Email and SMS marketing in 2026: deliverability, automation, segmentation and benchmarks

Email and SMS still return more revenue per pound spent than any other channel most brands run, but the rules changed hard in 2026. Mailbox providers now reject non compliant bulk mail outright, an AI layer sits between your subject line and your reader, and two European regulators have turned the open pixel into a consent decision. Deliverability and consent are the work now.

What actually changed in 2026

Three things, and they pull in different directions.

The first is enforcement. The Gmail and Yahoo sender requirements landed in February 2024 as guidance with soft consequences. Google’s own sender guidelines FAQ now describes temporary and permanent rejections for non compliant traffic, ramping from November 2025. Microsoft applied a similar bar to outlook.com, hotmail.com and live.com for senders above 5,000 messages a day, as dmarcian documented. Missing authentication is no longer a spam folder problem. It is a bounce.

The second is the AI layer. On 8 January 2026 Google put Gemini into Gmail for its three billion users: AI Overviews that summarise threads, Help Me Write free to everyone, suggested replies, and an AI Inbox that decides which messages surface first. Your campaign is now often read by a model before a human sees it. MarTech has argued this redefines what deliverability means, and that feels right. Reaching the inbox and being surfaced in it are separate problems now.

The third is consent creep into measurement itself. Two national regulators decided that the pixel you use to count opens is access to a device, not a neutral analytics artefact. More on that below, because it is the change most teams have not budgeted for.

Deliverability and the sender rules that now bite

Start here. A brilliant segmentation strategy delivered to a spam folder earns nothing.

The technical baseline is not complicated, and it is not optional. Publish an SPF record that covers every service sending as your domain. Sign with DKIM using a 2048 bit key. Publish DMARC, aligned to the domain in your visible From address. Google’s guidance is explicit that alignment, not just the presence of a record, is what it checks. Send over TLS, keep valid forward and reverse DNS on your sending IPs, and format messages to RFC 5322.

Then the behavioural half. Marketing and promotional mail needs RFC 8058 one click unsubscribe headers, and the request has to be honoured inside 48 hours. Transactional mail is exempt from that requirement. Spam complaints are the tightest constraint: Google wants you under 0.10% and treats 0.30% as the point where it stops helping you. One bad reactivation campaign to a two year old segment can push a domain over that line in a single afternoon.

Sources: Google sender guidelines FAQ, MarTech on the Google, Yahoo and Microsoft rules, dmarcian.

Why the open rate is finished as a metric

Apple Mail accounted for 62.26% of opens Litmus tracked in July 2026, against 27.03% for Gmail, from a sample of over a billion opens. Mail Privacy Protection pre fetches images for a large share of those, so the open you recorded may have been a proxy server in a data centre and not a person. Litmus estimates the distortion touches 55% to 60% of all opens.

That was already a measurement problem. In 2026 it became a legal one.

France’s CNIL adopted a recommendation on email tracking pixels on 12 March 2026 and published it on 14 April 2026. Pixels generally require prior consent under Article 82 of the French data protection act. There are narrow exemptions: authentication security, individual open tracking used strictly to clean a database or adapt the channel, and demonstrating compliance with a legal information duty. Addresses collected before publication had until 14 July 2026 under a transitional regime.

Italy went further and faster. The Garante adopted Provision No. 284 on 17 April 2026 and published it in the Gazzetta Ufficiale on 29 April, with a six month transition that closes on 28 October 2026. Consent is required, in the regulator’s phrasing, where pixels are used to carry out

“behavioral assessments or analyses”

If you send to French or Italian subscribers, you have a decision to make before the end of October, and most ESPs will not make it for you. Either add a separate, granular consent for open tracking at signup and in the preference centre, or switch open tracking off for those markets and measure on clicks and orders. I would do the second for most brands. You lose a metric you could not trust anyway, and you avoid building a consent flow that depresses your signup conversion rate.

Consent and compliance across four regimes

Most brands on this site’s readership send into at least two jurisdictions. The rules are not converging, so a single global template will fail somewhere.

