Mailchimp's AI: Free, in Beta, English Only, and Gated to Two Tiers

Mailchimp's AI costs nothing extra, is in beta, works in one language, and is switched on as an add-on rather than included. All four of those are published, and together they describe a very different posture from the rest of this category.

Only one hundred and ten people a month search for Mailchimp's AI, which makes this the quietest part of a very loud brand. It is also the part where the published terms are most worth reading before you plan around it.

OUR TAKE: The pricing posture here is unusually generous and the availability posture is unusually narrow. There are no AI credits, no metering and no overage rate anywhere in this vendor's published material, which is the opposite of the nearest comparable product. What there is instead is a beta label, two tiers of gating, and a restriction to English in select countries. Plan for the restriction, not for the price.

What the vendor actually calls it

The brand is Intuit Intelligence, and the assistant inside it is Intuit Assist^[23]. The marketing page describes the premise in one line: Mailchimp with Intuit Intelligence turns store data into clear answers and recommends the next move^[23].

That framing matters for a buyer. This is positioned as analysis and recommendation over your own commerce data rather than as a content generator with a word count, which is why there is no credit meter to read.

## The four published conditions

There are two sentences, not one, and they do not say the same thing. That difference is the most useful detail on this page.

On the AI page, about the analytics brand: "Intuit Intelligence functionality (beta) is available to certain users with Premium, Standard and Legacy plans."^[23] Three limits. Beta, certain users rather than every user on those plans, and Premium, Standard and Legacy, which excludes Essentials and Free.

On the pricing and comparison pages, about the assistant: "Intuit Assist functionality (beta) is available to certain users with Premium, Standard and Legacy plans in select countries in English only."^[3] The same three limits, plus two more. Select countries, which are not enumerated, and English only.

So the assistant is restricted geographically and linguistically and the analytics layer, on the page that describes it, is not. Whether that is a real difference in availability or simply a difference in which footnote was written we cannot tell, and we are not going to merge the two sentences into one to make it tidy.

The price is stated for both, with a qualifier worth keeping: "Access to Intuit Assist is available at no additional cost at this time."^[3] The AI page says the same of Intuit Intelligence^[23]. At this time is the vendor's hedge, not ours.

No metering, and why that is unusual

We looked specifically for AI credits, included units, usage allowances and overage rates across this vendor's pricing pages, comparison table, help centre and legal terms. There are none.

Hold that against ActiveCampaign, the one direct competitor we have measured in full. There, the AI that suggests is included and the AI that acts is metered from zero included units on all four tiers, with a published rate per unit. A buyer of that product has to forecast AI consumption. A buyer of this one does not.

The trade is visible in the comparison table. The Generative AI Features row reads "No additional cost add-on" on Premium and Standard, and "Not included" on Essentials and Free^[3]. So the feature is free and it is an add-on: something you enable rather than something that is simply there.

## What the features are

The vendor names several, and the useful way to read them is by what they act on.

Analytics AI explains which campaigns and automations are moving revenue, and gives the plain-language reason why^[23]. Write with AI works directly in the email editor, in both the new and classic versions^[23]. Send-time optimisation looks at historical click-rate trends and predicts the best times to send, based on patterns from the past year^[23]. Segmentation analyses customer data to build segments from behaviour and engagement^[23].

Two of those four appear as their own rows in the plan comparison table, gated identically: Send Time Optimization and Anomaly Detection both read "Included" on Premium and Standard and "Not included" on Essentials and Free^[3].

The Claude connector

One integration is worth singling out because it changes where the work happens. The vendor publishes a connector that brings what you build in Claude into Mailchimp, with a demonstration that requires no signup^[23].

A footnote under that demonstration states that it is based on the Standard plan and that availability of features and functionality varies by plan type^[23]. We report that as a fact rather than as a quotation, because the sentence lives in the page's markup and not in its rendered text, and we do not put quotation marks around a string we cannot verify in what the vendor actually served. The practical point stands either way: the demonstration is of the second-highest tier, and it is not a promise about Essentials.

## The three performance claims, and their footnotes

The AI page leads with three numbers. Each has a footnote, and the footnotes are more informative than the headlines.

Up to thirty times return on investment, which the footnote qualifies as "Based on all e-commerce revenue attributable to paid plan users' Mailchimp campaigns from August 2024–August 2025. ROI calculation requires an e-commerce store connected to a Mailchimp account. Results vary."^[23]

Up to one point seven times more attributed revenue from abandoned-cart flows against bulk emails, based on internal data from January 2025 to January 2026, also requiring a connected store, also with results varying^[23].

Up to two times higher click rates for users who sent both email and SMS against email only, measured on United States users across roughly three years^[23].

All three are the vendor's internal data on its own paid users. Two of the three require a connected commerce store to apply at all. None is an independent study, and the vendor does not claim otherwise. We report them as published claims with their conditions attached, because a number with its footnote removed is a different number.

How the gating looks in the table

Reading the comparison table row by row is more useful than reading the AI page, because the table is where the tiers are stated.

Feature row	Premium	Standard	Essentials	Free
Generative AI Features	No additional cost add-on	No additional cost add-on	Not included	Not included
Send Time Optimization	Included	Included	Not included	Not included
Anomaly Detection	Included	Included	Not included	Not included
Multivariate Testing	Included	Included	Not included	Not included

The pattern is the same across all four rows, which tells you something simpler than the AI page does: the line is drawn between Standard and Essentials, not between Premium and everything else. Twenty dollars a month at the entry band buys the whole published AI surface. Thirteen buys none of it.

That is a cheaper entry to AI features than anything else we have measured in this category, and it is worth weighing against the beta label rather than separately from it.

