You Are Buying Back Your Own Customers and Calling It Performance

The industry read the study, nodded, and went back to bidding on its own name.

Twelve years later, branded search is still the campaign with the best ROAS in most ad accounts. It is the line item the CFO loves, the screenshot that opens the quarterly review, the proof that performance marketing works. And a growing pile of incrementality research says a large share of that performance is a heist. Not fraud in the legal sense. Something more comfortable and more expensive: a measurement system that hands your ads credit for customers you had already won, then bills you for the privilege.

How the Heist Works

The mechanics are simple enough to explain to your CFO in one elevator ride, which is part of why nobody wants to.

A buyer already knows your company. Maybe a colleague recommended you. Maybe they read your content for six months. Maybe an AI assistant mentioned you last Tuesday. They open Google and type your name, which is the digital equivalent of walking toward your front door. Your ad appears above your own organic listing. They click the first thing they see, which is the ad, and you pay for a visit the free listing would have caught two seconds later. Then they convert, because they were always going to convert, and your attribution model hands the entire conversion to the ad that happened to be standing nearest the register when the sale rang up.

Last click attribution does not measure contribution to the journey. It measures presence at the finish line. A brand ad is always present at the finish line, because the finish line is your own name.

The platforms know this, and their measurement tools are not built to confess it. As the analysts at Recast put it when reviewing Google’s own research promoting brand ads, that research is Google telling you to spend more on Google. The referee owns one of the teams. Their attribution models overcredit brand terms because overcrediting brand terms sells more brand terms. Expecting the dashboard to expose the problem is expecting Google to grade its own homework and volunteer the failing mark.

The Numbers Nobody Puts in the Board Deck

The incrementality research on branded search is not subtle, and almost none of it makes it into budget meetings.

Growth consultancy Mammoth Growth reports working with companies where more than 80 percent of conversions attributed to brand ads would have happened without them. Picture that in dollars. A company spending one million a year on branded search and celebrating 100,000 signups is, in that scenario, paying a million dollars for roughly 20,000 signups it would not otherwise have. The other 80,000 were already walking in the door.

The pattern repeats across the measurement industry. One incrementality testing firm describes geo holdout tests where pausing branded search cut conversions by only 15 percent, meaning 85 percent of the attributed conversions were redundant spend. Search agency iQuanti, which runs these tests for large brands, puts the honest range at 20 to 50 percent incrementality for branded search, depending on how hard competitors bid on your name. Read that range again. Even at the generous end, half the conversions your dashboard credits to brand ads were never caused by them.

The gap between the number you report and the number that is true can be enormous. Measurement platform Cometly walks through the math: a campaign can show a 15 dollar CPA in your attribution dashboard while the true incremental cost per acquisition is 30 dollars, because the dashboard counts every conversion the ad touched while the honest number counts only the conversions the ad caused. Your reported efficiency can be double reality, and you would never know without testing.

Even the giant agency holding companies have started saying it out loud. WPP’s data arm Choreograph published research finding that paid brand traffic was cannibalizing organic search that would have captured the same visitors, and called on advertisers to run their own incrementality experiments rather than trust platform reporting. When the people paid to spend your media budget start publishing warnings about a channel, the warning is overdue.

And this heist has a modern accomplice. Performance Max campaigns sweep branded queries into their results by default, which means plenty of teams who believe they do not bid on brand are paying for it anyway, inside a black box that reports one blended number and never itemizes the theft.

If you want the counterfactual at scale, Airbnb ran it in public. During 2020 they slashed performance marketing, watched traffic hold, and shifted the budget toward brand. The demand did not live in the ad account. It lived in the brand, and the ad account had been charging rent on it.

Why 2026 Made the Heist Bigger

Here is where this stops being a 2013 story and becomes the exact problem sitting in your account this quarter.

I wrote recently about what happens when AI mentions your brand and then recommends a competitor for the purchase. There is a second effect of AI answers that nobody has priced in. When ChatGPT, Perplexity, or an AI Overview mentions your company, the buyer’s next move is often a branded search to verify you exist and find your site. AI does not send a click. It sends a name. That name becomes a branded query, and your brand campaign intercepts it, and your dashboard records another heroic conversion for paid search.

So branded query volume rises because of work AI did, your brand campaign takes credit for it, and the ROAS on your proudest line item inflates even further. The channel that was already overclaiming is now claiming AI’s work too. Meanwhile, zero click AI results keep shrinking the pool of generic search clicks, which makes branded clicks one of the few reliable click streams left, which gives the platforms every incentive to defend that revenue and none to help you measure it. I covered the trust side of this shift in why being visible is not the same as being trusted, and the money side is the same story. The measurement is drifting further from reality at the exact moment the spend is going up.

The One Defense That Holds, and When

Before the paid search professionals sharpen their pitchforks, the honest version includes the case for the defense, because there is one, and it is real.

The legitimate argument for brand bidding is conquesting. If competitors bid on your name, your ad defends the shelf. Skip the auction and a rival’s ad sits above your organic listing, skimming buyers at the moment of highest intent. This is why the credible incrementality range is 20 to 50 percent rather than zero, and why the answer is conditional rather than universal. One measurement firm documented the same advertiser running brand campaigns in two countries, where the campaign was genuinely incremental in the competitive market and completely redundant in the mature one. Same brand, same tactic, opposite truth, and only a test could tell them apart.

So the sin is not bidding on your own name. Companies in contested categories sometimes should. The sin is spending on it for years without ever once testing whether it does anything, then presenting the untested number as your best performance. Choreograph’s guidance matches what I tell clients: do not shut everything off blind, reduce or pause segments and measure what actually moves. The goal is not zero brand spend. The goal is knowing the true price of the insurance you are buying.

Run the Two Week Test

This is the part your agency will not suggest and your dashboard cannot show you. A brand holdout test is cheap, fast, and it converts an argument into a number. Here is the playbook.

Pick a geo holdout. Choose a few matched markets and pause all branded search campaigns there for a minimum of two weeks. Longer if your sales cycle is long. Leave comparable markets running as the control.
Clean the test first. Check Performance Max and add brand exclusions, or the black box will quietly keep bidding on your name and poison the result.
Measure total conversions in the test markets across every channel, not channel level numbers. The question is not what happened to paid search. The question is what happened to the business. Watch how much traffic the organic listing recaptures.
Compare against the control markets to strip out seasonality, then compute the honest number: spend divided by the conversions you actually lost. That is your incremental cost per acquisition, and it is the only CPA that belongs in a board deck.

Then act on what you find. If the test shows low incrementality, cap or cut brand spend and move the budget to the places that create demand instead of collecting it: generic search terms, shopping, content, and the SEO that was catching this traffic for free before the ad jumped the line. If the test shows competitors conquest you hard and the incrementality is real, keep the campaign, and now you know exactly what the defense costs. Either outcome beats the current state, which is a confident number nobody has ever checked.

The Real Lesson

Attribution is a story the platform tells you about itself, and every platform casts itself as the hero. I made the broader argument in The KPI Mirage: the metrics that look smartest in the deck are often the ones measuring the least. Branded search is the sharpest case of it in most accounts, because the number is spectacular, the CFO loves it, and the entire thing rests on an assumption nobody tested, that the ad caused the customer instead of billing for one who had already decided.

Your brand built that demand. Your content, your reputation, your product, your years of showing up. The campaign standing at the finish line did not win those customers. It invoiced you for them. Run the test and find out how big the invoice has been.

For more on marketing measurement that survives contact with reality, visit the Transmyt blog.