Should AI Set Your Email Send Frequency or Should You Cap It

AI controlling send volume and cadence versus human frequency caps is a narrower marketing decision than the framing suggests. In the documentation we read for Klaviyo, Braze, Iterable and Salesforce, we found no feature that sets volume without a boundary a person configured: where a model is involved at all, it works inside a range, a cap or a journey path someone approved. The useful question is which numbers the model should own, and the answer runs through the spam-rate thresholds Gmail and Yahoo publish. This page assumes you know what email deliverability is and want the frequency decision on its own.

What do the frequency features in Klaviyo, Braze, Iterable and Salesforce decide?

The table records what each vendor's own documentation said when we read it on October 7, 2026.

Klaviyo's product name invites the wrong assumption. Its help article tells readers to "Think of Smart Sending like a countdown timer on each profile", and we found no mention of a model in it. Klaviyo also states that "Messages skipped due to Smart Sending are not rescheduled automatically", which matters when a flow email loses its slot to a campaign sent earlier the same day.

Braze documents a similar rule engine with more controls: up to 10 capping rules per workspace, counted by calendar day in the user's time zone, and "When rules conflict, the most restrictive, applicable frequency capping rule is applied to your users." Its self-serve agent stops short of frequency, even though the Decisioning Studio overview lists frequency among the dimensions Braze's decisioning agents experiment across. In the Decisioning Studio Go setup guide, Braze says the agent "picks days and times within the frequency you set", and the reference table lists frequency as "Single selection (no frequency decisioning)".

Iterable is the one of the four where the model's number becomes the cap. With Frequency Decisioning (formerly Frequency Optimization), Iterable has the marketer enter a minimum and a maximum, and each user's cap moves inside that range: people who open and click can get a higher cap, and people who don't engage drift toward the minimum. A user with no recent sends has no data, "so Iterable uses the maximum cap you set", which puts long-quiet and newly added contacts at your ceiling.

Salesforce takes a third route. Its model card says Einstein Engagement Frequency powers the Frequency split activity in Journey Builder, lists unsubscribes and spam complaints among the model's inputs next to opens and clicks, and needs the audience to receive at least five variants of your messaging, meaning sending patterns with different intervals, over a 28-day period before it can model anything.

Do Gmail and Yahoo's bulk sender rules limit how often an AI tool can send?

The Gmail and Yahoo bulk sender rules put no send frequency on AI email tools or on any other sender; we found no weekly or daily message limit in either provider's requirements. What they publish are thresholds on outcomes. In its sender FAQ, Google defines a bulk sender as "any email sender that sends close to 5,000 messages or more to personal Gmail accounts within a 24-hour period". The sender guidelines from Google add the number that governs cadence: "Keep spam rates reported in Postmaster Tools below 0.10% and avoid ever reaching a spam rate of 0.30% or higher."

Yahoo draws its line at the upper figure ("Keep your spam rate below 0.3%") and adds the closest thing either provider has to a cadence rule: "Honor the frequency of the list's intent. Don't start sending daily emails to subscribers of your weekly or monthly mailing." An engagement-optimizing model can break that quietly, because a subscriber who signed up for a weekly digest and clicks every issue looks like someone who can take more.

Google's FAQ adds the enforcement: spam rate is calculated daily, bulk senders above 0.3% are ineligible for mitigation until the rate stays below 0.3% for 7 consecutive days, and "Starting November 2025, Gmail is ramping up its enforcement on non-compliant traffic".

How do the Gmail and Yahoo spam filters treat a frequency that AI picked?

An AI email sending frequency gets no separate treatment from the Gmail or Yahoo spam filter in the sender requirements either provider publishes. We found no reference to AI or automated cadence in them, only to complaint rates, authentication, unsubscribes and volume patterns. Two of those are frequency problems.

