The deliverability rules have not become stricter since February 2024. The consequences of ignoring them have. The volume threshold, the complaint rate ceilings and the unsubscribe deadline are all exactly where they were. What changed is that in November 2025 the largest mailbox provider began applying real rejections: “Gmail is ramping up its enforcement on non-compliant traffic. Messages that fail to meet the email sender requirements will experience disruptions, including temporary and permanent rejections.”

That distinction matters for a B2B newsletter, because it means the work is known and finite rather than a moving target. It also means a list that was quietly degrading for two years may now be visibly failing.

Three of the rules commonly quoted alongside the real ones turn out to have no primary source at all, and they are named at the end.

The thresholds, and the one that never expires

The published requirements split into two tiers, and the boundary between them is stickier than most people realise.

The tier boundary. “A bulk sender is any email sender that sends close to 5,000 messages or more to personal Gmail accounts within a 24-hour period. Messages sent from the same primary domain count toward the 5,000 limit.”

And the part nobody plans for. “Senders who meet the above criteria at least once are permanently considered bulk senders … Bulk sender status doesn’t have an expiration date … Changes in email sending practices will not affect permanent bulk sender status once it’s assigned.”

Which has a practical consequence. One large announcement to a merged list, sent once, moves your domain into the stricter tier for good. Subdomains aggregate to the primary domain, so splitting sends does not undo it.

What every sender owes, at any volume. SPF or DKIM, valid forward and reverse DNS, TLS, a spam rate below the threshold, and message formatting per the internet message format standard.

What the bulk tier adds. SPF and DKIM, a published DMARC record, alignment between the visible sender domain and one of the authenticated domains, and one-click unsubscribe on marketing mail.

With a detail that saves an argument. The minimum acceptable policy is the weakest one: “Your DMARC enforcement policy can be set to none.” Publishing p=none satisfies the requirement. It is widely misreported as requiring enforcement.

The two tiers of mailbox provider sender requirements and the permanence of bulk classificationThe two tiers of sender requirements published by the largest mailbox provider, both effective from the first of February 2024, together with the permanence of classification into the higher tier. The lower tier applies to all email senders regardless of volume and requires authentication by either the sender policy framework or domain keys identified mail, a valid forward and reverse domain name system record for sending addresses, use of a transport layer security connection for transmitting email, maintenance of spam complaint rates below the published threshold, formatting of messages according to the internet message format standard, and avoidance of impersonation in the visible sender header. The upper tier applies to bulk senders and adds four further requirements, namely authentication by both the sender policy framework and domain keys identified mail rather than either, publication of a domain based message authentication reporting and conformance record, alignment between the organisational domain in the visible sender header and either the authenticated sender policy framework domain or the authenticated domain keys identified mail domain, and support for one click unsubscribe on marketing and subscribed messages. The boundary between tiers is defined as sending close to five thousand messages or more to personal accounts within a twenty four hour period, with messages from the same primary domain counting toward that limit, meaning subdomains aggregate rather than being counted separately. The classification is permanent, the documentation stating that senders meeting the criteria at least once are permanently considered bulk senders, that bulk sender status does not have an expiration date, and that changes in email sending practices will not affect permanent bulk sender status once assigned. The practical consequence is that a single large announcement sent once to a merged list moves the domain into the stricter tier permanently and cannot be undone by subsequently splitting sends across subdomains. A detail frequently misreported is that the minimum acceptable authentication policy is the weakest available, the documentation stating that the enforcement policy can be set to none.Two tiers, and one door that only opens one wayEvery sender, any volumeSPF or DKIMValid forward and reverse DNSTLS for transmissionSpam rate under the thresholdFormatted per RFC 5322Bulk senders addSPF and DKIM, not eitherA published DMARC recordAlignment with SPF or DKIMOne-click unsubscribe on marketingMinimum policy accepted: p=none.The boundary is a one-way door”Senders who meet the above criteria at least once are permanently considered bulk senders.”Which means one big send changes your rules forever”Close to 5,000 messages or more” in 24 hours, aggregated across subdomains. Splitting sends afterwards does not undo it.
Two tiers, one boundary, and a classification that never expires once you have crossed it. Source : Google Workspace Admin Help, Email sender guidelines and the sender requirements FAQ (2026)

Complaint rate: two numbers, not one

This is the metric that decides whether your newsletter arrives, and the guidance contains two figures that get collapsed into one.

