The Revision Tax: Why AI Made Your Marketing Slower, and the Gate That Fixes It

Two surveys this year measured the same thing from opposite ends, and between them they explain why the AI tools most marketing teams bought have not made those teams faster.

Optimizely surveyed more than 2,000 marketing leaders and described a revision tax: the time saved on drafting is consumed by reviewing, correcting and refining what the tools produced. Typeface surveyed more than 200 VP-level marketing leaders and found that 88 percent said their teams can now generate content quickly, while the share of campaigns taking one to two months to launch rose from 5 percent to 34 percent in a single year. More AI, faster drafts, slower campaigns.

These are vendor surveys with their own products to sell, so hold the percentages loosely. The mechanism they describe is not in doubt, because it is arithmetic.

The arithmetic

A marketing process built for people is built around a scarce input: the draft. A draft used to cost a writer a day. Because drafts were expensive, the process could afford to treat each one carefully: three or four stakeholders reviewing it in rounds, a designer redoing the image when the copy changed, a builder rebuilding the email after the approver’s wording change. The review cost was large per draft and the number of drafts was small, so the total was tolerable.

Now the draft costs minutes. The marketer who used to produce one produces three variants and a channel version of each. Every one of those enters the same review process, with the same stakeholders, the same rounds and the same rebuilds. The review cost per draft did not fall. The number of drafts went up by a multiple. Total review time rose.

That is the revision tax: the review process was sized for scarce drafts and is now being run on abundant ones. It is why the complaint from marketing teams is so consistent and so confusing to the people who bought the tools. The tool did exactly what it promised. The process absorbed the gain and then some.

There is a second, subtler cost. Because drafts are cheap, more of them are speculative: produced to see what the tool does rather than because the brief called for them. Each speculative draft that enters review costs full review attention and was never going to ship. We looked at the version of this where the tool is a general-purpose assistant and the output lands in a marketer’s lap with nowhere to go in ChatGPT wrote my campaign, now what.

Why the usual fixes do not work

Less AI. Returns the team to scarce, expensive drafts and the same slow review path. The campaigns take as long as they did in 2023 and the tools sit unused. Nothing was learned.

More reviewers. Adds capacity to the review station and adds a handoff to every piece. The rounds get longer, not shorter, because each new reviewer has opinions, which we worked through in from ten stakeholders to two approvals.

Better prompts. Produces better first drafts, which is genuinely useful, and does not change what happens to them. A better draft still enters the same rounds. Knak’s 2026 survey put it well: the problem is everything that happens after the idea is approved.

A rule that AI output gets reviewed harder. The most common response and the most expensive. It formalises the tax. Every draft now gets extra scrutiny because of where it came from, regardless of whether it needs it.

None of these touch the arithmetic. The only fix that does is to change what review costs per draft.

The gate

Review costs what it costs because three different questions are being asked of each finished draft, by several people, in rounds: is this the right thing, is it safe, is it good. The revision tax falls when those are separated and only the last one is left for a person.

Is this the right thing? Decided at the brief, once, before any drafting. The audience, the offer, the message, the claims we will make, the dates. Anyone whose opinion matters on those gives it at the brief. Then drafts are reviewed against the brief, which is a one-round question, not against each reviewer’s fresh view, which is infinite.

Is this safe? Decided by a check, before a person looks. Are the claims on the approved list? Do the links resolve? Is the wording inside the voice rules? Are the disclosures present? Does it render? Is the image on-brand? A review agent runs that list on every draft, flags what fails, and the person sees only drafts that passed. Most of what the rounds used to catch is on that list, and it was the most expensive attention in the building doing it. What the check finds that people miss is in how a review agent catches what humans miss.

Is this good? Decided once, by one named approver per piece type, at one gate, on a draft that has already passed the first two questions.

The review cost per draft falls from several people in rounds to one person, once, on a pre-checked piece. The arithmetic flips: drafts are cheap and review is cheap, so abundant drafts are an advantage instead of a burden. Three variants cost three quick decisions, not three review cycles.

What has to be true for the gate to hold

The gate only works if two things exist, and most teams that have tried it and failed were missing one.