## What the AI acts on, and why that matters

Three of the four named features read your own data rather than generating from a prompt.

Analytics AI explains which campaigns and automations are moving revenue, in plain language, so the output is an attribution argument rather than a draft^[23]. Send-time optimisation looks at historical click-rate trends and predicts the best times to send, based on patterns from the past year^[23]. Segmentation analyses customer data to build segments from behaviour and engagement^[23].

Only Write with AI is a generator, and it works inside the email editor in both the new and classic versions^[23].

That balance is the reason there is no credit meter. A feature that explains your own numbers does not consume a measurable unit the way a text generator does, and the vendor has priced accordingly. It also means the quality of what you get depends on how much data you have already put in. The vendor says as much: the recommendations are grounded in your store's real sales, campaigns and customer behaviour, and sharpen the more you use the product^[23].

The demand is small, and that is informative

One hundred and ten searches a month for this product's AI, against 6,600 for its pricing and 368,000 for the bare brand name. A further twenty for its AI features specifically.

Compare the question people actually ask: seven hundred and twenty a month want to know what "cleaned" means, which is a billing concept. Six and a half times the demand for the AI.

We report that because it is a check on our own emphasis as much as on the vendor's. A reader arriving at this product is overwhelmingly more interested in what it will charge them than in what its AI can write. The AI page leads with a thirty-times ROI claim; the search data says the audience is somewhere else entirely.

## Three questions to settle before you plan around it

Ask whether your account qualifies, in your country, in your language. All three conditions are in one published sentence and all three are narrower than the plan name suggests.

Ask what happens when the beta ends. The price today is nothing, stated plainly, with no commitment past the beta label.

And ask for the connected-store requirement in writing if the ecommerce claims are what sold you. Two of the three headline numbers require a connected store by the vendor's own footnotes, and one of them requires it to calculate at all.

Who this AI is actually for

The pattern across everything above points at one customer: a Standard or Premium subscriber with a connected ecommerce store, operating in English.

If that is you, the proposition is strong. The analysis runs on your own sales data, the recommendations are grounded in what you have sent and sold, there is no usage meter to watch, and the two automation-side features that the comparison table gates are already yours.

If you are on Essentials, the generative features and both gated automation features read "Not included" and the AI brand does not apply to you at all. If you are outside the select countries or working in another language, the published condition says the same thing more bluntly.

## What it costs to be wrong about this

A buyer can get the AI question wrong in two directions here, and they are not symmetrical.

Overestimating is the cheaper mistake. If you choose Standard partly for the AI and then discover your country or language is outside the published condition, you still have a twenty-dollar plan at the entry band with scheduling, A/B testing, dynamic content and send-time optimisation. The AI was never a line item, so losing it does not leave a hole in the invoice.

Underestimating is dearer. If you choose Essentials at thirteen dollars to save seven, the comparison table marks the generative features, send-time optimisation, anomaly detection and multivariate testing all as not included^[3]. Moving up later means moving tier, and by then your promotional twelve months are partly spent on the lower one.

So the seven dollars between Essentials and Standard at the entry band is the cheapest decision in this whole product, and the AI is only part of what it buys.

How this compares with the one rival we have measured

ActiveCampaign is the only direct competitor we have scored in full, and its AI model is the opposite shape.

There, the AI that suggests is included and the AI that acts is metered, with zero included units on all four tiers and a published rate per unit. A buyer has to forecast consumption. The upside is that availability is not gated by country or language in the way it is here, and the feature is not in beta.

Here, there is no meter to forecast and no rate to model. The limits are who you are and where you are, not how much you use. On our scale ActiveCampaign scores 3.2 on the AI axis against 3.0 for Mailchimp, and the gap is smaller than the difference in the models would suggest, because neither vendor publishes anything measurable about depth.

For a heavy user, free and gated beats included and metered. For a light user on a cheap tier, the reverse. There is no general answer, which is why we score the axis rather than recommend a winner.

What is not published

Three gaps worth naming before you build a plan on this.

Which countries. The phrase is select countries, and it appears on the Intuit Assist condition rather than on the Intuit Intelligence one, with no list anywhere on the pages we read.

Which users. The phrase certain users with Premium, Standard and Legacy plans is narrower than all users on those plans, and the qualifying criterion is not stated.

And how long the beta lasts. Beta is stated on both features, with no timeline and no commitment about pricing after it ends. Since the current price is zero, that is the one number most worth watching.

What a reasonable plan looks like

Put all of it together and the sensible posture is narrow but genuinely positive.

If you are on Standard or Premium, in an eligible country, working in English, with a connected commerce store, the AI here is the best-priced in the category because it has no price. Enable the generative add-on, let send-time optimisation and anomaly detection run, and treat Analytics AI as an attribution second opinion rather than an oracle.

If any one of those four conditions fails, plan the tier on everything else it buys and treat the AI as a possible bonus. The seven dollars between Essentials and Standard is justified by scheduling, testing and dynamic content on their own.

And if you are modelling a budget, model zero for AI and zero for AI overage, because that is what the vendor publishes today. Watch the beta label rather than the price. A feature that is free in beta is a feature whose price has not been announced.

## How to evaluate it in practice

Three things to do rather than assume.

Confirm eligibility for your own account and country before you price the tier around the AI. The published condition makes availability a per-account question, not a per-plan one.

If you are comparing against ActiveCampaign, compare the models rather than the feature lists. One includes the AI and meters the actions; the other includes both and gates by tier. For a heavy user, free and gated beats included and metered. For a light user on a cheap tier, the reverse.

And treat the thirty-times figure as what it is: the vendor's own aggregate across a year of paid users with connected stores, with results varying by its own statement. Useful as a direction. Useless as a forecast.