There is a number from this week that every marketing leader should sit with for a minute. 2.31 out of 7.
That is how morally objectionable the American public finds the idea of replacing a search marketing strategist with software.
The scale ran to 7. Across 940 occupations, only file clerks scored lower.
The study measures permission, not capability
James Riley at Harvard surveyed 2,357 respondents and asked them to rate how objectionable it would be to hand each of 940 jobs to a machine. Greg Jarboe covered the findings for Search Engine Journal this week.
Read the design carefully, because the distinction is the whole story. Nobody was asked whether AI can do the work of a search marketer. They were asked whether they would mind.
They would not.
That is a different category of finding than the ones our industry has been arguing about since 2023. Capability studies tell you what the technology can reach. This one tells you what the public will accept once it gets there.
And the public has already signed off on us.
I have watched the counter-argument play out in agency rooms for three years. Clients want a human. Buyers want judgment. Relationships cannot be automated.
Every one of those is a competence claim wearing a moral costume. Riley's data pulls the costume off.
Objection is conditional, and the condition is performance
The second finding is the one that should settle the internal debate.
When respondents were asked to imagine a more advanced AI that outperformed humans at lower cost, support for automation nearly doubled. It moved from roughly 30 percent of occupations to 58 percent.
So the objection people hold is not a floor. It is a waiting room.
Elisabeth Paulson's research points the same direction with far higher stakes attached. Across studies covering loan decisions and pretrial release, among respondents who believed algorithms outperformed humans, 56 percent and 54 percent chose the algorithm.
Those are decisions about someone's credit and someone's liberty. If perceived competence moves the answer there, it will move the answer on keyword strategy without anybody writing a think piece about it first.
Preference tracks competence. It does not track category, and it does not track how much you like your job.
I wrote in an earlier edition about why AI was not cutting your team yet. The reason was never public conscience. It was that the output did not clear the bar.
The bar is being cleared where we felt safest
Procter and Gamble ran a study across 791 product developers. Ideas that ranked in the top 10 percent were three times more likely to come from AI-assisted teams than from individuals working alone.
Top decile ideas. Not throughput, not first drafts, not raw volume. Idea quality, which is precisely the ground marketers retreat to when the automation conversation gets uncomfortable.
Be careful how you read that. It is one company, in product development, under study conditions, and a headline productivity figure from someone else's environment does not transfer to your team just because it was published.
But the direction is not ambiguous, and direction is what you plan against.
The number is worse for agencies than for in-house teams
This is the part the coverage skipped, and it is the part that decides who feels this first.
An in-house search marketer sits inside a P&L, carries institutional memory, and is attached to a dozen relationships that have nothing to do with search. Removing that person is a reorganisation.
An agency search strategist is a line on an invoice. Priced explicitly, reviewed annually, and compared against alternatives by a procurement team whose job is to find cheaper equivalents.
Same role, same 2.31 score, completely different exposure. When the cost comparison arrives, the invoice line loses first because it is the only one anybody can see clearly.
That is not a reason for agencies to panic. It is a reason to stop selling the thing that appears on the invoice as hours.
What actually protects the role
If public objection will not hold the line, and competence is closing, then the defensible part of the job is neither of those.
It is accountability.
The work that survives is the work where a name sits on the outcome. Someone who can be called at nine in the evening when blended CAC moves the wrong way, and who can explain from memory which of last Thursday's four changes caused it.
A model can produce the recommendation. It cannot carry the consequence, and consequence is what clients are buying when they say they want a human.
Practically, that means splitting delivery in two.
Repetitive work with no judgment attached goes to the machine without ceremony. Ad copy variants, feed hygiene, bulk meta descriptions, first-pass keyword clustering, reporting assembly.
Trust-sensitive work keeps a human name on it. Anything a client forwards to their board, anything that makes a factual claim about their business, anything where being wrong costs them a customer rather than an impression.
Riley's own conclusion lands in the same place. Automate the repetitive tasks nobody enjoys, keep human attribution where trust is at stake, and understand that the technical competence gap is the only thing left to hold.
The uncomfortable part for our industry is that we have spent a decade selling the other thing. Hours, presence, and the reassurance of a face on a weekly call.
2.31 out of 7 is the market telling you what it thinks that is worth.
The public is not going to defend your job. Build the version of it that does not need defending.