The sentence that carries both. “Keep spam rates reported in Postmaster Tools below 0.10% and avoid ever reaching a spam rate of 0.30% or higher.”

What each level costs. “The user-reported spam rate’s impact on delivery is graduated, and rates of 0.3% or higher have an even greater negative impact on email inbox delivery. Even today, user-reported spam rates greater than 0.1% have a negative impact on email inbox delivery for bulk senders.”

And the recovery rule, which is the most operationally useful line published. “Beginning June 2024, bulk senders with a user-reported spam rate greater than 0.3% will be ineligible for mitigation … Bulk senders will be eligible for mitigation when their spam rates remain below 0.3% for 7 consecutive days.”

One measurement subtlety, from the other major provider. “Spam rate is calculated in our system based on mail delivered to the inbox.” The denominator is delivered mail, not messages sent, which means a poorly delivered send can post a worse complaint rate than a well delivered one with the same number of complaints.

Put in absolute terms for a B2B list. At 5,000 recipients, 0.30 percent is fifteen complaints. Fifteen people clicking the spam button on one send is enough to lose eligibility for support.

Which reframes the list-building question entirely. The cost of a badly acquired subscriber is not a wasted send. It is a shared reputation that every other recipient pays for.

The two published spam complaint thresholds and the consequences and recovery attached to eachThe two distinct spam complaint rate thresholds published by the largest mailbox provider, the consequence attached to each, and the mechanism by which eligibility is restored after breaching the higher one. The published guidance contains both figures in a single sentence, instructing senders to keep spam rates reported in the provider’s postmaster tools below zero point one zero percent and to avoid ever reaching a spam rate of zero point three zero percent or higher. The impact is described as graduated rather than binary, with the documentation stating that rates of zero point three percent or higher have an even greater negative impact on inbox delivery, and that even today user reported spam rates greater than zero point one percent have a negative impact on inbox delivery for bulk senders. From June 2024, bulk senders with a user reported spam rate greater than zero point three percent became ineligible for delivery mitigation, meaning the provider will not intervene to assist with delivery problems while the rate remains above that level. Eligibility for mitigation is restored when spam rates remain below zero point three percent for seven consecutive days, and the rate is calculated daily. A measurement subtlety published by the second largest provider is that its spam rate is calculated based on mail delivered to the inbox rather than on messages sent, meaning the denominator is delivered mail, so a send with poor inbox placement can post a worse complaint rate than a well delivered send generating the same absolute number of complaints. Expressed in absolute terms for a business list of five thousand recipients, the zero point three zero percent line corresponds to fifteen complaints, so fifteen recipients marking a single send as spam is sufficient to lose eligibility for delivery support. The consequence for list building practice is that the cost of a poorly acquired subscriber is not a single wasted message but a shared sending reputation which every other recipient of that domain’s mail subsequently pays for.Two numbers that get collapsed into oneThe target0.10%Above this already “has a negative impacton email inbox delivery for bulk senders.”The line0.30%“Avoid ever reaching” it. Above it, deliverysupport and mitigations become unavailable.How you get back”Eligible for mitigation when their spam rates remain below 0.3% for 7 consecutive days.” Calculated daily.In absolute termsOn a 5,000-person list, 0.30% is fifteencomplaints on a single send.And the denominator surpriseOne provider calculates it “based on maildelivered to the inbox”, not on mail sent.A badly acquired subscriber does not waste a send. They spend a reputation every other recipient shares.
0.10 percent is the target, 0.30 percent is the line. Seven consecutive days below it restores eligibility. Source : Google Workspace Admin Help, sender requirements FAQ (2026)

One-click unsubscribe, and why most implementations are inert

This requirement is widely implemented and frequently implemented in a way that does not work.

What is required. Two headers, exactly: List-Unsubscribe-Post: List-Unsubscribe=One-Click and a List-Unsubscribe header containing an HTTPS URI.

What does not count. “Other types of one-click unsubscribe, such as mailto and URL unsubscribe links, don’t meet our one-click unsubscribe requirement.” And: “One-click unsubscribe links that link to a landing or other type of web page don’t comply with RFC 8058.”

The deadline. “Process and honor unsubscribe requests within 48 hours.” The other major provider phrases it as two days. Build to the shorter.

