, stay below the volume where deferrals began for 24 hours. After that, raise volume slowly, with a common daily increase of 25% to 100% depending on various factors.
The same page recommends even pacing. Its example is that sending 60 emails as fast as possible, waiting a minute and repeating is a poor pattern, while a steady message every second is better. For a 4.7.28 error, the sender guidelines advise sending nothing for at least 10 minutes, then using a single connection and adding connections one at a time. Yahoo names the same cause for its unsolicited-volume errors: an IP set up recently without a slow ramp. A brand new domain has no history, so it deserves the slowest ramp of all.
Checking PTR, HELO and TLS
Infrastructure faults look like deferrals and blocks together. Google requires the public IP of a sending SMTP server to have a PTR record that resolves to a hostname, and that hostname to have an A or AAAA record resolving back to the same IP. It points to the Google Admin Toolbox Dig tool for the check. Spamhaus asks the same of the HELO name, and Microsoft's 5.7.506 reply flags a HELO that names the wrong server. If you send through a hosted platform, it owns these records, and you only check them when you run your own server.
The shell function below performs the forward-confirmed reverse DNS check. The first two addresses are Google's mail server and Google's public DNS, which match. The third is a documentation address with no PTR record.
check_ptr() {
ip="$1"
name=$(dig +short -x "$ip" | head -n 1)
if [ -z "$name" ]; then echo "$ip: no PTR record"; return; fi
if { dig +short "$name" A; dig +short "$name" AAAA; } | grep -qx "$ip"; then
echo "$ip: $name resolves back to the same address"
else
echo "$ip: $name does not resolve back to $ip"
fi
}
for ip in 209.85.220.41 8.8.8.8 192.0.2.10; do check_ptr "$ip"; done
# 209.85.220.41: mail-sor-f41.google.com. resolves back to the same address
# 8.8.8.8: dns.google. resolves back to the same address
# 192.0.2.10: no PTR record
A Repeatable Order for Any Drop
When the cause of your email deliverability issues is not obvious, work from the narrowest explanation to the widest and change one variable at a time. Google's own troubleshooter for senders asks four questions: was your email server compromised, was the email unauthenticated, were your subscription practices incomplete, and was an affiliate the cause. The table turns that into a working order and adds the evidence you can collect at each step.
| Step | Check | Where to look |
|---|---|---|
| 1 | Group replies by provider and cause | Your bounce log and the script above |
| 2 | Read a real message's headers | Authentication-Results, from Show original |
| 3 | Confirm the DNS behind each verdict | SPF lookup count, DKIM selector, DMARC record |
| 4 | Check the provider dashboards for the dates of the drop | Postmaster Tools, SNDS, the Yahoo feedback loop |
| 5 | Look for a listing on a blocklist | Spamhaus checker, public DNSBL lookups |
| 6 | List what changed: an import, a new tool, a volume jump, a DNS edit | Your change log |
After a fix, wait before you judge it. Postmaster Tools reports that data typically updates within 24 hours, and compliance changes can take up to 7 days to show. Google adds that once the number of spam reports drops, Gmail may need time to adjust its spam classification of your email. If you were over the 0.3% line, the seven consecutive days under it that the FAQ requires is the earliest point at which delivery help returns. No provider promises a recovery date, so a slow recovery does not by itself mean the fix failed.
Keep Email Deliverability Issues From Returning
Most deliverability faults return because nobody looked after the change that caused them. Add a check to each step that can break authentication: count SPF lookups whenever a new sending service joins, test the DKIM selector after any key rotation, and read the DMARC aggregate reports for sources you do not recognize. Review the Postmaster compliance status and your Yahoo and Microsoft feedback reports on a regular schedule, and keep a change log that records imports, template changes, new tools and volume jumps beside the date.
Keep the address file clean at the source. Confirm new subscribers, verify imported files before the first send, remove every contact that returns a permanent error and stop writing to contacts who never engage. Every provider quoted in this guide gives some version of the same advice: write only to people who asked for your email, and make leaving easy. Most of the codes above are the receiving servers' way of reporting that this rule was missed.
FAQs
Why do my emails go to spam when every authentication test passes?
Passing authentication proves who sent a message, not that anyone wanted it. A domain with a high complaint rate, a purchased list or a sudden volume jump can pass every test and still be filtered. Look at the spam rate, the list source and recent sending changes before you touch DNS again.
Is a deferral the same as a bounce?
No. A deferral is a 4xx reply that asks the sender to try again, and a bounce is a permanent failure that the sender should not repeat unchanged. Platforms usually retry deferrals for days, so the message may still arrive. A run of deferrals is an early warning that a block may follow.
Should I remove an address after a 5.7.x error?
Not on the first. A 5.7.x reply usually describes your sender identity, reputation or a block on your IP address, so removing the recipient leaves the cause in place. Repair the cause first, then retry. Drop an address on a 5.1.x reply, or when it keeps returning permanent errors from several providers.
Do I need Google Postmaster Tools?
If you send to Gmail addresses, yes, because it is the only place Google reports your spam rate, compliance status and delivery errors. It covers only email sent to Gmail users, and it may show no data when your volume is too low. A small sender will often have to lean on bounce text and seed tests instead.
How long does it take to remove a blocklist listing?
It depends on the list. Spamhaus lets XBL entries expire after a period once the behavior stops, and it routes other listings through its reputation checker, with remediation steps. A removal request does not help while the cause persists, so fix the compromised account, open relay or bad import first.
What spam rate is too high?
Google asks senders to stay below 0.1% and to avoid ever reaching 0.3%, and adds that rates above 0.1% already reduce inbox delivery for bulk senders. Yahoo draws the line at 0.3% and measures it against inbox placements. Amazon starts a review at a complaint level of 0.1%. The numbers differ because each provider counts a different population.
