The Accountability Sink in AI Advertising

Meta announced this week that they plan to use AI to automate the entire advertising workflow, from ad creation to placement, by the end of 2026. The pitch to brands is that Meta can handle the entire advertising process from creation, iteration, customization, testing and placement of ads using AI and do it cheaper, faster and more completely than people. Meta, along with Microsoft, Google and, now, Amazon and TikTok, are among if not the largest platforms online with, by extension, the largest advertising machines in the world, reaping in billions of dollars from brands spending on their platforms. Meta and the companies above have greater potential user than any other single website or network of sites online. That claim will stand so long as the open web continues to close down and the experience of being online involves being engaged on few select platforms.

Money, market share and to flex their AI capabilities are the biggest reasons for Meta’s announcement. Of the billions of dollars being spent on online advertising, millions of those dollars are dollars not being spent on Meta ads. That is money going going to two different places: other ad platforms and to pay middle men, freelancers or digital marketing agencies, to manage Meta ads on behalf of brands. With this announcement Meta is signaling it’s planning to after the latter with this announcement and hoping to gain more of the former along the way.

Operators, the media buyers, digital marketing professionals, online advertising professionals, general marketing professionals and others that have different titles but perform similar duties, play an important role to brands when comparing them to giving Meta, or Google or any other major platform with an advertising arm and a walled audience, total control over a brand’s online advertising. Operators offer accountability to brands. If you, the operator, make a mistake, you have to be accountable to that mistake. The brand is the check in place to hold operators accountable for their actions, decisions and the results of them. If Meta’s AI is the operator, and Meta “makes a mistake”, there’s no one to hold accountable. Meta can, and likely will, ignore you and then point to it’s AI and shrug. AI, for brands looking to advertise on Meta and, broadly, any brand looking to advertise on any platform’s properties that professes to use AI as their means to serve ads and, now, create them, creates an accountability sink.

The accountability sink deepens when you remember that Meta has different incentives than the businesses that use Meta ads and the people that manage Meta ads for brands have. As anyone whose interacted with Google sales reps can testify, getting brands to spend more and adopt new features are the two measures of success. Both, often, are goals that are not aligned to a brand’s goals. Brands will use online ads as a means to an end. The ends being, among others but usually, getting new prospective customers to do a thing (buy, signup, enroll, visit) for the lowest cost possible. The math needs to work in order to spend more. By not spending wildly or on blind faith like trust our AI isn’t smart business sense - unless what you think you’re looking at is a reason to spend more.

Meta Ads data has always been kind of off. Anyone whose used the platform can confirm that simply by looking at the provided dashboard or doing any compare-and-contrast with another reporting tool. Independent researchers have noticed these inconsistencies, too. Studies have been conducted while studying then-Facebook's advertising ecosystem was still possible and found that Meta had overstated, inflated and distorted Reach metrics, offered misleading, duplicative or outright overly generous conversion attribution modeling, and provided a litany of other misleading advertiser-facing and non-advertiser facing inaccurate metrics. Though there hasn’t been formal research performed, anyone that has even tried to setup Meta Ads’ setting up Meta Ads conversion tracking properly can testify that the setup is rife with pitfalls and opportunities to think you’re setting up tracking correctly and therefore think the performance metrics are correct. Anyone whose read Careless People will also know that Meta’s upper brass prioritizes initiatives that drive user growth, market share and profit over any other initiative eons ahead of things they won’t fix it because they don’t need to fix it - a pattern Google fits, too. It’s reasonable to think these issues above still exist and were never resolved.

The accountability deepens again when AI is introduced. AI helps Meta to smooth over and absolve them of any data, tracking, creative choices, ad placement choices, reporting discrepancies, whether you’re reaching a real person or not, modeling or accuracy. It also helps to bury the reality that it’s simply harder to reliably track and reach the people brands would like to reach and only like to reach with their ads. If you’ve noticed aggressive encouragement to try different audience expansion tools from Meta, LinkedIn Ads, Google Ads or programmatic services, it’s a legitimized, official-looking way of saying we don’t know how to find or can’t reliably see exactly these people, but the computer might. Please pay to find out.

So long as walled platforms can’t be audited on behalf of brands to ensure that their ads are winding up on real placements in front of real people (and the auditors themselves aren’t crooked), the internet continues to consolidate around fewer and fewer platforms, people have goods and services to sell and believe the most effective way of reaching prospective customers is advertising, Meta along with Google, TikTok and others, don’t have a reason to clean up their platforms. Brands have no choice but to deal with the hand they’re dealt.

Bringing this back exclusively to Meta though the same could be said of Google, Microsoft and TikTok, if nobody, including Meta, knows how their AI works, reported performance metrics are kind of right on the high end and unreliable at their worst with few ways to validate or invalidate their accuracy, brands know less about to who, where and in what way their ads are serving, there’s little to no moderation of the quality of placements in or outside of the platform, and now the ads themselves could be created outright from just a few inputs, where does that leave us - the industry, Meta, brands and operators? Does knowing and controlling these things matter? After all, if AI gets the Moore’s Law treatment, the cost to run resource-intensive computing will only get cheaper and more accessible and maybe giving control over to platforms will make the most sense in the future - short, medium or long term. There’s enough thought pieces speculating about the future so I’ll stop myself here.

But for now, for the industry and Meta and those like it, their pitch to brands is it all but says without saying we don’t know anything and have no control. Anything that goes wrong is the AI’s fault. But we’ll take your money so you can find out if this works for you or not. And oh, do you actually care where you get your performance from? Or do you care that you got it? And some brands, almost exclusively e-commerce brands (that’s an article for another day), seem to see success with letting Meta run most of the show for them. For brands, it leaves no choice but play ball and wait and see alternative to the online advertising model emerges or AI achieves a state that it’s a mixture of transparent and reliable. For operators, there’s the magical promise of having all kinds of new free time to do “other things” instead of toiling with the ad machines. Their value will and always will be judgement. They will be the ones making the calls if any of this is working. They likely won’t have any control to make it better except for altering a few inputs, but it is their judgement and attention to detail that will help keep accountability sinks in advertising in check.

Credit for thumbnail:

https://www.analyticsvidhya.com/