For years, the industry has treated somebody typing your company name into Google as the most obvious signal in advertising. They searched for you because they want you, so get out of the way, take the cheap click, and move on. A few weeks ago, I wrote about how AI Overviews are changing what happens on a results page before anyone clicks anything: more back-scrolling, more comparing, more reconsidering, across every kind of search. Branded search is where I’d test that first, because it’s the query type everyone assumed was too obvious to be affected.
Turns out, it isn’t.
Somebody typed your name. They didn’t leave.
Clickstream data no longer supports the “get out of the way” part of the sequence. What’s happening now is that somebody types your company name, and twenty-one seconds later, longer than it sounds, they are still on Google’s page rather than on your website: 45.8% of people searching a brand name with an AI Overview present are still active at that mark, against 12% when no AI Overview appears. Their cursor covers more of the page instead of tracking toward one particular link, which makes up about 40% for those searchers. AND THEN they scroll back up to re-examine something they’d already passed nearly twice as often, from 23% of their scrolling to 44%.
The data shows people lingering, but it doesn’t explain why. A preregistered field experiment by Stephanie Wang and her co-authors doesn’t answer that either, but it does point us in a direction: given a full AI Mode experience instead of ordinary results, people ran longer sessions while click-through to publishers fell, and they rated both the experience and their trust in it lower. If you read that the way I did, you probably laughed, because the short version is that Google kept more clicks, and people enjoyed it less 😅. Which brings me back to this piece: the branded result page is no longer somewhere you can assume you’re being handed the click.
Should you bid on your own brand? Cheap performance alone cannot answer it.
Before I get into numbers, I want to separate two things I caught myself blurring together while I worked through this: how my own accounts compare to the industry’s benchmark panel, and how brand compares to non-brand inside each of our own accounts. They’re two entirely different questions, so I’ll label them accordingly.
What branded search benchmarks are showing
Optmyzr’s Q2 2026 benchmark report, across more than 20,000 accounts and 250,000 campaigns on a matched panel, puts median branded CPA at $42.59 against non-branded at $41.67, branded now marginally higher. Sam Tomlinson, quoted in the report, put words to what that undoes: “For 20+ years of PPC, non-branded CPAs ran substantially higher than branded, which is why advertisers defended brand so heavily.” Branded CPC rose 5.21% year over year against 3.18% for non-branded, the median year-over-year change in each account, the same convention the whole report uses rather than a blended total. Branded CTR also runs well ahead of non-branded on the panel, 9.15% against 5.77%.
Two things are worth flagging about this dataset: It was pulled about fifteen days after the quarter closed, and conversion metrics understate at that distance, by Optmyzr’s own admission. And the branded classification is a campaign-name rule, which can catch hybrid formats like Performance Max.
What happened in our own Google Ads accounts
I looked at some of the Search-only accounts we manage, each with an explicit own-brand campaign-name designation, excluding Performance Max, Shopping, Display, and manufacturer or product-brand campaigns. Our data ran through August 20, about seven weeks past quarter close, so conversions had more time to settle than the panel’s fifteen-day window.
Inside that set, brand still won on cost across the board. Branded CPC was cheaper than non-branded in all accounts, and branded CPA was lower in 85% of the accounts where both sides converted. I also compared year over year; the median change in branded CPC was about +12.4%, against +24.0% for non-branded, the same per-account-median convention as the Optmyzr numbers above.
Branded CTR told the same story, more loudly. The panel puts branded CTR at 9.15% versus 5.77% for non-brand. Across our Search-only brand campaigns, branded CTR ranged from 15.6% to 46.2%.
As you can see, it’s quite the deviation from the panel on cost and on attention, but we also have to acknowledge that I’m comparing apples to oranges. The gap between the data volume I worked with and the 20,000-account cross-industry benchmark is pretty huge, so naturally, the comparison results will be too. What it tells me is that when campaigns are cleanly isolated and given time to convert, branded traffic stayed cheap and easy to click, in our own book of work… but, naturally, I have many more questions.
Cheap branded clicks are not incremental conversions
Evaluating branded search through CPC or CPA, or even CTR, conflates three different things into one word, “efficient”: cheap clicks, easy attention, and conversions that mean something. Once I looked at conversion rate specifically, brand and non-brand stopped moving together.
The Optmyzr panel shows branded CVR consistently ahead of non-branded, 5.10% against 3.91%. Across our accounts, it’s a coin flip: branded CVR beat non-branded in half of the accounts and lost in the other half, ranging from around 49% down to about 0.16%. A brand search can come from an existing customer looking for a login page, someone seeking support, or someone still researching rather than buying. That doesn’t automatically make the click worthless, especially if competitors are sitting on the query, but it does mean a high CTR and cheap CPC aren’t evidence of incremental demand.
