Your social media advertising agency should read the incentive first

The Gartner number every social media advertising agency now has to answer for#

Foxes do not argue with a scent, they follow it and stay sceptical of where it leads. That is the posture a social media advertising agency needs now, because the trail has changed shape underfoot. By 2028, more than 70% of global ad spending and 80% of US ad spending will flow through self-serve advertising platforms in which AI "materially influences" media buying, costs and outcomes, according to Gartner. That is up from 50% in 2025, a doubling of the platform's own grip on the budget in three years.

The prediction, first published by Gartner's own research and independently confirmed with matching figures and the same named analyst by Raconteur, is not a forecast about better tools. It is a forecast about who holds the steering wheel. A social media advertising agency that still frames its job as picking the right platform is answering last decade's question.

Picture the ordinary Tuesday version of this. A campaign manager opens a dashboard, and a small module suggests raising the daily budget by a tidy, round percentage. The suggestion arrives with a confidence score and a plain-language reason attached, both generated by the same system that benefits if the manager clicks accept. Nobody wrote that module to deceive anyone. It was simply built to grow the platform's own number, and a social media advertising agency's whole value now sits in the gap between that suggestion and the client's actual outcome.

Sit with that shape for a second: the target line is not a stretch goal set by a marketer, it is a platform's own prediction of how much of every brand's budget its own AI will soon touch. A social media advertising agency that treats the platform's dashboard as neutral is trusting the fox to mind the henhouse and calling it strategy.

Self-serve is the word doing the quiet work in Gartner's own definition, and it is worth pausing on. A self-serve platform hands the buyer the console, the sliders, and an AI wizard offering suggestions, but the model behind the wizard is trained on the platform's own data, tuned to the platform's own goals, and updated on the platform's own schedule. A social media advertising agency operating inside that console is not choosing between neutral options, it is choosing how hard to push back against a system built by someone else, for someone else's benefit first.

Why a social media advertising agency has to read the incentive first#

Gartner VP Analyst Eric Schmitt put the mechanism more plainly than most vendors would like. "It's kind of unsurprising that a lot of the AI wizard recommendations that I see surfacing in these user interfaces start with, 'You should spend more with us,'" he told Marketing Dive. His diagnosis of why is sharper still: "The AI was created in service of the platform, rather than the buyer... When the two competing objectives meet, maximising price for the platform and minimising cost for the buyer, it's a no-brainer to see which the AI will express bias toward."

This is not a conspiracy theory dressed up as analysis, it is a straightforward description of what the automation was actually built to optimise. Google's own help documentation says Performance Max campaigns use its AI to automatically funnel your budget toward the highest performing inventory and placements across Search, Shopping, YouTube, Display and Gmail. Funnelling is the honest word for it: money flows toward wherever the platform's own model decides it converts best, and the platform grades its own homework.

None of this makes the automation useless, and a good paid social advertising agency does not throw the tool out. It makes the recommendation a claim rather than a fact, which is a different thing to buy. The regulatory ground under this is shifting too: the FTC's enforcement policy statement on personalised pricing covers consumer-facing personalised pricing rather than ad-platform spend bias specifically, but it is evidence of the same wider unease about algorithms quietly setting a price nobody can see the working for.

A preprint on algorithmic collusion of pricing and advertising on e-commerce platforms, not yet peer-reviewed, models exactly the dynamic Schmitt describes: pricing and advertising algorithms on a shared platform can drift toward outcomes that favour the platform without any human ever agreeing to it.

The pattern is not confined to one platform. Google's own wizard funnels budget within its own inventory; other self-serve consoles run the identical logic in their own walled gardens, each one grading its own homework and calling the grade a recommendation. A paid social advertising agency working across several of these consoles at once sees the pattern plainly: every wizard nudges toward more spend on itself, never toward a rival platform, and rarely toward spending less at all.

The maths for a social media advertising agency is now unavoidable: an AI recommendation that always trends toward more spend is not a neutral signal, it is a sales pitch wearing a spreadsheet.

What Reddit's own numbers prove about speed#

None of this is theoretical, and none of it needs Reddit to be accused of anything to prove the point; it is simply the clearest public evidence of how fast advertisers are already being funnelled into AI-recommended, self-serve flows across the whole industry. Reddit COO Jen Wong told the company's Q2 2026 earnings call that "the number of advertisers using Max grew over 60% from Q1, while Max revenue grew over 150% during the same period." A Reddit SEC exhibit filed the same quarter put total advertising revenue at $625 million, up 74% year-over-year.

Read that chart the way a fox reads fresh tracks in the frost: the direction is not in doubt, only the pace is surprising. Advertisers are pouring into these automated lanes faster than most measurement teams can build a fence around them, and a social ads agency that waits for the platform to slow down first will simply arrive last.

Schmitt's own closing line to Raconteur ties the Reddit numbers and the Gartner forecast back together neatly: "CMOs need independent evidence that they're getting the returns they expect." That sentence is the whole brief for a social media advertising agency in five words: evidence, not expectation.

What a paid social advertising agency should demand before it trusts the dial#

This is the practical part, and it does not require walking away from automation. It requires paid social management that treats every AI recommendation as a hypothesis, then tests it before the budget moves.

The keyword that keeps surfacing in every one of these conversations is trust, and trust is precisely what a social ads agency should refuse to give a platform for free. ai powered ad campaigns are worth running, they are simply not worth running unexamined; the platform's own model is forecasting itself a bigger share of every budget it touches, and only independent measurement tells you whether that share is earned.

In practice, good paid social management is a cadence, not a one-off audit. Every AI recommendation gets logged the week it arrives, checked against a holdout or a prior period the following month, and reviewed on a quarterly rhythm alongside the rest of the account. A social media advertising agency that only checks the AI's homework once a year is grading a moving target with a stale answer key, and the platform's own model will have retrained itself twice over by the time the review lands.

Our own PPC management is built on that exact scepticism: folkfox bids like a fox, not a fire hose, spending where the scent is real and pulling back the moment a recommendation smells of the platform's own quarry rather than the client's. The paid social services work sits alongside it, treating a self-serve AI wizard's advice as one input among several, never the final word.

For any social media advertising agency choosing between a platform's dashboard and a client's actual return, the choice should not be close. The dashboard is graded by the platform. The client's return is the only scoreboard that pays folkfox's invoices, or anyone else's.

Measuring a bias the platform will never report on itself#

A social media advertising agency cannot audit what it cannot see, and no self-serve platform publishes a dashboard tile for "how often our AI recommended spending less." That absence is itself a data point. If a spend-down recommendation is rare enough to be notable when it happens, the AI is not neutral, it is house-trained.

Set that baseline before the next AI wizard prompt arrives, because a shift you cannot show is a shift no finance director will fund. A social media advertising agency that logs every platform recommendation, whether it was followed, and what independently verified lift resulted, builds the one asset a self-serve platform cannot hand a client itself: a paper trail that outlives the dashboard.

Gartner's own curve keeps climbing whether or not any single agency is ready for it: 50% of spend through AI-influenced self-serve platforms in 2025, 80% of US spend forecast by 2028. A social media advertising agency that waits for a slower year to build this discipline is waiting for a year the platforms have no reason to deliver. The agencies that build the habit now, while the share is still under a hundred percent, are the ones a client can still trust to say no.

The fox does not trust the easy path through the hedgerow just because something else laid the scent. It checks the ground first. That, in the end, is the whole brief for paid social management once the platform's own AI starts recommending how much of your budget it deserves.

Frequently asked questions#