The first is the daily calculation. A rate measured each day means the complaint allowance scales with that day's mail. The hypothetical below assumes the rate is spam reports divided by mail delivered to inboxes that day, the basis Yahoo describes; Postmaster Tools reports the figure Google uses.

On the flow-only day, 18 reports put the domain at the level Google says to avoid ever reaching, so a frequency change needs watching on the quiet days as well as the campaign days.

The second is the ramp. Google's guidelines say to "Avoid introducing sudden volume spikes if you do not have a history of sending large volumes", and warn that immediately doubling previous volume could result in rate limiting or reputation drops. A per-user optimizer switched on with a generous maximum is one way to produce that jump.

Who owns spam compliance when an automation tool picks the frequency?

AI email automation tools can choose a sending frequency, and spam compliance with Gmail and Yahoo stays attached to the sending domain, which means it stays with you. The tool's settings and the providers' rules meet in three places:

• Unsubscribes. Yahoo tells bulk senders to "Honor unsubscribes within 2 days", and Google recommends fulfilling requests within 48 hours.
• Exemptions. Klaviyo never skips transactional messages, Iterable leaves them out of its caps, and Braze lets a campaign override the rules. Google's FAQ says message recipients determine the nature of the messages they receive, so labeling a promotion as transactional to get past a cap moves the risk into the complaint rate.
• Caps that don't add up. In its Frequency Management article, Iterable says caps apply to each channel and each message type independently and don't combine to an overall limit. Braze notes that by default a campaign that does not obey frequency caps does not count toward them either.

A starting cap policy and the signals to watch before loosening it

For AI-optimized send frequency in email marketing, the frequency capping best practices we would carry from 2025 into 2026 fit in one table. The numbers are our starting assumptions for a brand that mails a few times a week; no vendor or mailbox provider publishes them.

Loosen one segment at a time, by one send a week, and only while all of these hold:

1. Gmail's Postmaster Tools spam rate has stayed under 0.10% on every day of the test, low-volume days included.
2. Yahoo's Complaint Feedback Loop, which Yahoo recommends for monitoring complaint rate, shows the domain below 0.3%.
3. The unsubscribe rate in the optimized group is no worse than in a holdout kept on the fixed cap. Iterable reports lift, unsubscribes and send skips against a control group.
4. Skips are rare. A cap that skips a large share of each campaign's audience means the calendar is overbooked, and a higher cap only hides that.
5. The change in weekly volume stays far short of the doubling Google warns about.

If you are sizing what an extra weekly send is worth, our email marketing statistics page collects the sourced revenue figures, and how much email marketing costs covers platform pricing by list size. The email deliverability checker and the lifecycle email playbook are further resources, and cap design is part of our lifecycle and demand generation work.

Does advertising split AI and human control the same way?

In advertising, AI frequency capping and human control over cadence split in a similar way in 2026, with one difference from email: in Google Ads Display campaigns a person picks either platform optimization, which Google labels recommended, or a manual cap. For those campaigns, Google Ads says "You can allow Google Ads to optimize how often your ads show to users, or you can set a frequency cap manually to manage the number of impressions on the campaign, ad group, or ad." We did not review other ad platforms for this page.

Email carries the heavier penalty, because spam reports count against the domain every later message depends on. Let the model choose the number for each contact, and keep the range, the exemptions, the rule for contacts with no history and the pace of change in human hands.

Sources

• Google: Email sender guidelines
• Google: Email sender guidelines FAQ
• Yahoo Sender Hub: Sender Requirements and Recommendations
• Klaviyo Help Center: Understanding Smart Sending in Klaviyo
• Braze Docs: Rate limiting and frequency capping
• Braze Docs: Set up your Decisioning Studio Go agent
• Braze Docs: BrazeAI Decisioning Studio
• Iterable Support Center: Frequency Decisioning
• Iterable Support Center: Frequency Management
• Salesforce Help: Einstein Engagement Frequency Model Card
• Google Ads Help: Use frequency capping