These aren’t normalized businesses or conversion actions, so the raw CVR range isn’t a performance ranking across accounts. That’s exactly why I care more about the pattern inside each account than whether one account’s 49% is “better” than another’s 0.16%.
The click still behaves like brand, but the conversion doesn’t, not consistently. This now means the two things: brand efficiency used to bundle together, cheap attention and real conversions, have come apart in half of what we manage.
“It’s brand, it’s cheap” used to end the conversation. It’s now the start of one.
When bidding on your own brand actually becomes defence
If high branded CTR and low CPCs don’t guarantee incremental conversions, why defend brand keywords at all?
It’s not because the clicks are cheap; it’s because of the conditional risk against competitors actively poaching your traffic.
I ran Referral Rock‘s brand terms through Bluepear, a brand-search monitoring tool, in April, and found five competitors advertising against them. Not on “referral software,” which is fair game and always has been, but on “Referral Rock reviews,” “Referral Rock free trial,” “Referral Rock demo,” and “Referral Rock login,” where the person typing it is already a customer trying to get into an account they pay for. When a brand owner leaves their own name undefended, the clearest signal in search sits there unclaimed.
Peer-reviewed evidence backs up both halves. Where competitors weren’t bidding, nine field experiments across eight small DTC brands and eBay’s own large-scale field experiment both found no statistically significant sales loss from turning branded search off. Organic absorbed the traffic, and the conversions the ads were credited with would likely have happened anyway, a straightforward selection problem. Where competitors were bidding, large-scale Bing auction experiments found rivals captured 18% to 42% of traffic when the brand owner didn’t defend, against only a few percent of clicks when no one was bidding at all.
Brand defence isn’t a universal revenue engine, and it isn’t dead weight either. It pays when someone is actually taking the traffic; it doesn’t pay because the click was cheap.
How to test whether branded search is incremental
Here’s what that actually looks like, using the same method Rand Fishkin walked through at SparkToro’s Zero Click Summer School, applied to branded search specifically.
Benchmark first. Know your branded account’s historical CTR, CPC, CPA, conversion rate, and qualified-lead rate, broken out the way I did above, not pooled across accounts that don’t share a business model.
Then split into test cells before you touch spend, not after: one where a competitor is actually bidding on your brand terms, one where none is, since that’s the condition that decides whether defence has anything to defend against. If you run across markets with different AI Overview prevalence, that’s a third useful split. The intervention below gets applied inside each cell separately, so you can see whether turning brand down costs you differently depending on what’s actually sitting on that page.
Reduce coverage, not just the budget. The point of the test is to meaningfully reduce how often your paid brand ad appears, so changing the budget only counts as an intervention if it actually does that. If the campaign wasn’t budget-constrained to begin with, cutting the daily budget by 50% can leave coverage almost unchanged. Whatever lever you use, verify that paid exposure actually moved before treating it as a test.
I also wouldn’t pause the campaign simply to make the experiment cleaner. Reactivating a Smart Bidding strategy can put it back into a learning period while Google recalibrates, although that doesn’t mean everything it previously learned disappears; historical conversion data can help it adjust faster. A controlled reduction lets you test what happens when paid brand coverage falls without adding a full stop-and-restart as another variable you then have to explain.
Track past the platform’s own conversion tags. Qualified conversion rate, not just conversion rate, filters out the logins and support traffic the Referral Rock example shows up in. Where you can, track total and new-customer revenue directly from your CRM or billing, since that’s the number that actually tells you whether a click was defended or just captured.
What a flat result actually tells you
Allow for the lag Google’s own guidance describes: conversion delays that run days to weeks. Don’t call the test on day seven if your sales cycle runs longer.
And be honest about what a flat result means. If revenue drops in the contested cell, that’s brand defence earning its keep. If revenue holds in the uncontested cell, that’s non-incremental capture; organic was already going to take the click. If revenue holds in the contested cell too, that’s suggestive, not proof: a short window or a small account can fail to detect a real effect just as easily as it can confirm there wasn’t one, the same caution the peer-reviewed studies above carry with them.
The old assumption
For twenty years, “it’s brand, it’s cheap” ended the conversation; It’s now the start of one. Cheap was never proof of anything on its own, and in half of our accounts, the conversion-rate edge that used to make cheap feel earned isn’t showing up either. What’s still true is that a brand search is the most valuable intent signal in the account, and it’s now landing on a page that no longer hands you the click for free, with competitors circling underneath. I don’t think that applies to every account. It adds up to a reason to test your own, as I laid out above, because the Optmyzr panel and I both had to hedge on this one, and you might not have to.