Two US points are worth pinning down because the reporting on them has been messy. Twenty states now have comprehensive privacy laws in force, per MultiState’s February 2026 tracker. And the FCC’s “revoke all” rule, which would make an opt out from one message type apply to every message from that sender, has been delayed twice. It is now due on 31 January 2027. What did take effect on 11 April 2025, and remains binding, is the requirement to honour opt outs made by any reasonable method within ten business days.

India’s timeline is the one most Indian brands are underestimating. The DPDP Rules were notified in November 2025 with an eighteen month phase in, consent manager registration opening around November 2026 and full compliance by 13 May 2027, with penalties reaching INR 250 crore, as India Briefing sets out. If your list was built without a stored consent record and a notice trail, the remediation work is a year long, not a sprint. Start now.

List building that does not poison your sending domain

Every deliverability disaster I have seen started as a list building shortcut. Purchased data, a co registration deal, an unchecked import from a trade show. The complaint rate arrives about six weeks later and the domain takes three months to recover.

What works is boring. A single field signup on high intent pages, double opt in for any market where you need a consent record, real time address verification at the point of capture, and a welcome message that arrives within sixty seconds so the person recognises you. Incentives are fine, but a generic 10% off attracts discount hunters who never buy at full price; a genuinely useful asset attracts people who wanted the thing you sell. If your signup form sits on a page you have already optimised for conversion rate, treat the form as a separate test surface rather than assuming the page result carries over.

And be honest about frequency at the point of capture. “Two emails a week, product news and offers” sets an expectation you can meet. “Join our newsletter” sets none, so the first promotional send reads as a surprise, and surprise is what people report as spam.

Segmentation that moves revenue rather than reports

Most segmentation decks are demographic. Most revenue comes from behaviour and recency.

The segments that consistently pay are engagement recency (opened or clicked in 30, 60, 90, 180 days), purchase recency and frequency, product or category affinity, average order value band, and predicted churn risk if your platform calculates it. Suppress anyone with no engagement in 180 days from routine campaigns and put them in a dedicated winback track. That single move usually lifts campaign rates and cuts complaints at the same time, because you have stopped sending to people who forgot you exist.

Beyond that, the useful frontier is joining email behaviour to site and product data. A subscriber who viewed a category three times this week and did not buy is a different person from one who opens everything and never clicks. Getting that join right is an analytics and data plumbing job before it is an email job, which is why so many segmentation projects stall.

The automation flows worth building, in order

Klaviyo’s 2026 benchmark data, drawn from more than 183,000 customers, makes the case better than any argument. Flows clicked at 5.58% against 1.69% for campaigns, and generated roughly 41% of email revenue from 5.3% of sends, with revenue per recipient close to eighteen times higher. If you have limited time this quarter, spend it on flows.

Two flows deserve more care than they usually get. The welcome series is where deliverability is won, because engagement in the first week teaches the mailbox provider whether people want you. Send the first message immediately, the second within 48 hours, and make at least one of them ask for a reply or a preference. The sunset flow is where deliverability is protected. Give a lapsed subscriber two clear chances, then suppress. Do not run an annual “we miss you” blast to everyone dormant; that is how domains get burned.

SMS, WhatsApp and RCS

SMS is not email with fewer characters. It is a permission you can exhaust in a fortnight.

Benchmark ranges compiled by ClickMinded from Klaviyo, Postscript and Attentive data put median SMS click through around 8.7%, with flow based messages near 10%, roughly double campaign performance. Klaviyo’s own “good” ranges sit at 8.9% to 14.5% click, 0.6% to 1.4% unsubscribe and 1.0% to 2.0% placed order rate. When opt outs climb past about 3.5%, the cause is nearly always frequency rather than creative.