And the requirement almost nobody implements correctly. The standard is explicit: “senders MUST apply at least one valid DKIM signature to the message … The List-Unsubscribe and List-Unsubscribe-Post headers MUST be covered by the signature.” Without that, “the mail receiver SHOULD NOT offer a one-click unsubscribe for that message.”

Which means an unsigned header does nothing at all. The unsubscribe button simply does not appear. This is invisible from the sending side, and it is the most common reason a correctly configured header produces no effect.

Why the standard exists at all is worth knowing. “Anti-spam software often fetches all resources in mail header fields automatically, without any action by the user … To prevent accidental unsubscriptions, senders return landing pages with a confirmation step.” One-click exists to remove that confirmation step safely, using a POST that scanners will not issue.

With the recipient-side logic stated in the standard itself. “If an unsubscription process is too difficult, the recipient’s alternative is to report mail from the sender as junk.” A hard unsubscribe converts directly into the complaint rate that decides your delivery.

One scope limit. It applies to marketing and subscribed messages only: “Transactional messages are excluded from this requirement.”

The three providers do not agree

Building to one set of rules is not the same as building to all of them, and the differences are specific.

On volume, one publishes a number and one refuses. “A ‘bulk’ sender is classified as an email sender sending a significant volume of mail. We will not specify a volume threshold.” Any numeric threshold attributed to that provider did not come from it.

On mailto, they disagree outright. One states that mailto links do not meet the requirement. The other says “the mail-to: method is acceptable.”

On scope, the third is narrower than both. Microsoft’s requirements for high-volume senders cover authentication only: SPF must pass, DKIM must pass, and DMARC with “at least p=none and align with either SPF or DKIM.” It publishes no spam-rate threshold, no unsubscribe requirement and no TLS requirement, listing unsubscribe links and list hygiene as recommendations rather than requirements.

And its threshold wording differs from its own blog. The support page defines a high-volume sender as one who sends “5,000 or more email messages to Microsoft consumer email services” with no time window stated, while the announcement said “more than 5,000 emails per day”. Both texts are live.

What that produces in practice. Build to the strictest reading of each rule and you satisfy all three: both authentication methods, alignment, HTTPS one-click with a DKIM-covered header, a 48-hour unsubscribe, and complaint rates under 0.10 percent.

Comparison of published sender requirements across three major mailbox providersComparison of the sender requirements published by three major mailbox providers, showing where they agree and where they materially differ. On volume thresholds, the first provider publishes a figure of approximately five thousand messages or more to personal accounts within twenty four hours and treats classification as permanent, the second explicitly declines to publish any figure stating that it will not specify a volume threshold and instead assesses senders at the authenticated domain or visible sender domain level using all available information, and the third states five thousand or more messages in its support documentation without a stated time window while its announcement blog states more than five thousand per day, leaving both texts live and inconsistent. On authentication, all three require both the sender policy framework and domain keys identified mail for high volume senders, together with a published reporting and conformance record whose minimum acceptable policy is none, and alignment between the visible sender domain and at least one authenticated domain. On spam complaint thresholds, the first and second publish a figure of zero point three percent while the third publishes none. On unsubscribe requirements, the first requires a one click mechanism using specified headers and rejects both mailto links and links to landing pages, the second requires a functioning list unsubscribe header while stating that the mailto method is acceptable, and the third lists functional unsubscribe links under additional recommendations rather than requirements. On transport layer security, the first requires it for all senders at any volume, while neither the second nor the third states any such requirement in its published requirement lists. On message formatting, the first cites the internet message format standard, the second cites both that standard and the simple mail transfer protocol standard, and the third states no formatting requirement. The practical resolution is to build to the strictest reading of each individual rule, which satisfies all three simultaneously.Three providers, three rulebooksRequirementGmailYahooOutlook.comA published volume threshold~5,000/dayRefuses to state one5,000, window unclearSPF and DKIM and DMARC at volumeYesYesYesA published spam rate threshold0.3%0.3%NoneOne-click unsubscribeRequired, no mailtoRequired, mailto okRecommendedTLS requiredYes, all sendersNot statedNot statedThe resolution is simple, if unglamorousBuild to the strictest reading of each rule and you satisfy all three at once: both authentication methods, alignment,HTTPS one-click with a DKIM-covered header, 48 hours, and complaints under 0.10 percent.
One publishes a threshold, one refuses to. One requires unsubscribe headers, one only recommends them. Source : Google Workspace Admin Help, Yahoo Sender Hub, and Microsoft support and blog (2026)

Three rules with no primary source

Alongside the documented requirements sit three widely repeated rules that do not come from anyone.