A written standard. Voice, look, claims. The check needs something to check against, and the approver needs something to approve against. If “on-brand” lives in the designer’s head, the review agent cannot apply it and the designer is back in the loop on every image. The standard is short, the specialists write it, and the whole of this series depends on it; the setup is in expand what your marketing team can do.

Copy before images, in one place. A large share of revision cycles are images being redone because the copy changed after they were made. If images are generated from approved copy, inside the brand look, in the same place the copy lives, that cycle does not exist. It also removes the handoff to a designer and back, which was a wait on each side.

Teams that put the gate in without the standard get a check that flags everything or nothing. Teams that put it in without fixing the copy-then-images order keep paying the tax on every visual.

Where the saved time goes

Here is the part that matters to the CEO, because “faster review” is a cost story and the point of the tools was never cost alone.

When the review cost per draft falls, the hours come back to the marketer who was spending them in rounds, and the question is what they do with them. The answer that compounds is that they carry more of each piece themselves. The writer who no longer waits on three review rounds and a design cycle ships the post, its images and its email build. The demand gen manager who no longer waits on creative review runs the test this week. Each marketer’s expertise starts to set what they can do, rather than the review committee and the specialists they used to wait on. That is the productivity story that the revision tax was hiding, and it is in why your marketing team is so slow.

Measure it with two numbers before and after: rounds per piece, and handoffs per piece. The first should go from three to five down to one. The second from eight to twelve down to two. Both are in the productivity metrics a CEO should read.

Where Marqeable fits

Marqeable is built with the gate in the line. A marketer writes the brief and it is approved once. Agents draft every piece of the campaign from it, copy and then images, in the team’s voice and brand look. A review agent checks the claims, the links and the brand before anyone looks. The marketer adjusts, one named person approves, and the same workspace sends the email and the text, publishes the post, runs the follow-up and traces revenue to the exact message. The human is the last reviewer, not the first, and the drafts being cheap is finally the advantage it was supposed to be.

Frequently asked questions

What is the revision tax?

The term Optimizely used in its 2026 research with more than 2,000 marketing leaders for the pattern where the time saved by AI on drafting is consumed by reviewing, correcting and refining what it produced. The net time saved approaches zero, and in teams that generate many more drafts than before it goes negative.

Why has AI made marketing campaigns slower?

Because it made one step cheap and left the others alone. Drafts used to be scarce, so a process could afford to review each one in rounds by several people. Drafts are now nearly free and arrive in multiples, and the same review process is applied to every one. Typeface’s 2026 survey found the share of campaigns taking one to two months to launch rose from 5 percent to 34 percent in a year while 88 percent of leaders said their teams generate quickly. The draft got faster; the path from draft to live did not.

How do you reduce the time spent reviewing AI content?

Three changes. Approve the brief and the brand rules before drafting so reviews are against an agreed standard rather than fresh opinions. Run a review agent that checks claims, links, voice, disclosures and rendering before a person sees the piece, so the human looks only at what passed. And replace review rounds by several stakeholders with one named approver per piece type at one gate.

Should we use less AI to avoid the revision tax?

No. Less AI returns you to scarce, expensive drafts and the same slow review path. The tax is a property of the process, not of the volume. Keep the cheap drafts, move the review work that is mechanical to a check that runs before the human, and spend the human’s attention once, on whether the piece is good.

The bottom line

AI made drafts free and left review as expensive as it was, so a process sized for scarce drafts now runs its rounds on abundant ones. That is the revision tax, and it is arithmetic, not a tool problem. Decide the brief and the rules upstream, let a review agent run the safety checks before a person looks, and give the one remaining judgment to one approver at one gate. The review cost per draft collapses, the saved hours go to marketers carrying whole pieces, and the tools you bought start producing what they were bought for: more leads generated, and more of them won as customers.

See each station of the factory: build campaigns from one brief, review and approve content, run follow-up with automations, answer visitors with website chat, and tie revenue to the exact message with attribution.

Marqeable runs your campaigns, answers every visitor, text, and email in seconds, and turns them into booked jobs and meetings - even at 9pm on a Saturday. We’re in private beta with a small early cohort. Get early access