The channel mix question is regional. WhatsApp carries most conversational commerce in India, Brazil and much of the Gulf, and it is metered per conversation, so a template heavy broadcast strategy gets expensive fast. RCS is the interesting one for 2026 because Apple added support in iOS 18, which finally gave brands a rich messaging format that reaches both platforms; Messente’s tracking shows business adoption climbing on the back of that. Treat RCS as an upgrade path for your existing SMS programme, not a new channel to staff.

For most brands the rule is simple: email carries the story, SMS carries the deadline. Use SMS for cart recovery, restock alerts, delivery updates and genuinely time bound offers. Nothing else.

Design, AI and writing for an inbox that reads first

Since January 2026 the first reader of a Gmail message is frequently a model producing a summary or an inbox priority score. That changes the craft in ways that are not obvious.

Put the offer and the deadline in the first hundred words of the body, in plain text, not only inside a hero image. Keep a real text version, because a message that is entirely images gives a summariser nothing to work with and gives spam filters a reason to be suspicious. Use a consistent, recognisable From name, because sender recognition is now part of how the inbox decides what to surface. Preheader text should complete the subject line rather than repeat it.

Dark mode still breaks more emails than any other rendering issue, so test it. Apple’s Mail categorisation, which files most marketing into Promotions or Updates and can collapse repeated sends from one domain into a stacked digest, means your third message of the week is competing with your own second message, as MarTech has covered.

On AI in production: use it for variant generation, subject line drafting and translation, and keep a human on the final read. Klaviyo reports AI driven product recommendations lifting click rates to 3.75% on average and 8.79% for the strongest accounts, the clearest measured win in the category right now. What these tools do well and badly is covered in our guide to AI in digital marketing. AMP for email remains a niche with weak client support. Skip it.

Benchmarks: what good looks like in 2026

Treat all of these as direction, not targets. Your list age, acquisition source and send frequency move these numbers more than your industry does.

Sources: WebFX 2026 benchmarks (drawing on Campaign Monitor and Mailchimp data across more than 30 billion emails), Klaviyo 2026 benchmarks, ClickMinded SMS compilation. Remember the open rate column is contaminated by Apple’s pre fetching, so a rising open rate on an Apple heavy list means very little on its own.

Choosing an ESP

Switching costs are high enough that this decision deserves more than a feature grid. Validity’s acquisition of Litmus, reported by MarTech, signals where the tooling side is heading: deliverability, data quality and design testing bundling together.

One practical filter: ask any shortlisted vendor how it handles per market suppression of open tracking, and how quickly it can produce a per subscriber consent record with a timestamp and source. The answers will separate the platforms fast, and both questions are now compliance requirements rather than nice to have features.

KPIs and a 90 day plan

Report on four things: revenue per recipient, click to delivered rate, list growth net of unsubscribes and bounces, and complaint rate by mailbox provider from Google Postmaster Tools. Open rate can stay on the dashboard as a directional signal for deliverability changes, clearly labelled as unreliable. Everything else is noise. If email revenue is being credited generously by last click, reconcile it against a holdout test at least twice a year; the method is the same one used across measurement and attribution generally.

The first ninety days, if you are inheriting a programme:

Weeks 1 to 3. Audit authentication for every sending service, including the ones marketing forgot about. Set DMARC to p=none, read the reports, then move to quarantine. Split subdomains. Check complaint rate by provider. Verify one click unsubscribe headers are present on marketing sends.

Weeks 4 to 6. Suppress everything with no engagement in 180 days. Build or rebuild the welcome series and cart abandonment flow. Turn off open tracking for French and Italian subscribers, or add granular consent, before 28 October.

Weeks 7 to 12. Segment by engagement recency and mail accordingly. Add post purchase and winback flows. Run your first holdout test. Fix the signup form and the preference centre, and start testing subject lines against click to delivered rather than opens.

None of that is glamorous. All of it compounds, and a programme with good plumbing and average creative beats the reverse. That gap has widened every year since the sender rules arrived. Our digital marketing overview covers where email sits against paid and organic, and the ecommerce marketing guide goes deeper on retention economics. For what shifted this month, see our email, privacy and data news for September 2026.

Last researched and updated: 7 September 2026.