There is no published bounce rate ceiling. Searching the sender documentation of all three major providers returns guidance and no threshold. Gmail says to “automatically unsubscribe recipients who have multiple bounced messages” and to reduce volume if messages start bouncing. Yahoo says to “remove invalid recipients from your list promptly.” Microsoft lists bounce management as a recommendation. None states a number.

And the number people cite is not in the document they cite it from. A five percent hard bounce threshold is widely attributed to an industry working group’s best practices document. That figure does not appear in it.

What that document does contain is a usable rule. Remove an address “if it bounces consecutively at least two times over two weeks or more. This should account for any server issues on the receiving side that could have caused an erroneous bounce.” Worth noting the document is from 2015.

There is no plain-text-part requirement. Gmail’s formatting requirements cite the internet message format standard and say that HTML messages should follow HTML standards. No published requirement obliges a multipart alternative text version.

And there is no published Yahoo volume threshold, because Yahoo says explicitly that it will not publish one. Every number attributed to it is somebody’s inference.

None of this makes the underlying advice bad. Suppressing repeated bounces, sending a text part and keeping volumes modest are all sensible. They are just not requirements, and presenting them as such makes the real requirements harder to take seriously.

Three widely repeated deliverability rules that no mailbox provider publishesThree rules widely presented as email deliverability requirements which no mailbox provider actually publishes, established by full text search of the published sender requirements of the three major providers. The first is a maximum acceptable bounce rate. No provider publishes a numeric threshold. The largest provider instructs senders to automatically unsubscribe recipients who have multiple bounced messages and to reduce sending volume if messages begin bouncing or being deferred, then increase slowly again. The second largest instructs senders to monitor hard and soft bounces and inactive recipients, to reduce invalid recipients through confirmed opt in, and to remove invalid recipients promptly. The third lists bounce management among additional recommendations rather than requirements. None of the three states a percentage. Furthermore, a five percent hard bounce threshold widely attributed to an industry working group’s sender best practices document does not appear in that document, whose text contains no such figure. What the document does contain is an operational rule, namely that an address should be removed if it bounces consecutively at least twice over two weeks or more, allowing for receiving side server issues that could produce an erroneous bounce, and that document dates from February 2015. The second unpublished rule is a requirement to include a plain text alternative part alongside hypertext markup language content. The largest provider’s formatting requirements cite the internet message format standard and state that hypertext markup language messages should follow hypertext markup language standards, without any requirement for a multipart alternative text version. The third unpublished rule is a volume threshold at the second largest provider, which states explicitly that it will not specify a volume threshold, meaning any numeric figure attributed to it originates elsewhere. None of these three pieces of advice is unsound, but presenting sensible practice as a published requirement makes the genuine requirements harder to take seriously.Three rules nobody actually published”Keep bounces under 5 percent”No provider publishes any bounce threshold. And the figure does not appear in the industry document it isusually attributed to.”Always include a plain text part”Not in any published requirement. The formatting rules cite RFC 5322 and say HTML should follow HTML standards.”Yahoo’s threshold is X messages""We will not specify a volume threshold.” Every number attributed to it came from somewhere else.What the industry document does say, and it is usableRemove an address “if it bounces consecutively at least two times over two weeks or more.” Published in 2015.
Sensible advice, presented as requirements. One of them cites a document that does not contain the figure. Source : Full-text search of the published sender requirements of three providers, and the 2015 industry best practices document (2026)

What changed underneath, in May 2026

One thing did move, and it moved at the specification level rather than at the provider level.

The DMARC specification was replaced. The 2015 document was obsoleted in May 2026 by a successor that also replaces a 2021 update. Citing the old number as the current DMARC specification is now wrong.

And the status changed with it. The old document was Informational. The new one is Standards Track.

Three substantive changes are worth knowing. The policy tag went from required to recommended, a record without a policy tag now defaults to none, and the sampling tag was removed entirely and marked historic.

The semantics shifted too. The old text told receivers what to do. The new one describes what the domain owner considers: none means “the Domain Owner offers no expression of preference”, reject means the owner “considers all such failures to be a clear indication that the use of the domain name is not valid.”

With the receiver’s discretion made explicit. “The final handling of any message is always a matter of local policy and is left to the discretion of the Mail Receiver”, and receivers “SHOULD NOT reject messages solely because of a published policy of ‘reject’.”

