Even OpenAI Still Buys Meta Ads and Sends You Emails. Your Channels Aren’t Dead, Your Measurement Is.

Digital Marketing TransformationB2B MarketingLeads Generation Marketing
Written by Jaslyin Qiyu
Cover of The 2026 Advertising Platform Map report by Mad About Marketing Consulting
Free Report

Get The 2026 Advertising Platform Map

Where do B2B and B2C budgets really work in 2026? This report compares the major platforms, shows what the AI labs’ channel mix teaches CMOs, and sets out why email still earns its place.

  • Scale and growth for Meta, Google, Amazon, TikTok, LinkedIn, Reddit and ChatGPT Ads
  • B2C and B2B playbooks with benchmark ranges
  • Email in 2026: return, deliverability and consent
  • An Asia-Pacific view, a four-question channel test and nine CMO actions
Download the ReportPDF, 11 pages

I noticed two things recently. The companies building the most advanced AI on earth are advertising on Facebook and Instagram. And one of them is emailing me, though I never remember subscribing to anything.

Meanwhile, my feed is full of marketers declaring organic reach dead, email dead, and Meta ads too expensive to bother with. If the AI labs run the oldest playbook in digital marketing, maybe the problem isn’t the channels.

The labs run a portfolio, not a platform

Start with what these companies actually do. A recent analysis of Meta’s public Ad Library by Sprites.ai counted 1,211 Claude ads on Meta since February 2026, 379 of them through 84 creator pages, mostly promoting Claude Code and Cowork. The same analysis counted 629 OpenAI ads on Meta, 560 on Google and 302 on LinkedIn. These are vendor-compiled counts, so read them as directional. The pattern is still clear.

Look at what each channel is doing:

  • Meta buys cheap, scaled reach with creator-style video.
  • Google captures intent, with task-specific landing pages for specific searches.
  • LinkedIn runs account-based campaigns aimed at open sales opportunities.
  • ChatGPT Ads is a new intent layer, launched in February 2026 and reportedly at a $1 billion annualized run rate by August.
  • Email is the owned channel where acquired users get nurtured.

None of these companies picked a winner. They built a portfolio where each channel has a job. When a marketer tells me “X is dead,” I now ask which job X was hired to do and who measured it.

Email: the unglamorous winner

Industry compilations citing DMA and Litmus put email’s return at $36 to $42 for every $1 spent, with ecommerce closer to $45. HubSpot’s 2026 State of Marketing puts email level with organic social as the second most used channel, and 75% of marketers plan to hold or raise their email spend.

Only 12.5% of companies measure email ROI adequately, and 36% don’t measure it at all (Litmus, as compiled in 2026 roundups).

A channel that looks this good, and that most teams can’t prove, is not a success story. It is a measurement gap with a flattering headline.

Why your numbers lie

Three things have quietly broken the dashboards most teams still use.

  1. Deliverability. Validity’s benchmark found roughly 1 in 6 legitimate marketing emails never reaches the inbox. Since February 2024, Gmail and Yahoo have required authentication (SPF, DKIM and DMARC) from bulk senders.
  2. Inflated opens. Apple Mail Privacy Protection pre-loads tracking pixels, so a machine “opens” the email before a human does. Estimates of the inflation vary widely, from a few points to 15 to 25 points depending on how Apple-heavy your list is. Open rate has become a vanity metric.
  3. AI inboxes. Gmail’s Gemini features can decide how prominently your message appears even after it lands. One vendor estimates up to 40% of mail reaching Gmail inboxes is pushed into low-visibility groupings. Treat that figure as directional, but the direction is real: engaged lists win, big passive lists lose.

Cold B2B email has fared worst. One vendor analysis reports open rates on cold outreach falling from 35 to 45% in 2022 to 12 to 18% by early 2026, with reply rates on generic sequences dropping from 8 to 12% to 1 to 3%. Permission-based B2B email holds up far better, with click-through averaging roughly 2 to 4%.

So email isn’t dead. Spray-and-pray email is. If your report still leads with open rate, you are managing a number that no longer measures a human.

The consent bill

This is the part that got my attention. OpenAI updated its privacy policy on May 1, 2026, adding explicit language about direct marketing and sharing limited identifiers with marketing partners. WIRED tested two free accounts and found the marketing privacy settings switched on by default. The opt-out sits under Settings, Data Controls, Marketing Privacy. The policy also recognizes Global Privacy Control and includes a marketing communications unsubscribe right.

I want to be careful here. That update concerns ad targeting and marketing data sharing. I could not find a source describing exactly how consent for OpenAI’s marketing emails is captured, so I can’t tell you it was buried in the terms. What I can tell you is the pattern, and it is not unique to OpenAI.

Default opt-out is generally lawful in US-style regimes like CAN-SPAM, provided unsubscribe requests are honored. The EU and UK are stricter and generally expect consent, with a limited carve-out for existing customers. Singapore’s Spam Control Act requires an unsubscribe facility for unsolicited commercial electronic messages. None of this is legal advice, so check with counsel before you build a program on it.

Here is my point for CMOs. Legal and wise are different tests. Default-on consent buys volume and spends trust. For a consumer app growing at breakneck speed, that trade may pencil out. For a bank, an insurer or a healthcare brand, where trust is the product, it rarely does. The cheapest email list is the one people chose to join.

A channel portfolio test

Before you declare any channel dead, or fund it, ask four questions:

Reach

Where do we buy attention at scale, and what does it cost per meaningful result?

Intent

Where are people already asking, including inside AI assistants?

Owned

What can we reach without paying a platform each time?

Proof

Can we show what each channel produced without trusting its own dashboard?

If a channel fails the proof question, you haven’t found a dead channel. You’ve found an unmeasured one. That intent question matters more each quarter. Forrester found that 94% of business buyers now use AI in their buying process, which means your visibility inside AI answers is becoming a demand generation job, not a side project.

What to do on Monday

  1. Retire open rate as a headline KPI. Report clicks, replies, pipeline and revenue.
  2. Fix the plumbing. Confirm SPF, DKIM and DMARC are in place, and clean your list for engagement, not just bounces.
  3. Stop generic cold sequences. Replace them with permission-based nurture and signal-triggered outreach.
  4. Run a capped ChatGPT Ads test. Put 2 to 5% of your search budget behind it, pick a handful of high-intent use cases, and set a 90-day kill-or-scale rule against your Google Search cost per acquisition.
  5. Audit your consent records. Know what you collected, when, how, and whether you could defend it to a regulator or a journalist.

The takeaway

The AI labs aren’t proof that old channels are immortal. They are proof that discipline beats fashion. They run a portfolio, test creative every couple of weeks, measure what they can prove and own the relationship once they’ve paid to start it. Do that, and most of your “dead” channels come back to life.

Mad About Marketing Consulting helps leadership teams turn marketing, customer experience and AI ambitions into working, measurable systems. If your channel mix and your dashboards are telling different stories, let’s talk.

Explore MAMC advisory

Sources

Ad counts and most email statistics come from vendor compilations and should be read as directional. Legal descriptions are general and are not legal advice.

Jaslyin Qiyu
Next

NPS Isn't Your Revenue Problem. The Missing Alignment Is.