Which is worth understanding before you tighten your own policy. Moving to reject does not compel anybody to reject. It expresses a position, and it can break forwarded mail and mailing lists while doing so.

Changes introduced by the May 2026 replacement of the email authentication reporting specificationChanges introduced when the specification governing domain based message authentication, reporting and conformance was replaced in May 2026. The original specification, published in March 2015 with informational status, was obsoleted along with a 2021 update by a successor published in May 2026 with standards track status, meaning citing the former document number as the current specification is now incorrect. Three substantive changes affect implementation. First, the policy tag changed from required to recommended, so a record lacking it remains syntactically valid. Second, a record without a policy tag is now treated as though it specified a policy of none, whereas previously the absence of a required tag rendered the record invalid. Third, the sampling tag which allowed a domain owner to apply the policy to only a percentage of failing messages was removed entirely and is now marked historic in the tag registry, so any record still publishing it is using a deprecated element. The semantics of the policy values also shifted from instruction to assessment. The former text framed each value as a request to the receiver, with none requesting no specific action, quarantine requesting that failing mail be treated as suspicious, and reject requesting that failing mail be rejected during the transaction. The current text frames each value as a statement of the domain owner’s view, with none meaning the owner offers no expression of preference, quarantine meaning the owner considers such mail suspicious while acknowledging it may be valid, and reject meaning the owner considers all such failures a clear indication that the use of the domain name is not valid. The receiver’s discretion is made explicit, the specification stating that final handling of any message is always a matter of local policy left to the discretion of the mail receiver, and that receivers should not reject messages solely because of a published policy of reject.The specification itself movedThe 2015 document is obsoleteReplaced in May 2026, along with a 2021 update. Status also changed: Informational to Standards Track.Policy tagRequired, then recommended.A record without it now defaultsto none.Sampling tagRemoved entirely and markedhistoric in the registry.SemanticsFrom telling the receiver whatto do, to describing what theowner considers.And the receiver’s discretion, stated outrightReceivers “SHOULD NOT reject messages solely because of a published policy of ‘reject’.”Which is worth knowing before you tighten yours. Moving to reject compels nobody, and it can break forwarded mailand mailing lists while expressing a position.
A new document, a status upgrade, and a shift from instructing the receiver to describing the owner's view. Source : RFC 9989, May 2026, obsoleting RFC 7489 and RFC 9091 (2026)

What a newsletter actually costs

The recurring cost is not the sending platform. It is four pieces of maintenance that nobody budgets.

Authentication upkeep. SPF, DKIM and DMARC records that stay correct as tools change. Every new sending tool added to the stack is a chance to break alignment, and a broken alignment is a silent failure.

Complaint monitoring. Somebody has to look at the postmaster data after each send and act on a rising rate before it crosses the line, given that recovery takes seven consecutive days.

Suppression handling within the window. Forty-eight hours is not a batch job that runs weekly. It is an automated path from the header to the list.

And list decay. Business addresses go stale faster than personal ones, because people change employers. A list you have not cleaned in two years is not the list you think you are sending to.

One thing that is not a cost. Open rate reporting, which has not measured opens reliably since privacy protections began pre-fetching images. Building a newsletter programme around it means optimising a number that is partly generated by machines.

Which leaves the metrics worth keeping. Clicks to a named destination, replies, unsubscribes, complaint rate, and whatever downstream event you actually care about. All four survive image blocking, and the last one is the only one that pays for the programme.

What to do with this

Check your alignment first, because it is the single point of failure that produces no visible error. Send yourself a message from each tool that mails on your behalf, look at the headers, and confirm the visible sender domain matches an authenticated one.

Then verify that your one-click header is covered by your DKIM signature. If it is not, the button does not appear, and no amount of correct header syntax fixes that.

Put the complaint rate somewhere a person sees it weekly, with 0.10 percent as the line rather than 0.30 percent. Recovery from the higher figure takes a week of clean sending, which is a week of a newsletter you cannot send.

And stop reporting open rate as a result. Report clicks, replies and complaints, and let the open rate sit in the platform where it belongs. Wiring those clicks through to an event that pays is the same measurement problem as on bought traffic, where we set the tracking up before any budget is released and report on cost per lead rather than on impressions.

The related pieces are cold email and what US law allows and consent on lead capture forms.